Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb Visible Photon imaging and 3D IT R. BARBIER IPNL, IN2P3 - Lyon University Outline: Introductory remarks on the expected vertical integration benefits for Imaging Sensors Trends in Photon Imaging for fluorescence microscopy Single Photon Imaging Sensors EMCCD and EBCMOS Back thinned CMOS for imaging and vertical integration

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 2 Introductory remarks Vertical integration will offer new possibilities for designing imaging sensors At the level of the system integration : 1. 4-edge buttable => large sensitive surface (interesting for X-ray-gamma) 2. backside access to the sensor for readout : increase the fill factor New integration solution At the level of the sensor design : 1- sensitive substrate can be selected and optimized separately : etching - post- processing - diode optimization for Noise and CCE. 2- Each layer (Analogue and digital) can be optimized separately (different techn.) 3- Pixel by pixel parallel access gives more interesting but more complicated readout strategies : sparsification - clustering - time stamping dynamics photon / photo-electron But the readout strategy has to be optimized for the application and it depends on the hit density

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 3 Applications : Trends in Fluorescence Microscopy Fluorescence Microscopy is a very promising field of instrumentation the field is moving very fast with new techniques : The molecule under study are tagged by fluorescent dye –Excitation with a laser and emission in another wave length –Proteins : Green Fluorescent Proteins (genetically modified drosophilia) –Qdots : CdSe(ZnS) In cellular and molecular biology the main goal is to understand molecular and cellular interactions : - property of diffusion and transport, interaction, immobilization, information about structure, mobility, elasticity Methods to observe and to quantify : Fluorescence Life Time Imaging : FLIM Fluorescence fluctuation Microscopy : FCS Single Particle Tracking : PALM Each methods has his own optical systems Confocal Spinning Disk (multi pinholes) Total internal Reflection Microscopy Selective Plane Illumination Microscopy … Synaptic activity with Qdots tracking M. Dahan CdSe(ZnS) with TIRF

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 4 Trends in Fluorescence Microscopy Reduce photo-toxicity and photo-bleaching means reduce excitation power of the laser therefore the sensitivity has to be increase. Observe dynamics : Millisecond time scale Combine variables of observation : Lambda, time lapse, 3D Increase resolution : To go beyond the diffraction limit (resolution < 50 nm with PALM) Increase the number of functionalities : Windowing - averaging First conclusion Imaging techniques needs Imaging Sensor with FASTER READOUT MORE SENSITIVITY SMARTER But Faster readout without increasing the readout noise Better resolution with few photon Single photon sensitivity keep dynamics for counting => ADC What could we expect from 3D IT to work with these trade offs ? PALMClassical TIRF

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 5 Single photon sensitive pixel detector Two ways Vacuum Tube100% Silicon CCD Technology Impact ionisation : EMCCD (E2V Tech. / Texas In.) Avalanche (Geiger mode) pnCCD (MPI) Photocathode + High Voltage +silicon ICCD-EBCCD-EBCMOS-EBMCP-HPD… Back thinned CMOS/CCD Photocathode : S20-GaAs Quartz window EBCMOS / EBCCD R3R1 dcR2R3 R1 V Impact ionisation Electron Multipling CCD principle Pixels Storage Shift registerGain register Output Amplifier G=(1+P) 512 =1000 P=1.3% - The readout Noise is no more a problem - The remaining noise comes from the electron emission from the cathode

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 6 State of the art in “photon counting” EMCCD technology Background events Clock Induced Charge (CIC) effect: Noise T=-85°C - Fake rate = 0.5 % per frame For 30 ms integration time -> fake rate = 647 Hz/mm2 The background events increase with frame rate ! Obtained with a long exposure time > 1s For each readout True Photon counting ? No Maximum frame rate for full frame (1024x1024) 35 fps « Solutions » : Binning 2x2; 4x4 = decrease resolution and Max is 130 fps Read few lines = smaller array size but decrease the FOV Frame Rate : limited Statistical fluctuation of the gain Counting from 1 to 10 ph. is impossible with an EMCCD ! Statistical fluctuation in the gain register !!

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 7 EBCMOS : Principle of detection Visible photon from the Microscope E Photo-électron photocathode HV Back-thinned CMOS carrier Few mm Few cm 2 The backthinned MAPS is a photoelectron tracker Single Photon(ph.e.) Sensitivity Spatial Resolution; with proximity focussing Gain is given by High Voltage CMOS etching down to epi layer 15 microns Post processing of the back thinned (dead layer < 60 nm) Large Scale Mega-pixels CMOS Pitch: 5-20  m => 1-4 cm 2 FOV Ultra-fast Clock: MHz signal processing : CDS - Sparsification and/or ADC

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 8 Back thinned MIMOSA5 Single photoelectron sensitivity of the back thinned MIMOSA 5 Example of Photon Spot tracking with back thinned MIMOSA5

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 9 Single photo-electron tracking : demonstrator EBMIMOSA5 EBMIMOSA5 demonstrator : tests in Microscope facility: Dark count 100 Hz/mm ph.e. / 27 ms / 2.5 cm 2 = ms integration time Sensitive surface : 2.5 cm 2 / 18mm cathode diameter Frame Rate : 2 fps (usb2 board) fps are possible in principle with processing on Board We use this demonstrator to investigate on microscope facility the pros and the cons of ebcmos devices and we will determine the requirements for the next prototypes LUSIPHER PROJECT Large-scale Ultra-fast SIngle PHoto-electron trackER: Vacuum Tube: – S20 multi-alkali cathode – HV < 2.5 kV – GAP < 1 mm Chip Designed in July 2007 by W Dulinski Techno 0.25 micron The chip is alived and is going to be etched and post processed 10 um pitch and medium scale chip: 400x800 pixels; 8 // analog outputs; 40 MHz clock frequency With sparsification on Board we should readout 1000 frames per second.

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 10 Point Spread Function (PSF) Spatial resolution with single photo-electron measurement: –The spot is focused on the center of one pixel –T = 20°C HV = 8 kV  8 kV 1D Resolution Fit Dark count due to photo-electron emission of the cathode : HV = 6 kV, T = 10 °C 650 ph.e. / 27 ms / 2.5cm 2 => 100 Hz/mm 2 One photo-electron : Pixel seed SNR ~ 20 cluster 3x3/5x5 Integrated Image 2 micron spot focussed on the cathode plane Binary output no clustering algorithm

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 11 A true photon counting Multi-photon spectrum: Photo-electron peaks: Poisson distribution 5x5 countour ph.e Reconstructed. energy and position of the 3 photo-electrons (off line algorithm) If we want true photon counting we need ADC (3-4 bits)

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 12 Tracking for Imaging Sensor Tracking for Vertex detector is a spatial tracking Layer 0 Layer 1 Layer 2 MIP Z X Y Frame n Frame n+1 Frame n+2 Time Lapse X Y Online/Offline High hit density in Layer 0 Fluorescent dye Dark count static dynamics Tracking for Imaging sensor is a temporal tracking Noise is randomly distributed Signal is repeated in time Noise or wrong hit Hit from the track

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 13 3D IT approach Analog Control Memory Digital pixels Digital layer : counter and/or pixel memories : noise supression Reduction of the data flow Idea : keep the memory of the signal hits into a sliding temporal window of 2 or 3 frames Memory 0 Memory 1 Memory 2 X Y Fluorescent dye static dynamic Noise is randomly distributed Signal is repeated in time

Vertical Integration Technologies for HEP and Imaging Sensors - Ringberg Castle 6-9April 2008 / rb 14 Analog What could we expect from 3D IT for EBCMOS ? Digital Diode with good CCE minimization of the cross talk Etched and Post process layer Preamplifier - shaper - CDS Discriminator at one photo-electron ADC (3-4 bits) for counting and clustering Memories : integration mode on few frame (temporal tracking) will be very interesting Readout decision Carrier 3D IT bring pixel-by-pixel parallel access pixels Bonding issue and integration into the vacuum tube : Vertical Integration will simplify the tube processing and back thinned CMOS integration into the carrier