1. 3D sensors for tracking detectors: present and future applications C. Gemme (INFN Genova) Vertex 2013, Lake Starnberg, Germany, 16-20 September 2013 Outline: 3D tracking detectors A real case: IBL AFP CMS 1

2. 3D detectors for tracking detectors 3D technology is very attractive for future tracking detectors. Intrinsically radiation hard and able to sustain large fluences Low depletion voltage and power consumption Good resolution and time response Fabrication process is complex, bringing to low yield and high cost. In this talk I want to review the current status of the possibility to built a full-scale detector for future tracking system. Experience of IBL Close Future and perspectives 18/09/2013C. Gemme- 3D Tracking detectors2

3. IBL: 3D- oriented Overview The IBL (Insertable B-Layer) is the first upgrade of the ATLAS Pixel detector and it is going to be installed in LS1. See Fabian Huegging’s talk for details. Sensors (planar and 3D) matched to new rad-hard FE-I4 electronics. First “large” scale 3D tracking system, composed of eight 3D modules at high eta on 14 staves surrounding the beam pipe at ~ 3 cm from the interaction point. Each 3D sensor matches a FEI4 read- out chip, 2x2 cm 2 size. Overall 112 3D modules on detectors. 18/09/2013C. Gemme- 3D Tracking detectors3

4. Two vendors: CNM and FBK 230um thick 3D n in p-type sensors. Rad-hard to 5 10 15 n eq /cm 2 NIEL 4 IBL: CNM and FBK Sensors FBK CNM

5. Requirements 18/09/2013C. Gemme- 3D Tracking detectors5 Main requirements: Slim-edge: Dead region in z < 225 um Fully Irradiated 3D CNM JINST 7 (2012) P11010

6. Requirements 18/09/2013C. Gemme- 3D Tracking detectors6 Inefficiency at columns (not sensitive areas) Overall efficiency 97.5% (99%) at 0 0 (15 0 ) 15 0 corresponds to IBL tilt angle. Main requirements: Slim-edge: Dead region in z < 225 um End of lifetime Efficiency >97% Fully Irradiated 3D CNM JINST 7 (2012) P11010

7. Requirements 18/09/2013C. Gemme- 3D Tracking detectors7 Main requirements: Slim-edge: Dead region in z < 225 um End of lifetime Efficiency >97% End of lifetime Depletion voltage lower than 200 V and good charge collection. Fully Irradiated 3D FBK JINST 7 (2012) P11010

8. IBL: Summary of Production 3D 18/09/2013C. Gemme- 3D Tracking detectors8 ProducerWafers Produced & Tested Selected WafersUBM (by IZM) Wafers CNM514140 FBK70 (??)33 (2 Broken??)31 ProducerUBM (by IZM) Wafers Selected Good tilesDelivered Modules CNM40228167 * FBK31137114 * Preliminary as the Module production is still ongoing and will be completed in ~ one month. Wafers are selected for UBM if at least 3 tiles out of 8 are good A tile is good if V bk >25 V

9. Modules Qualification Modules are assembled in Genova and Bonn: Gluing of a flex hybrid Wirebonding Assembly test at 15°C  ASSY Burn in: at least 10x (-40°C,+40°C) Final Qualification at -10°C  FLEX Qualification and Ranking 18/09/2013C. Gemme- 3D Tracking detectors9

10. Breakdown voltage studies - I In the first batches, we have verified that CNM has low yield (~65%) after assembly due to V bk < 25 V or high leakage current. 18/09/2013C. Gemme- 3D Tracking detectors10 FBK CNM

11. Breakdown voltage studies - I In the first batches, we have verified that CNM has low yield (~65%) after assembly due to V bk < 25 V or high leakage current. The main reason is in the wafer measurement done by CNM to select the good tiles that is limited to the guard ring: Based on preproduction, expected 10-20% bad at the device level. 18/09/2013C. Gemme- 3D Tracking detectors11 FBKCNM Temporary column short Guard Ring FBK CNM

12. Breakdown voltage studies - II Since then, added a IV test after IZM delivery and assembled only the good ones at module level. CNM sensors not flip-chipped yet back to CNM for re-test after UBM  Found 50/72 (69%) of good tiles. Tested also the red tiles and recover many. Breakdown voltage distribution for the assembled devices results rather different for CNM and FBK. FBKCNM

13. Noise studies The electronic noise depends on Vbias, temperature and sensor type. Typical values are 140e- (FBK) and 120e- (CNM) at -10C, 3ke threshold and V bias = 20 V. SE POSSO AGGIUNGO I VALORI MEDI DI NOISE CALDO/FREDDO e LE DISTRIBUZIOni in SPARE The difference between noise with and without bias is used to identify disconnected bumps. A cut of 20e is very efficient at RT, while at low temperature detects 10% fakes for FBK. Therefore a source scan is used. 18/09/2013C. Gemme- 3D Tracking detectors13 CNM

14. Geometrical characteristics In FBK the edge is active and the occupancy in the edge pixels (rows and columns) is larger than for the ‘normal’ pixels. 14 CNM FBK 241 Am scans in self-triggering mode. In CNM the guard ring stops the charge collection at the edges and the occupancy is therefore the same as the internal pixels.

15. Cross-talk The test consists in injecting the maximum charge (~50 ke-) in two neighbouring pixels and measure if any hit is observed. The test is run routinely at 3 ke- threshold and usually the cross-talk is lower than ~3%. Some crosstalk has been observed in few CNM devices. Not relevant for data-taking performance. 18/09/2013C. Gemme- 3D Tracking detectors15 Injection Read-out Injection 50 x 250  m 2 Occupancy map Crosstalk pixels vs inj. charge

16. 3D yields Bump bonding problems in the first batches. Then failure rate at the assembly stage is ~28% (FBK) and ~39% (CNM). 18/09/2013C. Gemme- 3D Tracking detectors16 FBK CNM

17. AFP ATLAS intends to install a Forward Physics detector (AFP) in order to identify diffracted protons at ≈210 m from the interaction point in 2018. AFP combines a high resolution pixel tracking detector with a timing detector for the removal of pile-up protons. The tracking detector will consist of an array of six pixel sensors placed at 2–3 mm from the LHC proton beam. The proximity to the beam is essential for the AFP physics program as it directly increases the sensitivity of the experiment. Two critical requirements for the Pixel detector: The device has to cope with a very inhomogeneous radiation distribution from ~5 10 15 1 Mev neutrons to several orders of magnitude less. The active area of the detector has to be as close as possible to the LHC beam, which means that the dead region of the sensor has to be minimized. 18/09/2013C. Gemme- 3D Tracking detectors17

18. AFP: Inhomogeneous radiation Non-uniform irradiation of CNM 3D IBL devices done in IRRAD1 at CERN-PS. Good leakage after non-uniform irradiation. Non uniformly irradiated device performance evaluated at CERN ATLAS 120 GeV pion test-beam. Efficiency: 98.0% (irradiated side) by masking out dead and noisy pixel cells (due to front-end issues). 18/09/2013C. Gemme- 3D Tracking detectors18 http://dx.doi.org/10.1016/j.nima.2013.03.064 (a) (b)  = 98.9%  = 92.7% Efficiency folded into a 2 by 2 pixel area Threshold: 1700 e - Bias voltage: 130 V Temperature: -20 C

19. AFP: Slim edge 4 devices (2 CNM and 2 FBK) with ~100 um slim edge tested AFP_FBK_S1_R9  efficiency ~ 98.3% Need to do edge efficiency analysis. 18/09/2013C. Gemme- 3D Tracking detectors19  = 98.9%  ~ 98.3% AFP_CNM_S3_R5 CNM FBK 100 um Tight dicing in the bias-tab side

20. CMS- I 18/09/2013C. Gemme- 3D Tracking detectors20 Inter-electrode distance: 90 μm (1E), 62.5 μm (2E), 45 μm (4E) Several devices assembled using with the existing CMS pixel readout chip (ROC) type PSI46v2, which has an array of 80 rows x 52 columns of 100 um x 150 um readout pixels or PSI46dig

21. CMS-II 18/09/2013C. Gemme- 3D Tracking detectors21

22. CMS-III 18/09/2013C. Gemme- 3D Tracking detectors22

23. Conclusions 18/09/2013C. Gemme- 3D Tracking detectors23

24. Noise distribution 18/09/2013C. Gemme- 3D Tracking detectors24

25. Charge collection in 3D 18/09/2013C. Gemme- 3D Tracking detectors25

26. Xtalk vs Bias: 2ke threshold 26

27. Xtalk vs Threshold 27 1ke 1.5ke 2ke 3ke 4ke VCAL = 900