CMS Tracker Upgrade for HL-LHC Sensors R&D Hadi Behnamian, IPM On behalf of CMS Tracker Collaboration
Outline HL-LHC Tracker Upgrade: Motivations and requirements Silicon strip R&D: * Materials with Multi-Geometric aspects * Bias voltage * Annealing and temperature * Neutron, mixed irradiation with different fluences Investigation of charge collection Noise Study Device simulation of electric field Conclusion 2
Tracker Upgrade New Tracker Layout Need to adapt to HL-LHC radiation level, in particular silicon sensors. For 3000 at r = 22 cm: fluence = PS modules 2S modules Outer Tracker Readout Chip 2.5mm modules are located at 20<r<60cm (max ), 4mm modules are located at r>60cm (max ) Same senor technology for PS & 2S modules CMS Binary Chip (CBC) in 130nm CMOS Binary unsparsified readout The CBC threshold is expected to be between 4k and 6k electrons, thus we consider a 8k electron signal to be acceptable 3
Silicon Strip R&D Wafers have been produced by Hamamatsu Material Types: Float Zone (FZ): 320, 200, 120 μm Magnetic Czochralski (MCZ): 200 μm Epitaxial silicon: 100, 70, 50 μm Measurements are token before and after Irradiations (with 23MeV protons@kit, 23GeV protons@cern and reactor neutrons@jsi, Ljubljana) MSSD: multi-geometry strip structure p-in-n is known as n-bulk, n-in-p is known as p-bulk (with p-stop and p-spray strip isolation) 4
Charge Collection, n-bulk Charge collection measured with Sr90 at -20 C after different irradiations. Float Zone ( 300um)-no annealing Measurement comparison between ATLAS and CMS 600V 600V 900V 900V 600V 20500 18500 CMS measures higher signal - sensors with different thickness - sensors from different vendors - ATLAS results with only neutron irradiations while CMS used mixed ones Cluster Charge (e) 16500 CMS ATLAS CMS ATLAS CMS 2W@RT 14500 12500 10500 8500 6500 4500 2500 1E+14 1E+15 1E+16 Fluence (neq /cm 2) From these measurements: Generally 320μm n-bulk strip sensors could still be used up to ATLAS/Liverpool2: PoS(VERTEX 2008)036 5
Charge Collection, p-bulk ATLAS and CMS results show reasonable agreement Float Zone ( 300um)-no annealing 22500 p-bulk strip sensors show uniform drop trend The threshold is expected to be between 4k and 6k electrons, thus we consider a 8k electron signal to be acceptable 600V ATLAS(p) 600V ATLAS(n) 18500 16500 Cluster Charge (e) 600V CMS 20500 14500 12500 10500 8500 6500 4500 Signal in p-bulk sensors is higher than 8ke even at high irradiation fluences > 2500 1E+14 1E+15 1E+16 2 Fluence (n eq / cm ) ATLAS/Liverpool: NIM A 636 (2011) S56-S61 6
Charge Collection Comparison Bias voltage=600v, no annealing, p&n-bulks 200μm For fluence < - n-bulk sensors have higher seed signal than p-bulk then n-bulk strip sensors could still be used in this range of fluence For fluence > - n-bulk with low seed signal - p-bulk signal is higher than 8ke Bias voltage=600v, no annealing, p-bulk Cluster Charge (e) 23000 14000 Float Zone n-bulk Magnetic Czochralski n-bulk Float Zone p-bulk Magnetic Czochralski p-bulk 13000 Seed Signal (e) 12000 11000 10000 9000 8000 Float Zone 320um 7000 Float Zone 200um 6000 0.0E+00 18000 5.0E+14 1.0E+15 1.5E+15 2.0E+15 2 Fluence (neq /cm ) 13000 200μm strip sensors allow to reduce the material budget and keep charge collection at same level as 320μm at high fluence. 8000 3000 0.0E+00 5.0E+14 1.0E+15 1.5E+15 2.0E+15 2 Fluence(neq /cm ) At fluence > signals from 320μm and 200μm are comparable 7
Annealing Behavior at Fixed Fluence Study annealing functionality of seed signal at fixed fluences for different material and thickness. At fluence: Signal is more stable for p-bulk sensor 12000 All thin p-type samples work well and show 11000 seed signals >8ke- at 600V until above 20w@RT. 10000 16000 14000 Seed Signal (e) At fluence: FZ200N FZ320N FZ200P FZ320P Seed Signal (e) 12000 9000 8000 7000 6000 5000 4000 5E+1 10000 8000 5E+2 Annealing (h@rt) 5E+3 20 Weeks @RT 6000 4000 1E+2 FZ200P MCZ200P FZ200N 1E+3 Annealing (h@rt) 1E+4 20 Weeks @RT Signal on n-bulk sensor decreases after a few weeks @ RT to a low level. Magnetic Czochralski material show an increased signal with annealing 8
Noise Study Multi-geometry strip structure 12 11 10 9 8 7 70 80 240 120 70 80 W/P=0.33 6 240 5 4 3 2 120 70 80 240 W/P=0.23 1 120 W/P=0.13 Pedestal run with no particle passing the sensors at different voltage. RMS of Gaussian distribution=noise p=70 μm p=80 μm p=120 μm p=240 μm Float Zone 200μm n-bulk W/p=0.13 W/p=0.23 W/p=0.33 Float Zone 200μm p-bulk Generally, n-type Noise is higher Regions with higher pitch show a large noise at low bias voltage 9
Noise Random Ghost Hits (RGH) CMS has observed noise with Gaussian+large tail distribution for irradiated n-bulk sensors by scaning phase space (fluence, bias voltage, annealing). A RGH rate was defined by counting all signals >5σ and dividing by #strips and #events with two methods: Method A RGH=A +/Strips/Events RGH, can produce fake signals and increase occupancy for n-bulk materials. 10
RGH - p&n-bulks comparison Method A Float Zone 200μm p-bulk Float Zone 200μm n-bulk No effect was appeared in p-bulk sensors even after mixed irradiation. Simulation on surface and bulk damages can be help us to study this effect with more details. 11
RGH - Different irradiation types for Method A Float Zone 200μm n-bulk Compare only-neutron with mixed irradiation On ly N eut r on Effect is much reduced for neutron only irradiation This dependence on ionizing radiation hints towards a combined effect of bulk damage and surface charge. 12
RGH Temperature Dependency Method A Fix bias voltage and study temperature behavior of RGH at defined irradiation and annealing for different materials. For n-bulk material, there is correlation between leakage current and fake hits with temperature. 13
Random Ghost Hits Method B A- A+ RGH=(A +-A-)/Strips/Events Subtract the pick-up noise with this method 14
RGH @ Different Geometries by using Method B Multi-geometry strip structure 10 70 80 240 9 8 7 120 70 80 6 240 5 4 3 2 120 70 80 240 1 120 70 μm 80 μm 120 μm W/P=0.33 W/P=0.23 W/P=0.13 We measure the voltage at which the RGH appear (called turn-on voltage) We don't observe RGH effect in p-bulk sensors with 200μm even at higher bias voltage. Regions with high strip pitch are affected earlier. Turn-on-Voltage 12 11 1100 1000 900 800 700 600 500 400 300 W/P=0.13 1100 1000 900 800 700 600 500 400 300 W/P=0.23 1100 1000 900 800 700 600 500 400 300 W/P=0.33 15
Device Simulation Electric fields Simulations show higher electric fields at the strip edges for irradiated n-bulk than for p-bulk sensors with same geometry (except p-stop). This would explain why n-bulk sensors are more likely to discharge or break-down locally. Increasing oxide charge... increases max. electric fields in n-bulk but reduces max. electric fields in p-bulk. This would explain why RGHs are reduced for neutron irradiation with less ionization 16
Conclusion Signal and noise were studied on sensors produced by Hamamatsu. Sensors were irradiated to fluences expected at radius=20&60 cm. n- vs. p-bulk material: * n-bulk material gives a Higher signal up to, but at the maximum fluence tested, the p-bulk sensors measure a higher charge. * Signal on n-bulk material decreases after a couple of weeks at room temperature. * High noise, creating fake hits and increasing the occupancy. Float zone vs Magnetic Czochralski: Sensors made on Magnetic Czochralski material show an increased signal with annealing. 200 microns vs. 320 microns thickness: For fluences up to, the signal is comparable for both thicknesses. CMS decided to use p-bulk material for the Tracker and concentrate the future work on optimizing the technology and the geometry. 17
Back Up 18
Tracker Upgrade Motivations Need to adapt to HL-LHC radiation level, in particular silicon sensors. For 3000 at r = 22 cm: fluence = & dose = 0.4 Mgy Pile-up increases from ~ 20 events to 100-200 events per bunch crossing. up to 20 000 particles in the tracker. higher granularity needed to keep occupancy at 1% level. Keep L1 trigger frequency at 100kHz tracker data must be fed into L1 trigger. H ZZ eeµµ 20 pile-up events, corresp. To 200 pile-up events, corresp. to 19
Tracker Upgrade Motivations Detector suffers from its own material Photon conversions, Hadronic interactions ( detection efficiencies), Multiple scattering ( spatial resolution), Bremsstrahlung ( energy resolution) This drives many R&D areas Novel powering and cooling systems Reduced redundancy in layout Power savings in layout & electronics (strips ~ 0.2kW/m2) Reduce material in the tracking volume Improve performance @ low pt Reduce rates of nuclear interaction, γ conversions, bremsstrahlung Reduce average pitch Improve performance @ high pt 20
Currently, tracker information is used only in High Level Trigger (HLT) Reduction of L1-trigger rate by increasing thresholds not feasible (+ lose physics) Tracker info helps: improved pt-resolution, x10 = 20kHz e- matching, isolation, primary vertex identification pt-spectrum arriving at r=25cm η < 2.5 400 pile-up events Muon trigger rate L= M. Pesaresi, PhD Thesis Number of tracks CERN/LHCC/2002-26, CMS TDR 6.2 Trigger On Tracks Cannot read out full tracker data within latency (6.4µs) Identify high-pt tracks (~ 90% for pt < 2-3GeV and ship this data out only. 21
Stacked Layers High-pT tracks are more straight compare hit patterns in close layers ( stacks ) Needs pixelated detector layers Profits: Compatible with thin sensors (mass!) Provides z-information vertex identification Challenges: pt-layers will have high power consumption... will be costly... will be massive Correlation logic must take into account alignment, Lorenz angle etc. r ~ 100µm Pass Fail Upper sensor ~ 1mm B Lower sensor J. Jones, G. Hall, C. Foudas and A. Rose, LECC 2005 rφ 22
Outer Tracker Readout Chip Proposed successor of APV25: CMS Binary Chip (CBC) in 130nm CMOS Binary unsparsified readout: lowest power; synchronous; constant data volume Power per channel ~ 0.5mW (simulation for Csensor = 5pF) Compatible with both signal polarities, DC & AC coupling Challenges: noise immunity, fewer diagnostics, binary position resolution Design being finalized, layouting has started Expected for Spring 2010 Pre-amp Post-amp Strip (here n-in-p, DC) Peaking time 20ns M. Raymond and G. Hall, TWEPP2008 23
Leakage Currents Strip Sensors IV curve does not show a nice plateau after Vfd Chose 600V for comparison, which is the max. voltage of current CMS Tracker PS Leakage currents tend to be higher for 200µm sensors (up to ~40%) Absolute temperature measurement is difficult and leads to large errors protons neutrons mixed 24
Strip Sensors CV and Full Depletion Voltages Vfd vs. F increases faster for thick and for p-type sensors Points at 7e14neq/cm² are taken after mixed irradiation (3e14neq/cm² protons and 4e14neq/cm² neutrons plus additional annealing of ~19min@60 C) We see a kind of saturation for F>5e14neq/cm² 25
Mixed Irradiation Neutrons: 1 MeV Protons: 23 MeV 23 GeV (CERN PS) p-irrad (KIT) p-irrad (KIT) Short annealing - measurement Initial measurements Short annealing - measurement n-irrad (Ljubljana) n-irrad (Ljubljana) Mixed (p+n) dependence Long annealing and many measurements 26
Inter-Strip Resistance Both p-stop and p-spray show acceptable inter-strip resistance even after 1.3e16neq/cm2 27
Future Plans: 8 Sensors A new vendor for high volume production of silicon particle detectors Needs compatibility with 6 Need a layout which is suitable for Building module prototypes (2S and PS) Evaluate 8 production (n-on-p) P-stop studies Proposal for 1x2S, 1xPS-strips, several MPA (PS-pixel) sensors, different p-stop layouts 28
Open issues Strip isolation p-spray can be used and save money?. MCz p-spray type Study in more detail effect of [O] on p-type. Prevent charge loss between strips by design? up to fluences ~1.5x1015 all materials and polarities show up to 30% signal loss between strips. Study to investigate Long term effect. FZ 200um p-spray type Annealing: 156min@60 Voltage bias: 1000V 29