1. Wir schaffen Wissen – heute für morgen 6. März 2016 Paul Scherrer Institut Radiation Damage in Silicon Sensors for Particle Tracking Tilman Rohe

2. Outline Surface Damage Bulk or Crystal damage Dependence on particle type and energy Leakage current Internal electric field, N eff Effect of oxygenation Trapping, signal collection

3. Surface damage Definition: Changes at the interface between silicon and the covering insulation layer caused mostly by ionization in the oxide Why important? Inter pixel isolation Break down behavior Leakage current Parameters changing Fixed oxide charges –Electron hole pair generated by incoming  –Electrons very mobile –Holes: hopping through states in the SiO 2 bulk –Fall into traps close to the interface (dangling bonds? H-tempering) –Cause electron accumulation and short n- electrodes – Inter-strip isolation –Example shown as realized in BPIX Interface traps –Reduction of surface mobility –Increasing of surface currents - + + + + + + + + + ++ X-ray Gate EFEF EVEV ECEC - - - - - Gate (Al or Poly) Oxide Silicon

4. Measurement of MOS capacitor [Jiaguo Zhang RD50 Workshop, 2011, Liverpool]

5. Change with radiation Shift of the CV curve is caused by fixed oxide charge and interface traps Frequency dependence of CV is caused by interface traps only Increase of conductance during depletion is cause by the interface traps Determination of fixed charges is difficult –Determine energy levels and concentration of traps with other technique –Fit CV and GV curves using this values Done as part of the R&D for the European XFEL (HH) [Jiaguo Zhang, RD50 Workshop, 2011, Liverpool]

6. Dose dependence Gamma irradiation using 60 Co Ionizing radiation only No crystal damage No bias applied Fixed oxide charge saturates after about 20kGy Interface traps do not saturate Bias applied Concentration of both oxide charges and interface traps much higher Saturation of oxide charge much earlier Technology dependence Defect concentration –Depends much on crystal orientation –Process details In general: material shows lower concentrations than Sarturation of oxide charges [Wunstorf 96]

7. Inter pixel isolation Fixed oxide charge Creates a conducting channel between n-electrodes Determines the electrical field in the critical area close to the surface Technology parameters (dose of the isolation implant has to be adjusted) P-Stops (FPix) High boron dose (adjustment uncritical) Alignment important (Breakdown after irradiation if dose is too high) P-Spray, mod. p-Spray (BPix) HV-stability of un-irradiated device critical Boron dose to be adjusted –High enough to provide isolation –Low enough to enable HV operation of new devices (e.g. during module production) Narrow gaps possible (without moderation) –Punch through structures –Homogenous drift field at high voltages

8. Critical positions in the BPix layout Insufficient isolation at punch through dot fatal because it would short all pixels Has to be confident that production parameters a well chosen 7µm 20µm

9. Check of interpixel isolation Use special test structure which “imitates” the biasing structure Applied 0.6V (~0.5 * V ana ) Measured the current as function of the back side voltage –Some over depletion needed to separate bias grid from pixels –No dramatic change with radiation Chosen gap of 7 µm

10. Crystal damage Charged particles scatter electrons and lift them from the valence into the conduction band. Those electron-hole pairs are the signal. Those particles also hit Silicon nuclei and displace them –Interstitial-Vacancy (Frenckel) – pair, if exchanged momentum is small (> 25 eV) –Those defects are not stable. They move through the crystal. –Anneal –“Meet” other defects and form a stable one (V-O, V-V, etc.) –Spatially separated defects are called point defects –If the exchanged momentum is > ~5 keV the Silicon nuclei will cause more defects which will be located close together ( clusters ) –The ration of point defects to clusters depends on the kind of radiation particle V I 10 MeV protons 24 GeV/c protons 1 MeV neutrons [Mika Huhtinen NIMA 491(2002) 194]

11. Normalization of radiation damage NIEL-hypothesis : parameters characterising “radiation damage” scales with the deposited energy with does not create electron hole pairs ( non-ionizing energy loss ) – caveat is not entirely true “Norm” particles are 1 MeV neutrons Hardness factor  : ratio of the NIEL for the particle under consideration divided by the NIEL of an 1MeV neutron Values are calculated with simulation programs

12. Influence of point defects on sensor properties Trapping (e and h)  CCE shallow defects do not contribute at room temperature due to fast detrapping charged defects  N eff, V dep e.g. donors in upper and acceptors in lower half of band gap generation  leakage current Levels close to midgap most effective Characterisation of defects: DLTS method Formation of point defects can be influenced by impurities in the crystal –Oxygen forms stable V-O i defects, which are “less harmful” than mainly di-vacancies –Catch vacancies Use of Oxygen rich material (DOFZ, mCz) is of advantage if point defects are dominant (as the case for the pixels where pions heavily dominate)

13. Leakage current Why important? ROC operation (noise, dc-coupled preamplifier) Power dissipation (supplies and cooling) Volume current Silicon is an indirect semiconductor The intrinsic charge carrier density (n i ) is small Thermal electron-hole pairs are (at “normal” temperatures) created at deep levels close to the middle of the band gap. –Is characterized by the generation life time  g –A high life time is a quality measure of the substrate J vol ~ -e n i /  g * W Temperature dependence J vol ~ T 2 exp(E g (T)/2kT) or a factor 2 every 8K Surface currents Cannot be parameterized that easy –Can also display the same temperature dependence if caused by surface traps –Can be independent of T if “ohmic” e.g. caused by surface accumulation layers Breakdown current Avalanche break down has completely different temperature dependence

14. Radiation induced increase of leakage current Leakage current was measured with a very large number of small diodes with the guard current separated Volume current only Segmented sensors might be different Result Current density is proportional to the Fluence Damage constant  =  I / (V  eq ) independent of –Kind of particle –Fluence –Growth method of the crystal –Impurities of the crystal –NIEL hypothesis correct for I leak Leakage current of diodes is used for fluence calibration or as measurement of  [M.Moll PhD thesis]

15. Annealing of leakage current The radiation induced current reduces with time Short term: a combination of exponentials with different amplitudes and time constants Long term (> 1 month at RT): ~log(t) Annealing is strongly temperature dependent During operation low temperatures are preferred Outside operation a warm up is beneficial (but too long because of “reverse annealing”, see N eff )

16. Leakage current of CMS pixel detector Leakage current measurement in the detector could in principle be used to calculate The received fluence  (if T is known) The sensor temperature (if  is known) Unfortunately this leads to an overestimation of the received fluence (compared to counting clusters) or an unrealistically high temperature Annealing history not properly included Literature value of  “wrong” –Does not contain surface components of the current –Might be possible to estimate those from the slope of the IV curves at V>V depl

17. Internal electric field Why important? Defines the minimum Voltage for efficient operation In typical Silicon sensors the doping of the substrate is constant The electric field grows linearly with the depth and peaks at the junction The electrical potential increases with the 2 nd power of the depth – Growth of depletion zone ~ sqrt(V) The type of the material defines the space charge: – n-material: positive – p-material: negative Space charge (N eff ) can be measured with CV curves [L. Andricek]

18. Change of internal electric field with radiation Over many years the fluence dependence of N eff was studied using small diodes and CV-measurements  N eff(  ) = N eff (  =0) (1-exp(-c  )) –  1 st term describes the “donor removal” –Fluence needed to remove all donors depends on the resistance of the starting material 2 nd term describes the formation of negatively charged defects The point at which N eff reaches a minimum is called “type inversion” p-type means that the “average space charge” is negative The inverted substrate has “intrinsic” conductivity –Donors are “removed”/deactivated –Radiation induced charged states are too deep in the band gap to provide free charge carriers [M. Bruzzi, TNS, 2001] [M. Moll PhD thesis]

19. Effect of Oxygen content in silicon Effect of 24 GeV/c protons (CERN-PS) studied FZ-Silicon Neff clearly goes to very low numbers (“type inversion”) Defect creation rate high Diffusion oxygenated FZ (DOFZ) Defect creation rate strongly reduced “type inversion” still visible Cz and mCz material Neff does not come close to zero (“no type inversion”) Caveat : Difference only for charged hadrons, for neutron irradiation no difference CV measurements only measure the V depl After the radiation induced defects have a concentration comparable or larger than the (active) doping, the field is not anymore linear but displays 2 peaks ( double junction ) The term “type inversion” looses its meaning [M. Moll]

20. Mixed irradiation Change of N eff is different Cz material for irradiation with neutrons and charged hadrons Compensating effect is seen if samples are irradiated with both kinds of particles for Cz –In FZ effects add –DOFZ ? This is en effect of the different space charge, trapping is not effected Such mixed particle fields are found in the region of the strip tracker (esp. the TIB) [ F. Hartmann, Nucl. Instr. Meth. A, Vol. 628, (2011) 40-49]

21. Double junction from edge TCT measurements Polish the side of the sensor Shine in with laser and focus it below a known strip Measure the pulse with a fast amplifier rise time is a measure of drift velocity Integral of the pulse is a measure of efficiency Result here: Though the field develops from the front side, the more efficient region at high voltages is located close to the backside. –Electrons travelling from the back to the front are travelling faster than holes from the front to the back, thus having less chance of being trapped [N. Pacifico, CERN 2011

22. Annealing of space charge effects CV measurements were repeated after irradiation with heating steps in-between. Three components are distinguished Stable damage –Does not depend on annealing time –Defines the minimum N eff Short term –Beneficial annealing (~2 weeks at RT, depending on fluence) Long Term –Reverse annealing –Requires to keep detector cooled even if switched off –Amplitude of reverse annealing saturates for DOFZ material

23. Signal formation (i) (ii) Why is trapping important? Signal loss due to trapping sets the operation limit Signal is induced by the drift of charges Begins instantaneously Mechanism can be described by (purely geometrical) weighting field (Ramo) Signal induced when charge q drifts from x 1 to x 2 Q=q[  w (x 1 )-  w (x 2 )] Field is calculated by –setting electrode under consideration to potential 1 and all others to 0 –Solving the Poisson equation (without fixed charge) The smaller the electrodes the more the field is concentrated –In such cases most of the charge is induced at the last part of the particle drift path –Such devices are more sensitive to trapping If q is collected by neighbour (i), a “bipolar” signal is introduced with the total 0 Charge drifting along (ii) induces q

24. Trapping times The effective trapping time is reversely proportional to the equivalent fluence  eq : 1/  eff =  eq  found independent of Silicon resistivity [O], [C] up to 1.8e16 cm -3 Type (p / n) Wafer production (FZ, Cz, epitaxial) Electrons and holes have approximately the same trapping time, but mobility of electrons is 3 times higher  Electron devices are less prone to trapping drift velocity trapping I(t) difficult to integrate  (-10 o C, t=min Vfd) [10 -16 cm 2 /ns] 24 GeV protons (average ) reactor neutrons Electrons 5.6±0.24.1±0.2 Holes6.6±0.36.0±0.3 References G. Kramberger et al, Nucl. Inst. Meth. A481 (2002) 297. A.G. Bates and M. Moll, Nucl. Instr. and Meth. A555 (2005) 113. O. Krasel et al., IEEE Trans. NS 51(1) (2004) 3055. E. Fretwurst et al., “Survey Of Recent Radiation Damage Studies at Hamburg'',presented at 3rd RD50 Workshop, CERN, 2003.

25. Annealing of trapping times Trapping time decreases for holes increases for electrons Advantage for electron collecting devices

26. Measurement of signal collection Strip detectors on a large variety of substrates was irradiated with all particle types currently available (Protons: 24 GeV/c CERN-PS, 26 MeV/c Karlsruhe; reactor neutron Ljubljana; 300 MeV/c pions PSI) After high fluences, when signal height is mostly defined by trapping and less by differences in the space charge: Basically all materials behave equal (if electrons are collected) All kind of hadrons make the same damage n-in-n mCz at 1E15 (neutrons) at 900V sticks out –Based on 1 sample –Needs to be verified Seems that signal does not follow annealing of N eff n-in-p FZ [A. Affolder – RD50, 24th-26th May 2011, Liverpool] n-in-n MCZ 26 GeV/c p

27. Charge multiplication Sr-90 measurements with strip detectors (Ljubljana): Signal did not saturate with bias “CCE” reached values > 1 IV curves shoed strong increase with bias –Detectors were cooled to low temperatures (< -30º C) Effect was explained with localised charge multiplication compatible to the effect in APDs and triggered a large R&D activity within the RD50 collaboration which is still ongoing. Tried to do similar measurements at PSI (with the help of many PIRE students – thank you !! ) See clear saturation of the signal up to  ~1E15 For the sample with 2.8E15 the situation is not clear –Understanding of ROC operation after such high fluences is not, yet, complete Improved cooling box in order to reach ~ -25º C Many new samples wait to be mounted an irradiated [I. Mandic NIM A 612 (2010) 474-477]

28. Radiation induced operation limitations 2500 fb -1 Pixels“Strip like” limited by trappinglimited by space charge What here? ~6500 e - No charge sharing ~ fluence limit of present detector mainly pionsmainly neutrons benefit from “new” materials (mCz,...) data rate limit of present ROC

29. 6. März 2016PSI,6. März 2016 Seite 29 Tried to give overview on radiation damage in silicon sensors for particle tracking …surface damage has to be considered when choosing the technology but does not limit operation …increase of leakage current can only be influenced by the temperature. Power dissipation might set a limit. …decrease of signal due to trapping is the present limit. Way out: Charge multiplication? Other sensors (3D)? …change of internal electric field is complex and can be engineered by the choice of the ground material. Thank you