SCIPPSCIPP
Hartmut Sadrozinski, with
Scott Ely, Vitaliy Fadeyev, Zachary Galloway, Jeffrey Ngo, Colin Parker, Brett Petersen, Abe Seiden
SCIPP, Univ. of California Santa Cruz
Nicolo Cartiglia, Amadeo Staiano INFN Torino
Mara Bruzzi, Riccardo Mori, Monica Scaringella, Anna VinattieriUniversita de Firenze
4D Sensors: Unifying the Space and Time Domain
with Ultra-Fast Silicon Detectors
SCIPPSCIPP
“4D”
Ultra-Fast Silicon Detectors (UFSD) incorporate the time-domain into the excellent position resolution of semiconductor sensors
they provide in the same detector and readout chain• ultra-fast timing resolution [10’s of ps]• precision location information [10’s of mm]
2 questions need to be addressed for UFSD:• can they work: signal, capacitance, collection time vs. thickness • will they work: required gain and E-field, fast readout
We hope that we will answer the questions within an RD50 Common Project (Giulio Pellegrini)
A crucial element for UFSD is the charge multiplication in silicon sensors investigated by RD50, which permits the use of very thin detectors without
loss of signal-to-noise.
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 2
SCIPPSCIPP
Motivation for UFSD
Up to now, semiconductor sensors have supplied precision data only for the 3 space dimensions (diodes, strips, pixels, even “3D”), while the time dimension has had limited accuracy (e.g. to match the beam structure in the accelerator).
We believe that being able to resolve the time dimension with ps accuracy would open up completely new applications not limited to HEP
Proposal: Combined-function pixel detector will collect electrons from thin n-on-p pixel sensors read out with short shaping time electronics
Charge multiplication with moderate gain g ~10 increases the collected signal
Need very fast pixel readout
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 3
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 4
Benefits of Ultra-Fast Silicon Detectors
- Tracking: Identifying with high precision the temporal signature of different events allows for their association and it reduces random coincidences. Traditional tracking is often overwhelmed by combinatorial backgrounds, which can be dramatically reduced by adding a 4th dimension (time) per point.- Time of Flight (ToF): ToF is already used in many commercial applications such as ToF-enhanced PET and Mass Spectroscopy ToF, however with precision one order of magnitude higher than the goal of UFSD (~500 ps vs. ~50 ps). ToF is also used in particle physics as a tool for particle identification. ToF can also be used in 3D and Robotic Vision: the ability to accurately measure the travel time of light pulses reflected by an object at unknown distance is of paramount importance to reconstruct 3D images, fundamental in imaging and robotic vision. UFSD will offer a spatial precision of a few mm at low illumination power, allowing for battery operated, portable systems.- Particle counting: UFSD performance would allow developing new tools in single particle counting applications with unprecedented rate capabilities. For example, in the treatment of cancer using hadron beams, such a tool would measure the delivered dose to patients by directly counting the number of hadrons.Material science experiments using soft x-rays will benefit from the combination of high rate and precision location that UFSD offers.
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 5
Remote Sensing in HEPTime of flight determines position:
with σt=10 ps, σx= 3 mm, i.e. locate a proton within 3 mm in 20 cm long IPtx c
SCIPPSCIPP
TOF for Particle Identification in Space
The Alpha Magnetic Spectrometer (AMS) detector, operating in the International Space Station since 2011, performs precision measurements of cosmic ray composition and flux. The momentum of the particles is measured with high-resolution silicon sensors inside a magnetic field of about 1 m length.
A time resolution of 10 picoseconds, the “Holy Grail” of Cosmic Ray Physics: the distinction between anti-carbon ions and anti-protons can be achieved up to a momentum of 200 GeV/c.
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 6
SCIPPSCIPP
7
Range straggling limit for 200 MeV p
Future: 4-D Ultra-Fast Si Detectors ?
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013
Protons of 200 MeV have a range of ~ 30 cm in plastic scintillator. The straggling limits the WEPL resolution.
Replace calorimeter/range counter by TOF:Light-weight, combine tracking with WEPL determination
SCIPPSCIPP
8
A. Del Guerra, RESMDD12
Positron Emission Tomography PETStudy accumulation of radioactive tracers in specific organs. The tracer has radioactive positron decay, and the positron annihilates within a short Distance with emission of 511 keV γ pair, which are observed in coincidence.
Resolution of detector (pitch)Positron rangeA-collinearityParallax (depth)
T: true event S: Compton ScatterR: Random Coincidence
Resolution and S/N Effects:
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013
Perfect Picture:
SCIPPSCIPPReduce Accidentals & Improve Image: TOF-PET
t1
t2
Localization uncertainty:Dd = c x Dt /2
When Dt = 200 ps ➔ Dd = 3 cm
PET
TOF-PET
@ VCIK. Yamamoto 2012 IEEE NSS-MIC
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 9
SCIPPSCIPP
10
1. For a given acquisition time and dose to the patient, TOF can provide better image quality and improved lesion detection.
OR2. with TOF the scan time and dose can be
reduced while keeping the same image quality ( better clinical workflow and added comfort for the patient).
2/tx c
TOFNonx
TOF SNRDSNR
TOF – PET SNR Improvement
M. Conti, Eur. J. Nucl. Med. MoI Imaging (2011) 38:1147-1157
The improved source localization due to timing
leads to an improvement in signal-to-noise
and an increase in Noise Equivalent Count NEC x
DGainNEC
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013
SCIPPSCIPP
Can it work? UFSD Collected Charge
Signal = thickness*EPM (EPM = 73 e-/mmCollection time = thickness/vsat (vsat = 80 mm/ns)
For thickness > 5 um, Capacitance to the backplane Cb << Cint
For thickness = 2 um, Cb ~ ½ of Cint, and we might need bipolar (SiGe)?
Realistic gain & cap
Good time resolution
Thickness BackPlane Capacitance Signal Coll. Time[um] Pixels [fF] Strips [pF] [# of e-] [ps] for 2000 e for 12000 e0.1 2500 500 8.3 1.3 241 14461 250 50 83 12.5 24 1452 125 25 166 25.0 12 725 50 10 415 62.5 4.8 2910 25 5 830 125.0 2.4 1420 13 3 1660 250.0 1.2 7.2
100 3 1 8300 1250.0 0.2 1.4300 1 0 24900 3750.0 0.1 0.5
Gain required
Per 1 cm
WRONG
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 11
SCIPPSCIPP
Energy loss measurement for charged particles in very thin silicon layers
S. Meroli, D. Passeri and L. Servoli
11 JINST 6 P06013
Details of Collected Charge in Thin Sensors
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 12
SCIPPSCIPP
Correct UFSD Collected Charge
Gain needed for a sensor thickness of 10 µm
Pixel about 4
1 cm long strips about 20
(much less than APD’s or SiPM)
Realistic gain & cap
Good time resolution
Thickness BackPlane Capacitance Signal Coll. Time[um] Pixels [fF] Strips [pF] [# of e-] [ps] for 2000 e for 12000 e
2 125 25 80 25.0 25 1495 50 10 235 62.5 8.5 5110 25 5 523 125.0 3.8 2320 13 3 1149 250.0 1.7 10.4
100 3 1 6954 1250.0 0.29 1.7300 1 0 23334 3750.0 0.09 0.5
Gain required
Per 1 cm
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 13
SCIPPSCIPP
Impact Ionization
)/exp(* ,,, EbE hehehe
Eb
E
NN
hehehe
,,,
0
exp*
)*exp(*)(
A. Macchiolo,16th RD50 Workshop Barcelona, Spain, May 2010
Charge multiplication in path length ℓ :
At the breakdown field in Si of 270kV/cm:e ≈ 1 pair/um
h ≈ 0.1 pair/um
→ In the linear mode (gain ~10), consider electrons only
Raise maximum and minimum E-field
as close to breakdown field as possible
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 14
SCIPPSCIPP
Non-uniform E-Field across a pixel/stripNon-uniform Field across the implant results in charge collection difference
Example of electric field: Ф = 1.6·1015 n/cm2, U = 900 V
Even if non-uniformity of field across the implant is only 30%, a large fraction of the center of the implant does not exhibit charge multiplication before sensor breakdown !
Gregor Kramberger, 19th RD50 Workshop, CERN, Nov 2011 A. Macchiolo, 16th RD50 Workshop Barcelona, Spain, May 2010
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 15
SCIPPSCIPP
16
Charge Multiplication C.M.
M. Köhler, et al., Nucl. Physics B – Proceed. Supplements 215, 1, (2011), 247-249
G. Casse et al, NIM A 624, 2, (2010) 401–404
I. Mandic et al., Nucl. Instr. and Meth. A 612, (2010 ) 474-477.
Expectationfrom trapping extrapolation
RD50 investigated C.M. and found it in irradiated sensors only.C.M. exists if signal is in excess of pre-rad value
orin excess of trapping measurements
Advantage of thin sensors because of high E-field But indication of excess noise.
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013
SCIPPSCIPP
Epi, short drift on planar diode g = 6.5
Using red laser and ’s probes E-field and gainclose to the junction, where it counts. Diode gives uniform field.
J. Lange et al., Nucl.Instrum.Meth A622, 49-58, 2010.
Need uniform field: 3D or diode-like
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 17
SCIPPSCIPP
What about fast readout:
CERN fixed-target experiment (NA62) needs very fast pixel sensors: Gigatracker (GTK)
Prototype CFD system (INFN Torino) has ~ 100 ps resolution, predicted to be 30 ps in next iteration.
Optimized for 200mm sensors and hole collection (?), could it be re-designed for electron collection from 2 – 10mm sensors?
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 18
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 19
SCIPPSCIPP
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 21
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 22
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 23
SCIPPSCIPP
Firenze ps Laser Data: 50 um p-on-n 6kW-cm
Laser: 1.2 ps pulse width & 10 ns period,red 740 nm, penetrates ~6 um. Oscilloscope: 500 MHz bandwidth Charge collection of electrons moving away from laser spot
Terminal velocity ~ 100um/ns, i.e. expected collection time ~500ps for high fields
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 24
FWHM becomes constant at about 150 V biasRise times RT and trise are ~ independent of bias: Bandwidth of system
Above 120V bias, the field > 25kV/cm, i.e. large enough to saturate the drift velocity, i.e. constant collection time.
SCIPPSCIPP
25
Study of ’s from Thorium 230 50 um thick P-type epi diodes
IV Measurements for diode 2 & 3 during C-V
Epi Diode IV Curve: Diode 2 & 3
1.E-10
1.E-09
1.E-08
1.E-07
1.E-06
0 50 100 150 200 250 300 350
Bias (V)
Cur
rent
(A)
Diode 2Diode 3 Challenge for manufacturers:
For charge multiplication, need breakdown voltage >1000V!!
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013
SCIPPSCIPP
Do NOT expect: Charge multiplication
Pulses from Th 230 alpha’s
Expect collection time of ~ 500ps for over-depleted bias (drift velocity is saturated)large rise time dispersion for low bias voltage
SCIPPSCIPP
Doping Profiles & Biasing
Assume doping distributions with maximum 2 doping distributions, uniform in depth,“pads fields”
p+ substr.
p- epi
+ n
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 27
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 28
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 29
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 30
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 31
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 32G. Kramberger, Preliminary!
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 33G. Kramberger, Preliminary!
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 34G. Kramberger, Preliminary!
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 35G. Kramberger, Preliminary!
SCIPPSCIPP
Rapid rise of signal up to 100V during the depletion of the sensor.“plateau” beyond this : gain almost constant beyond ~50V ?
Explanation of rapid rise and level
Toy Simulation
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 36
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 37
Why/How does C.M. work stably?
Planar vs. 3D: “smart money” is on planar, although 3D has a uniform 1/r field around electrodeit seems to be harder to get the proper implant profile for the high field section.
Breakdown associated with charge multiplication in high fields:Our experience with drift chambers tells us that you need a “quenching” mechanism That’s why the “magic gas” worked, a mixture of 4 gases:It added to the high gain of Ar the quenching of CO2, isobutane, and methylal or DME.
The fact that we observe gain in low-resistivity (low-R) silicon at RT and low currents as well as in irradiated sensors with large numbers of defects and high currents seems to indicate that the quenching comes from the large numbers of defects which provide trapping of excessive charges.
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 38
4D UFSD Score Card March 2013
Gain in irradiated sensors (pads + strips) √ 2009
Gain in un-irradiated pads √ 2013
Gain in un-irradiated strips/pixels Fall 2013 ?
Fast Readout ASIC Proto type (60 ps) 2012
Fast Readout ASIC (50 ps) 2014
Pixel System (Sensor, ASIC, DAQ) 2016
SCIPPSCIPP
Hartmut F.-W. Sadrozinski: UFSD SCIPP 2013 39
Quo Vadis with 4D UFSD?
Fast developments:
First Presentation June 2, 2012 in Bari for RD50November 2012: Common Project within RD50 accepted
Research project funded by DoE within the SCIPP Instrumentation Section
2 EU Research proposals submitted including Torino (Frontend ASIC), Vienna (DAQ)
March 2013: Gain of ~10 seen in un-irradiated diodesMarch 2013: submission of dedicated strip sensors for gain studies
(CNM)Spring 2013: Spanish grad student Marta to come to SCIPP
SCIPP’s role:
Application of SiGe in front-end (work with Torino)Simulations (Marta & Colin & ..)Pulse shape studies (need fast scope and increasingly faster pre-
amps!)Strip readout vs. pixel readout