L. Rossi 1

1
Pixel 2005 summary

• The workshop has tried to cover all about pixel
  detectors, looking at
• today (mostly hybrid p.) and tomorrow (mostly
  integrated p?) in HEP
• non-HEP applications (biomedical, astronomy,)
• other-then-silicon sensors
• Some statistics
• 50 talks and 19 posters
• half of the talks are on hybrid pixel or related
  to, this represents the work of 50 labs and
  large investments for construction.
• evolution of the field
• Compare with Pixel2000 workshop (Genoa)
• of talks dedicated to hybrid pixels decreases
  (85?50). Partly due to a stop of one project
  (BTeV), but also indicating that hybrid pixel is
  a mature technology and the challenges are now
  in
• Integrated solutions (MAPS, DEPFET, PCCCDS,,)
• New material (mostly for non-HEP applications)
• spin-off activities (pixels start to be
  competitive in some)

2
biases

• Make a conference summary talk is not easy (the
  last presentation was few minutes ago), make it
  of a workshop (more discussions) is even more
  difficult
• I have privileged new results (general bias),
  even when old results were more relevant
• I have privileged what I know better (personal
  bias)
• therefore some results (even important) may not
  have been reported (especially in the last
  session), I apologizes for this
• finally, it would not be surprising if there are
  mistakes
• General comments
• level of presentations quite good and nobody run
  overtime significantly, thats good. I would
  nevertheless advice to reduce of talks next
  time
• Even if the summary talk will be done by somebody
  else...
• level of invited talk excellent, I have learnt a
  lot. Some of those have been very useful to
  understand fields where pixel may bring (or get)
  some contributions.
• I would have appreciated more problems presented
  and discussed.

3
Hybrid pixel

• Mostly reports about status of construction of
  three detectors (ALICE, ATLAS and CMS) all to be
  operated at LHC.
• First time in Pixel conf. that we can hear
  reports about detectors in (more or less)
  advanced construction status, not just
  prototypes.
• Use in experiment of hybrid pixel designed for
  LHC has been already anticipated by NA60. M. Keil
  reported about the improvement in vertex
  reconstruction

• Used ALICE and (for high intensity runs) ATLAS
  chips and modules

4

• Also reported improvement (23 instead than 80
  MeV) in mm- invariant mass reconstruction.

• all that was expected, but nice to see

5
ALICE

• Less rad-hardness problems than ATLASCMS,
  challenge is transparency (lt1 X0/layer is the
  design aim)
• Thin detector and thinned electronics
• yield penalties (also in flip-chip)
• Thin services (multilayer Al on kapton)
• Central part Al/kapton, then Cu/kapton

Riedler
6

• Thin mechanical cooling support
• Very sophisticated CFRP shapes to host modules
  with minimal mass
• Prototype detectors tested on beam
• and simulation tuned to data (cl.size)
• but still without B-field (Elia).
• Production of all critical parts started, yield
  of (large) chip is good (52) and already 3900
  dies have been tested.
• Still to tune on the flip-chip yield (now 64)

Pepato
chips good per wafer
7
ATLAS

• Most advanced (50 modules built), start to
  worry about possible beam accidents.
• Gemme demonstrated that pilot beam accidents (the
  only acc. LHC experiments should in principle
  worry about) should not damage ATLAS pixel
  modules. An experiment was done with PS extracted
  bunches of 42ns and 1011 p, grazing a pixel
  module. This corresponds to gt30 times the energy
  released by a beam accident and did not harm the
  pixel module (so some safety factor)
• Andreazza presented a
• detailed analysis of test beam
• results of irradiated (to 500 kGy)
• production modules.
• New results on trapping
• indicate a carrier lifetime ?e 4.10.30.5
  ns

8

• High intensity operation has also been tested and
  overall losses are proven to be small even at the
  expected high luminosity in the B-layer.

Drop in efficiency of 1 starts to be seen at
occupancy gt0.21 hits/cp/bx
9

• Peric described in detail the ATLAS pixel FE
  chips now fully fabricated and tested (average
  yield 82 on 250 wafers).
• Design specs have all been met (also thanks to
  many global and local tuning bits).
• Some further options have been included and worth
  noticing. PH information through ToT,
  self-trigger capability both are important for
  detector understanding and calibration.
• A trick to recuperate low (i.e.late) hits has
  been implemented on chip and can be activated
  when needed.

10
CMS

• Is about to begin production
• Kaestli presented the test results on PSI46v2 the
  production version of the front-end chip
  (including analog data transmission).

11

• Erdmann showed results of modules built with
  PSI46v2 in high intensity beam (108/cm2)_at_PSI

• .e not fully satisfactory and there is some need
  for periodic reset of the chips. Analysis very
  new, may need some more thinking to sort out the
  problems (or some re-design?)

12
Bump-bonding

• Several firms active in the field (reports from
  AMS (now SELEX), IZM,VTT, PSI)
• Have been possible due to the volume of
  placements related to the the LHC projects
  (50k), but this still needs close follow-up from
  experimental groups to optimize process and yield
  
• Alimonti (for INFN and SELEX) reported a yield of
  68 (before rework) and 86 after rework at a
  production rate of 25 16-chip ATLAS pixel
  modules/week (now 800 modules done).
• Reasons for failure are multiple
• and often difficult to identify, but
• possible only if access to the full
• production chain is available
• Rework is mandatory in these
• yield situation. Still valuable even in
• the substantially higher yield obtained
• at IZM (Fritzsch) where 91?98.

13

• Rohe presented the in house bump-flip capability
  developed at PSI based on In reflow, bump yield
  better than 99.9 (flip-chip and rework yield
  still need more statistics).

• Solder bumping has been discussed by Savoilainen
  (VTT) who worked for ALICE and LHCb and presented
  prescriptions on yield optimization throughout
  the whole process.
• A very interesting overview of the bumping
  process and problems has been presented by Wolf
  (IZM) who went through all the possible
  techniques that are industrially applicable (but
  In).

14

• Electroplated bump deposition is the present
  choice for high density bumping and half of the
  ATLAS pixle modules have been fabricated with
  this technique.
• Future trends require abandoning Pb and trials
  are being done to define the metallurgy of a
  leadless process.

• HEP is a small special part of a large world bump
  market (0.5) and we have to follow the trends.
• The amount of information provided by Wolf is
  quite impressive and indicate the complexity of
  the problem
• Two points are worth mentioning
• IZM is building modules for HEP
• w/good yield and in large
• IZM is willing to continue collaborating with HEP.

15
Radiation hardness of sensors
Conclusions from the very complete and detailed
review of Moll

• At fluences up to 1015cm-2 (Outer layers of a
  SLHC detector) the change of depletionvoltage
  and the large area to be covered by detectors is
  the major problem.
• CZ silicon detectors could be a cost-effective
  radiation hard solution
  (no type inversion, use p-in-n
  technology)
• p-type silicon microstrip detectors show very
  encouraging results CCE ? 6500 e Feq
  4?1015 cm-2, 300mm, collection of electrons,
  no reverse annealing observed in CCE
  measurement!
• At the fluence of 1016cm-2 (Innermost layer of a
  SLHC detector) the active thickness of any
  silicon material is significantly reduced due to
  trapping. The promising new options are
  Thin/EPI detectors drawback radiation hard
  electronics for low signals needed
  e.g. 2300e at Feq
  8x1015cm-2, 50mm EPI,
  . thicker layers will be
  tested in 2005/2006 3D detectors
  drawback technology has to be optimized
  ..
  steady progress within RD50

16
Deep(er) submicron electronics

• Yarema presented (for Mekkaoui) the FNAL test
  chips realized in IMB 0.13mm to understand the
  technology as a first step for a new FE chip
  design
• Results are encouraging and we may aim at smaller
  pixels for new LHC b_layers (noise33-66e
  thr718-1160e and s(thr)93-170e) or other use
• Interesting (and worth further investigations)
  the 90Sr signal got out of a triple N-well diodes
  implanted in the matrix.

2000 e- signal
17
CCDs

• Bosiers made an excellent overview of CCD issues,
  showing (amongst many other interesting things)
  that CCD can also be pretty fast in image
  capturing (need local buffer), still not so much
  in data transfer.

An image sensor which captures 100 consecutive
frames at 1000000 frames/s, T.G. Etoh et al.,
IEEE T-ED, vol. 50, January 2003
18

• Quite some debate about CCD vs CMOS (for image
  applications). This is a question since long and
  not solved yet. Pros and cons have been be
  listed.

• CCD
• Perfect image quality
• High fill factor
• Low dark current
• Signal processing in charge domain
• Radiation hardness
• Read-out speed

• CMOS
• Low readout noise (line frequency)
• High-speed (parallel processing)
• System-on-chip
• Radiation hardness
• More complex design
• Expensive mask sets for 0.18mm and below

19

• Stefanov showed that the read-out speed
  limitation can be largely improved with
  P(arallel)C(olumn)CCD

• 60 e noise measured with sensor and
  column-parallel read-out _at_ 1 MHz
• There are still some gain variation over the
  matrix (a new version under design) and the
  rad-hardness to measure

PPCCD bumped on CMOS ASICS
20
Medical imaging

• Overdick illustrated the status of the art of
  industrial production of medical imagers
  (especially tomography, but also mammography and
  dental x-rays).
• Even if the field is moving to Direct Conversion
  (DICO) devices (like pixels) and to higher data
  throughput, I got the impression that there is
  not much phase space for pixel detectors there.
  Competition with flat panel sensors, CCDs and
  other well established detectors does not leave
  much room.
• Energy resolved x-ray in pixel counting devices
  may be a window of opportunity.

21
Astronomy DEPFETs

• Finger made a detailed and interesting overview
  of the IR astronomy sensors
• Quite challenging requirements
• Fast, low noise (few e-) _at_ T80K with good QE
  over 0.66-2.5mm, minimal interpixel cross-talk
  and need fancy sensors.

Example of a 2kx2k array for VISTA

• HgCdTe grown by LPE on CdZnTe substrate
• Pixel size 20 mm
• 16 parallel outputs
• Pixel rate 400KHz
• Frame rate 1.45 Hz
• 3-side buttable
• Multilayer ceramic mother board on metal
  pedestal
• Reference cells included in video data stream
• ESO is building 256 channel data acquisition
  system (IRACE)

22

• Frame store pnCCDs (Meidiger) for x-ray detection
  in astronomy (mission ROSITA to detect 50000 new
  sources of lt2 KeV and measure dark energy/matter
  in universe).
• pnCCD msec time resolution, 450 mm fully
  depleted, 75 x 75 mm pixel 256x256 array.
• For future avalanche pnCCD for single optical
  photon counting (i.e. an avalanche amplifier
  added to the on-chip FET). To be backsided
  illuminated through thin homogeneous window.

• Very good energy resolution 48eV(_at_277 eV)?133
  eV(_at_ 2 KeV) running cold (-80C), does not degrade
  much with T up to -25C.

20 frames/s, -75C
23
Pilatus

• Hybrid pixel used for x-ray diffraction studies
  in PSI on protein cristallography (Henrich )and
  metallurgy (Tokoyawa).

(c) 600?
a211
Thaumatin electron density mapResolution 1.4
Å
600C
?311
Time resolved diffraction pattern, to study
crystal formation in welding
?220
500C
(e) 400?
a211
a220
400C
a200
24
CMOS sensors

• Nicely summarized by Winter.
• CMOS s. provide a good balance between
  granularity, speed, radiation hardness and
  material budget and may find their application in
  future accelerators (e.g. ILC)

• Signal created in the O(15mm) thick epi-layer and
  collected by diffusion in O(100 ns).
• Electronics can be integrated on chip
• Industrial product (potentially cheap and good
  fab procedure established)

• The first detector was MIMOSA(s), now also CAP
  (for Belle upgrade) both have been qualified with
  particle beam and work fine (S/N20-30) (see
  Contarato and Varner, respectively).

25

• Recent significative progress on radiation damage
  (depends on design and on fabrication process)

26

• First results on a triple-well CMOS MAPS has been
  presented by Rizzo, it uses 0.13mm process from
  ST.

• This solution better insulates (through the deep
  n-well) the sensor cell from digital signal
  traffic and offers the possibility of implanting
  complex logic in the pixel cell.
• Preliminary results 1500 e- collected in 50 ns.
  S/N10 due to mismatch between sensor and
  electronics it is expected to rise to 30 in the
  next generation of the chip.
• Virtue can (in principle) implement complex
  trigger functions on chip.

Landau peak 80 mV
saturation due to low energy particle.
1250
2200
3000 (e-)
Threshold set cuts this region
27
New sensor material developments

• Fiederle made an exhaustive overview of CTe and
  CZT detectors and fabrication (included supplier
  info).
• processing of contacts and passivation
  (difficult)
• Note bulk material sensitive to T
• In bumping (low T and high density) or Au-stud or
  z-bond (polymere bump-bonding) both limited to
  200mm or other exotic techniques.
• A lot of semiconductor engineering is needed to
  improve the material quality. Progress, but slow.

• Some material quality problem is still there for
  diamond (sometimes good, not always)
• Kagan showed that (for good samples) the charge
  collection can be 25ke for 0.8mm tickness and
  that 300 mm diamond still gives 2500 e- after
  500 MRad (silicon?)

28

• Sherwood P. gave an inspired talk (with a nice
  historical introduction) on the 3D pixel
  detectors and outlined the characteristics that
  make this design apt for B-layer replacement in
  super-LHC.

depletion after 2 1015
Signal after 1 1015

• Still some more momentum (more people) needed on
  this activity to get it a realistic option for
  such an upgrade

29
Conclusion

• I have found the workshop interesting and useful.

• I have learnt a lot (expecially outside hybrid
  pixel)
• This is possibly because I had to follow all
  talks to prepare the conference summary

• and I have discovered that Bonn is a sunny town
  with mild weather

• (this is a preliminary result based on low
  statistics)

• see you (in FNAL?) for Pixel 2007