Aspects of electron and hole collection in irradiated silicon detectors

1
Aspects of electron and hole collection in
irradiated silicon detectors

• Gregor Kramberger
• Jozef Stefan Institute, Ljubjana, Slovenia
• (on behalf of RD50 collaboration)

2
Outline

• Motivation
• Basics of induced charge in silicon detectors
• Induced current
• EEw options
• Trapping induced charge sharing
• Signature of p readout
• Low Fluence (RD50 micron run mainly)
• Single particle irradiations
• Correlation of Vfd with CCE how well can Vfd
  predict Q-V
• Mixed irradiation
• Trapping probabilities for different particles
• High fluence puzzles
• CCE
• Reverse annealing
• Leakage current
• Conclusions

3
Motivation (I)

• SLHC tracker
• Inner tracker rlt27 cm (ATLAS)
• (pixel detector)
• r5 cm 1016 neq cm-2 90 c. had.
• r12 cm 41015 cm-2 65 c. had.
• r18 cm 1.51015 cm-2 50 c. had.
• r27 cm 1015 cm-2 35 c. had.
• Middle tracker 38ltrlt60 cm
• (short strips, macropixels)
• Required to withstand 91014 cm-2
• mix 30-70 at r38 cm
• 20-80 at r49 cm
• 15-85 at r60 cm
• Outer tracker 60ltr cm
• (long strips, 12 cm)
• Required to withstand 51014 cm-2
• mix gt90 neutrons

(4 layers)
3 cm strips
Complex environment requires approach with
individual solutions not only in geometry sense
but also in material choice!
4
Motivation (II)
FZ p-n
Standard, available, proven most used material
FZ n-n
main motivation can be operated partially
depleted, but
DOFZ n-n
even improved performance for charge hadrons

• PROBLEMS WITH LONG TERM ANNEALING AT LHC
• understood simulation reproduce the data for
  LHC

FZ n-p
Avoid complex double sided processing for n-n
MCz n-p
Combines the benefits of DO and n readout and
single side p.
MCz n-n, p-n
Completely different behavior positive space
charge for charge hadrons opening up additional
options/scenarios
Epi n-p,p-n
5
Motivation (III)

• In this talk only signal as the main limiting
  factor (with given noise) for successful
  operation will be addressed
• Building a tracker is very complex task and it is
  puzzle where many pieces should fit perfectly
  together
• Mechanics
• Cooling
• Connections
• Electronics
• Ability to work in radio-active environment.
  Those who have this experience know it is not
  pleasant experience

6
Induced charge (I)
sensing electrode Uw1
rh finished drift
r0
re finished drift
Uw0
7
Induced charge (II)
ATLAS Strip geometry
Contribution of carriers drifting away from
electrodes
Remember for the diode it is

• The thinner the detector the more similar it is
  to the diode
• equal contribution of electrons and holes the
  read-out side doesnt matter! Any movement of the
  drifting charge regardless of where in the
  detectors results in the same amount of induced
  charge.
• The choice of readout side in terms of signal is
  entirely determined by Uw

8
Induced charge (III)
What if the drift is not completed due to
carriers being trapped!
drift current of a single carrier
geometry factor peaked at electrodes

• electrons get less trapped
• they should drift to the strips/pixels and
• contribute most to Q (n readout for silicon)

• trapping term ( teff,eteff,h )
• drift velocity ( me3mh )

One should also consider the difference between
ve and vh gets smaller at high fields teff,e,h
can depend on E L. Beattie et al, NIM A421
(1999) 502.
9
U V
EEw V/mm2
Neff-1e13cm-3 U-1000V
ATLAS strip
depth mm
Uw
depth mm
depth mm
10
Options for the strip/pixel type and geometry in
Si
(a) very bad
(b) bad
(c) good
p-n-n
n-n-p
p
n
p-p-n
(d) the best
(e) the reference
n-p-p
n
p
diode
If irradiated detectors are fully depleted
(d)gt(b)(c)gt(a) If irradiated detectors are
over-depleted (d)(b)gt(c)(a) We will
investigate this cases for different materials at
SLHC.
11
Trapping induced charge sharing
trapping
absence of trapping

• this effect is far more important in irradiated
  detectors with p strips/pixels
• the amount of charge induced depends also on
  strip/pixel geometry

12

• The trapping induced charge sharing depends on
• temperature (larger for higher T)
• voltage (larger for smaller voltage)
• fluence (larger with fluence)
• position of the track (the effect is much larger
  for tracks close to strip border)
• Strip type p,n

n strips
topology of the clusters at high voltage cluster
charge of all is the approximately the same
p strips
diode
FE threshold
0
Charge sharing determines resolution! Both
mechanisms play a role TICS long range (not
only close to strip border) Diffusion short
range (close to strip border) n device
diffusion () trapping (-) p device diffusion
() trapping ()
n - higher signal in hit electrode p -
wider clusters
13
Residual true position-recalculated
position Resolution RMS of the residual
distribution
D280 mm, T-10oC, U450 V, Feq3x1014 cm-2,
pitch 120 mm
detection threshold
7000 e
6200 e
(noise1800 e)
5000 e
Resolution (noise1800 e) p-n 20 mm
n-n 27 mm
Smaller noise (better S/N) improves resolution
significantly
n optimized for efficiency , p for resolution
14
What about the p readout?There is no good
argument to use them from the point of view of
maximizing detection efficiency there should be
other reasons to do so

• CCE of n is always larger except at VgtgtVfd where
  the cluster charge is comparable - not necessary
  also the efficiency!
• p should be used if you can assure the full
  depletion during entire time you need a safety
  margin!
• An effect of EEw can be nicely seen when n-p
  and p-n are compared

T. Affolder et al., presented at 13th RD50
Workshop, 2008.
See Hartmuts talk ....
Micron RD50 detectors
G. L. Casse et al., presented at 12th RD50
Workshop, 2008.
15
Does the mathematical description give results
that are in agreement with measurements for
entire SLHC?
grazing technique

• Yes, for fluences up to 1015 cm-2
• ATLAS pixel (T. Lari NIM A518(2004) 349.)
• CMS pixel (V. Chiochia IEEE TNS 52(2005) 1067.)
• A general conclusion
• Pronounced double junction
• Trapping at the lower limit of TCT measurements
  (20)
• The measured electric field profile fed back to
  simulation reproduced the data well
• No, for higher fluences will be discussed

test beam
16
Low Fluence Performance(around 2x1015 cm-2 -
defined by understanding the device model)
17
RD50 Micron detectors

• Fz (topsil) and MCz (okmetic) wafers of pn type
  material
• n-on-n, n-on-p, p-on-n structures (pixels,
  strips, diodes)
• Strips ATLAS strips geometry 80 mm pitch
  (w/p1/3)
• Pads 2.5 x 2.5 mm2 , multiple guard rings

Neutron and Proton and Pion irradiation of SSD
and Diodes Systematic studies of Vfd, CCE, Ileak
for strip and pad detectors by Santa Cruz,
Liverpool, Ljubljana Enormous amount of data!
18
Neutron irradiation
open-strips analog, solid-pads, shadow-strips
binary
10 kHz, 20oC
10 kHz, 20oC
80min _at_ 60oC

• The beneficial effect of electron collection at
  low bias voltages, due to strip segmentation
  depletion growth in that region (at high voltages
  the effect becomes smaller saturation of drift
  velocity?)
• the charge on neighbors in binary is often below
  the threshold (high) reduction of visible
  mean charge (binary) most probable charge
  (analog) ? around 20 charge in analog comes
  from neighbors
• Unusually low introduction rate for acceptors for
  MCz gc0.008 cm-1 (factor of 2 lower than for
  all other)

• Very good agreement for strips and pads
• Prediction of CCE based on the C-V works well.
• Q-V is linear before for VltVfd

19
Pion irradiation (the most relevant particles)
10 kHz, 20oC
80min _at_ 60oC

• Annealing of the CV reveals (not shown) that
  predominately negative space charge is introduced
  for all but MCz-n where it is positive!
• Introduction rates of stable damage is
  comparable with previous/other measurements
• Strong acceptor removal for MCz-p after charged
  hadron irradiations.

• Very good agreement for strips and pads!
• Prediction of CCE based on the C-V works well.
• Q-V is linear before for VltVfd

20
Proton irradiation
open-strips analog, solid-pads
10 kHz, 20oC
80min _at_ 60oC

• Only Fz-p shown, but measurements done for all
  points on the left (and more)
• Again CCE for pads and strips agree up to
  4.871015 p cm-2 unexpected Vfd(Fp)
• Evolution of Vfd(Feq) seen also in CCE for strips?

• Very good agreement for strips and pads (Fz-p)!
• Prediction of CCE based on the C-V works well.
• Q-V is linear before for VltVfd

21
Annealing of Vfd for Micron samples
A confirmation of the assumed space charge
Prolonged reverse annealing for MCz with
different space charge for p and n Almost
identical behavior of FZ
22
Linear increase of charge with voltage for under
depleted means that active region grows linearly
with voltage.
wthickness of active area (green) irradiated
bulk highly resistive
fraction of weighting field seen
generated e-h pairs
Typical for Neffconst.? Is that a
coincidence? The double junction should grow in
the same way!
23
Vfd from C-V vs. Vfd from Q-V (pads, strips)

• Vfd from C-V is determined for pad detectors (
  80min _at_ 60oC end of beneficial annealing )
• Vfd from CV underestimates the onset of
  saturation in CCE by max. 100-150 V!
• after Vfd the collected charge continues to
  increase due to shorter drift
• due to growth of depletion depth from electrode
  side the offset is smaller than with p-on-n!
• The correlation holds for all investigated
  fluences in range of full depletion voltages up
  to 1000V!

24
Vfd from C-V vs. Vfd from dQ/dV

• What if we cant get Vfd,CCE due to breakdown. Is
  there a way to evaluate detectors?

For very high fluences account for trapping.
Even without trapping correction a reasonable
agreement can be achieved! A possible new way of
presenting performance of irradiated detectors!
25
Qsat-charge in over-depleted detectors

• For VgtgtVfd the charge loss depends only on
  trapping and not on the electric field vdrift
  approaches vsat

Open-pions Solid-neutrons
RD50 status report 2007
At the same equivalent fluence, charge hadrons
seem to be more damaging thus confirming the
teff measurements dQsat/dFeq600 e/1e14 cm-2 for
neutrons dQsat/dFeq850 e/1e14 cm-2 for
pions dQsat/dFeq800 e/1e14 cm-2 for protons The
measured trapping probabilities from TCT are
around 40-50 too large to give the agreement
with measured saturated charge!
There is no dependence of Qsat on material!
The over depletion is more important at lower
Vfd and less at high Vfd as the ltEgt is already
very high and drift velocities close to saturated
in large part of the detector
26
Q(500V) after short term annealing
FZ mostly

• Best detectors seem to be all on the line is
  that beatable ?
• Below the line are all FZ, except MCz-p which
  suffers from high initial doping concentration
  after neutron irradiation!

27
STABLE damage
NEG. SPACE CHARGE
POS. SPACE CHARGE
the opposite sign of neutron/proton damage
compensation of the SC in mixed fields
opposite sign gc and gY compensation SC during
annealing
, but things are not that simple (complex SC
profile, complex annealing)
28
Mixed irradiations Vfd (24 GeV p neutrons)
24 GeV protons
neutrons
21014 neutrons cm-2
Only 24 GeV p will be shown the behavior
confirmed also for 200 MeV pions
gc can be or -
always

• Proton irradiated samples were added 21014 n
  cm-2
• MCz n Vfd expected to decrease (opposite sign
  of gc)
• MCz-p ?
• Fz Vfd expected to increase (same sign of gc)

29
Mixed irradiations-Vfd

• The similar gc for 24 GeV p and neutrons result
  in linear increase of Vfd for FZ data
• The MCz-n improves after additional neutrons!
• The MCz-p shows that after high p fluence the
  additional damage is less than predicted.
• The shift in voltage corresponds to the expected
  140V!

80 min _at_ 60oC
80 min _at_ 60oC
80 min _at_ 60oC
30
Mixed irradiations leakage current
80 min _at_ 60oC
the line denotes a410-17 A/cm

• Leakage current as always follows linear
  dependence on equivalent fluence

31
Annealing of the Vfd 24 GeV p n
5 y at 20oC
Max. fluence 9e14 cm-2 corresponds to 1/3 of the
total fluence at r12 cm with proper mix!
32
Mixed MCz - CCE and annealing

• CCE confirms Vfd measurements improvement for
  MCz-n deterioration for MCz-p, FZ
• During the entire annealing the Vfd of MCz-n is
  better than proton irradiated sample only

33
Mixed FZ CCE and annealing
Vfd nicely reflects in CCE much worse
performance of Fz detectors! Generally we see
earlier breakdown at late stages of annealing of
Fz/MCz-p samples! Larger Neff -gt higher field
-gt earlier break down ?
34
Mixed - CCE strips

• Improvement also seen for strips
• Note the order of irradiation here was reversed
  first neutron then protons

35
Fluence after completed short term annealing for
different materials
Long term annealing increases the Vfd
Speculative
MCz superior
Long term annealing decreases the Vfd
Assumed complete initial doping removal!
36
High fluence puzzles (defined by
not-understanding the device model)
37
Puzzles at high fluences -100 CCE

• The CCE at fluences shows values that can not be
  explained by extrapolation of the measurements at
  lower fluence. In the calculation of the signal
  either
• active region (electric field) is different than
  expected
• trapping probability decreases
• mobility increases (not likely)
• So far these superb performance seen only with n
  segmented detectors by different groups mainly
  for neutrons!
• Liverpool (SCT 128A) by far the most measurements
• Ljubljana (SCT 128A)
• Valencia
• Santa Cruz (binary 50 ns)
• Pisa (APV 25 ns, unfortunately only up to 600V)

up to Vgt1000 V
38

• Signal vs. Bias Voltage MICRON RD50 n-p FZ
  detectors
• highest voltage limited by breakdown
• SCT 128A, 25 ns shaping, 90Sr signal
• ? 100 CEE seen also
  after 3x1015 n/cm2
• ? 15000 electrons after
  1x1016 n/cm2
• ? very good agreement with
  Lpool up to voltages of 1100 V
• ? break down occurs without fully
  visible saturation of CCE for thick detectors

Blue empty symbols Liverpool (different
annealing stages explain the difference in CCE)
300 µm
T -20C
300 µm
140 µm
Measured at T -40 C
39
Black measured, Red simulation
Black measured, Red simulation
ße 3.210-16 cm2/ns ßh 3.510-16 cm2/ns
No trapping, only Neff
SIMULATION FAILS COMPLETELY! Even if trapping is
off - the active region assumed by depletion is
not enough to reproduce the signal!
40
What about the CCE in pad detectors (investigated
up to 1016 cm-2)?

• Trapping probabilities should be the same
• Electric field in most of the detector also,
  except close to strips
• Alpha TCT (25 ns integration) - 100 CCE also
  observed

J. Lange, Diploma thesis, Uni. Hamburg,2008.

• Vertical bars denote full depletion voltage. The
  onset of charge saturation is matches the Vfd up
  to 41015 cm-2.
• At high fluences and high voltages there are
  indications rapid increase of charge indication
  of charge multiplication?

24 GeV p
41
Possible explanations of high CCE

• tt(E) - change of st,e,h at high fields
• Un-depleted bulk is resistive and electric field
  exists in the entire depth
  (at 1015
  cm-2 and RT, lower at low temperatures)
• Multiplication starts at E10 V/mm
• very likely to be achieved in at least some part
  of detector if ltEgt5 V/mm for under-depleted
  diode
• can happen in high field at n electrodes in
  segmented detectors even higher electric field
  due to implant edges multiplied charge drift in
  high weighting field (see V. Eremin et al., 13th
  RD50 Workshop)
• Higher Neff-gt higher E avalanche more likely

• But
• If 1. and 2. apply why do we see more charge in
  thin at high fluences than for thick sensors
  only possible if gain in trapping compensates
  difference in thickness!
• If 3 applies why dont we see CCE well in excess
  of 1 detector always break down there? What is
  the quenching mechanism to prevent break down
  trapping?
• there should be position dependence of efficiency
  measured in the test beam!
• detectors measured as diodes (strips bonded
  together) should performs better than pads

42
Puzzles at high fluences reverse annealing

• Long term annealing of FZ detectors should be
  harmful, but it is only of the order 30.
• Can be explained by compensation of reduced
  active region by positive annealing of electron
  trapping times G. Kramberger et al., NIM A571
  (2007) 608..
• Saturation of reverse annealing amplitude
  effective concentration of defects during reverse
  annealing saturates. At high fluences therefore
  becomes relatively less important.
• BUT
• Measured charge is larger than expected by
  reduced active region.
• Only in O rich materials was seen the
  saturation (it should be checked again)

Micron FZ n-in-n, 1.5E15 n cm-2
43
Puzzles at high fluences - Reverse current

• Confirms the hypothesis that undeleted bulk is
  resistive enough that sufficient electric field
  is present to separate the carriers

• But, why doesnt then the current depend on
  thickness? For pad detectors it does!

44
Conclusions

• n optimized for efficiency, p for charge
  sharing
• As long as a device model works we have
  reasonable understanding
• Vfd (CV) makes sense good correlation with CCE
• At the same equivalent fluence charged hadrons
  are more damaging
• Charge at 500V is reduced by 1000 e/1e14 cm-2
• Mixed irradiations confirm the expectations of
  additive damage ???
• At high voltage/fluence range we dont understand
  the device model need further studies probing
  electric field! Look for possible multiplication