Rainer Wallny

1
Silicon Detectors How They Work

• Rainer Wallny
• Silicon Detector Workshop at UCSB
• May 11th, 2006
• Slides ruthlessly stolen from
• Paula Collins, CERN
• Alan Honma, CERN
• Christian Joram, CERN
• Michael Moll, CERN
• Steve Worm, RAL

2
Outline

• Why Silicon ?
• Semiconductor Basics
• Band-gap, PN junction
• Silicon strip detectors
• Some Technicalities
• - Wafer Production
• - Wire Bonding
• Radiation Damage
• Effect on Vd
• Effect on Leakage Currents
• Conclusions

3
Tracking Chambers with Solid Media

• Ionization chamber medium could be gas, liquid,
  or solid
• Some technologies (ie. bubble chambers) not
  applicable in collider environments

Solid-state detectors require high-technology
devices built by specialists and appear as black
boxes with unchangeable characteristics.
-Tom Ferbel, 1987
3
4
Why Silicon?

• Electrical properties are good
• Forms a native oxide with excellent electrical
  properties
• Ionization energy is small enough for easy
  ionization, yet large enough to maintain a low
  dark current
• Mechanical properties are good
• Easily patterned and read out at small dimensions
• Can be operated in air and at room temperature
• Can assemble into complex geometries
• Availability and experience
• Significant industrial experience and commercial
  applications
• Readily available at your nearest beach

4
5
The Idea is Not Quite New
6
Pioneering Silicon Strip Detectors
6
7
Contemporary Silicon Modules
CDF SVX IIa half-ladder two silicon sensors
with readout electronics (SVX3b analog readout
chip) mounted on first sensor
ATLAS SCT barrel module four silicon sensors
with center-tapped readout electronics (ABCD
binary readout chip)
7
8
Large Contemporary Silicon Systems
DELPHI (1996) 1.8m2 silicon area 175 000
readout channels
CMS Silicon Tracker (2007) 12,000 modules
223 m2 silicon area 25,000 silicon wafers
10M readout channels
CDF SVX IIa (2001-) 11m2 silicon area 750 000
readout channels
8
9
New Production Paradigm
evolution DELPHI (888 detectors, 8
geometries) CDF (8000
sensors, 8 geometries) CMS
(25000 sensors, 15 geometries)
Each sensor treated individually, nurtured into
life in many hours of careful handling
P. Collins, Warwick 2006
10
Moores Law for Silicon Detector Systems
11
Large Silicon Detector Systems .
12
The Basics
13
Semiconductor Basics Band Gap
14
Semiconductor Basics Principle of Operation

• Basic motivation charged particle position
  measurement
• Use ionization signal (dE/dx) left behind by
  charged particle passage

15
Doping Silicon

• n-type
• In an n-type semiconductor, negative charge
  carriers (electrons) are obtained by adding
  impurities of donor ions (eg. Phosphorus (type
  V))
• Donors introduce energy levels close to
  conduction band thus almost fully ionized gt
  Fermi Level near CB
• Electrons are the majority carriers.

• p-type
• In a p-type semiconductor, positive charge
  carriers (holes) are obtained by adding
  impurities of acceptor ions (eg. Boron (type
  III))
• Acceptors introduce energy levels close to
  valence band thus absorb electrons fromVB,
  creating holes gt Fermi Level near VB.
• Holes are the majority carriers.

16
The pn-Junction

• Exploit the properties of a p-n junction (diode)
  to collect ionization charges

When brought together to form a junction, a
gradient of electron and hole densities results
in a diffuse migration of majority carriers
across the junction. Migration leaves a region of
net charge of opposite sign on each side, called
the depletion region (depleted of charge
carriers). Electric field set up prevents
further migration of carriers resulting in
potential difference Vbi Another way to look at
it Fermi-Levels need to be adjusted so thus
energy bands get distorted gt potential Vbi
17
pn - Junction














18
How to Build a Silicon Detector
19
Properties of the Depletion Zone

• Depletion width is a function of the bulk
  resistivity , charge carrier mobility and the
  magnitude of reverse bias voltage Vb

Vd d2 /(2???)
20
Properties of the Depletion Zone (contd)

• One normally measures the depletion behavior
  (finds the depletion voltage) by measuring the
  capacitance versus reverse bias voltage. The
  capacitance is simply the parallel plate capacity
  of the depletion zone.

capacitance vs voltage
1/C2 vs voltage
21
Leakage Current

• - Two main sources of (unwanted) current flow in
  reversed-biased diode
• Diffusion current, charge generated in undepleted
  zone adjacent to depletion zone diffuses into
  depletion zone (otherwise would quickly
  recombine)

negligible in a fully depleted device
Exponential dependence on temperature due to
thermal dependence of e-h pair creation by
defects in bulk. Rate is determined by nature and
concentration of defects.
21
22
Bias Resistor and AC Coupling

• Need to isolate strips from each other and
  collect/measure charge on each strip gt high
  impedance bias connection (resistor or
  equivalent)
• Usually want to AC (capacitatively) couple input
  amplifier to avoid large DC input from leakage
  current.
• Both of these structures are often integrated
  directly on the silicon sensor. Bias resistors
  via deposition of doped polycrystalline silicon,
  and capacitors via metal readout lines over the
  implants but separated by an insulating
  dielectric layer (SiO2 , Si3N4).

23
The Charge Signal
dE/dx)Si 3.88 MeV/cm, for 300 mm thick 116
keV This is mean loss, for silicon detectors use
most probable loss (0.7 mean) 81 keV 3.6eV
needed to make e-h pair Collected charge 22500 e
(3.6 fC)
24
But There Is Noizzzzzz ..

• Landau distribution has significant low energy
  tail which becomes even lower with noise
  broadening.
• Noise sources
• Capacitance ENC Cd
• Leakage Current ENC v I
• Thermal Noise ENC v( kT/R)

One usually has low occupancy in silicon sensors
?most channels have no signal. Dont want noise
to produce fake hits so need to cut high above
noise tail to define good hits. But if too high
you lose efficiency for real signals. Figure of
Merit Signal-to-Noise Ratio S/N. Typical
Values 10-15, people get nervous below 10.
Radiation Damage can degrade the S/N. Thus S/N
determines detector lifetime in radiation
environment.
25
Charge Collection and Diffusion

• Drift velocity of charge carriers v ?E, so
  drift time, td d/v d/?E
• Typical values d300 ?m, E 2.5kV/cm, ?e
  1350?h 450 cm2 / Vs, gives td(e) 9ns ,
  td(h) 27ns

26
Double Sided Detectors
Why not get a 2nd coordinate by measuring
position of the (electron) charge collected on
the opposite face?

• SOLUTION
• Put p-strips in between the n-strips.
• OR
• Put field plates (metal over oxide) over the
  n-strips and apply a
  potential to repel the electrons.

27
Guard Rings and Avalanche Breakdown
We have treated the silicon strip device as
having infinite area, but it has edges. What
happens at the edges?
27
28
Some Technicalities
29
Material Float Zone Silicon (FZ)

• Mono-crystalline Ingot

• Float Zone process

? Using a single Si crystal seed, meltthe
vertically oriented rod onto the seed using RF
power and pull the single crystal ingot

• Wafer production? Slicing, lapping, etching,
  polishing

• Oxygen enrichment (DOFZ)? Oxidation of wafer at
  high temperatures

30
Czochralski silicon (Cz) Epitaxial silicon (EPI)

• Czochralski silicon

• Pull Si-crystal from a Si-melt contained in a
  silica crucible while rotating.
• Silica crucible is dissolving oxygen into the
  melt ? high concentration of O in CZ
• Material used by IC industry (cheap)
• Recent developments (2 years) made CZ available
  in sufficiently high purity (resistivity) to
  allow for use as particle detector.

Czochralski Growth

• Epitaxial silicon

• Chemical-Vapor Deposition (CVD) of Silicon
• CZ silicon substrate used ? in-diffusion of
  oxygen
• growth rate about 1mm/min
• excellent homogeneity of resistivity
• up to 150 mm thick layers produced
• price depending on thickness of epi-layer but
  not extending 3 x price of FZ
  wafer

31
Wafer Processing (1)
SiO2
Photolithography ( mask align photo-resist
layer developing) followed by etching to make
windows in oxide
UV light
etch
mask
32
Wafer Processing (2)
B
As
Al
Photolithography followed by Al
metallization over implanted strips and over
backplane usually by evaporation.
33
Bringing It All Together

• Connectivity technology some of the
  possibilities
• High density interconnects (HDI)industry
  standard and custom cables, usually flexible
  kapton/copper with miniature connectors.
• Soldering still standard for surface mount
  components, packaged chips and some cables.
  Conductive adhesives are often a viable low
  temperature alternative, especially for delicate
  substrates.
• Wire bonding the standard method for connecting
  sensors to each other and to the front-end chips.
  Usually employed for all connections of the
  front-end chips and bare die ASICs. A mature
  technology (has been around for about 40 years).

4 x 640 wire bonds
200 wire bonds
Total 2700 wire bonds
34
Wire Bonding

• Uses ultrasonic power to vibrate needle-like tool
  on top of wire. Friction welds wire to metallized
  substrate underneath.
• Can easily handle 80?m pitch in a single row and
  40?m in two staggered rows (typical FE chip input
  pitch is 44?m).
• Generally use 25?m diameter aluminium wire and
  bond to aluminium pads (chips) or gold pads
  (hybrid substrates).
• Heavily used in industry (PC processors) but not
  with such thin wire or small pitch.

35
Radiation Damage
36
Radiation Damage in Silicon Sensors

• Two general types of radiation damage to the
  detector materials
• ? Bulk (Crystal) damage due to Non Ionizing
  Energy Loss (NIEL) - displacement
  damage, built up of crystal defects
• Change of effective doping concentration (higher
  depletion voltage,
  
  under- depletion)
• Increase of leakage current (increase of shot
  noise, thermal runaway)
• Increase of charge carrier trapping (loss of
  charge)
• ? Surface damage due to Ionizing Energy Loss
  (IEL) - accumulation of positive in the
  oxide (SiO2) and the Si/SiO2 interface
  affects interstrip capacitance (noise
  factor), breakdown behavior,
• Impact on detector performance and Charge
  Collection Efficiency (depending on detector
  type and geometry and readout electronics!)
• Sensors can fail from radiation damage by virtue
  of
• Noise too high to effectively operate
• Depletion voltage too high to deplete
• Loss of inter-strip isolation (charge spreading)
• ? Signal/Noise Ratio is the
  quantity to watch !

37
Run I Experience SVX Signal-to-Noise

• Radiation Damage limits the ultimate lifetime of
  the Detector
• Need S/N gt8 to perform online b-tagging with SVT
• Need S/N gt5 for offline b-tagging

38
Surface Damage
Metal (Al)
-
-
-

• Surface damage generation over time
• Ionizing radiation creates electron/hole pairs in
  the SiO2
• Many recombine, electrons migrate quickly away
• Holes slowly migrate to Si/SiO2 interface.
• Hole mobility is much lower than for electrons
• (20 cm2/Vs vs. 2x105 cm2/Vs)
• Some holes stick in the boundary layer
• Surface damage results in
• Increased interface trapped charge (see picture)
• Increase in fixed oxide charges
• Surface generation centers

-
-

Oxide (SiO2)

-

-
-
-




Interface (SiOx)
Semiconductor (Si)





After electron transport


-




After transport of the holes

• Electron accumulation under the oxide interface
  can alter the depletion voltage (depends on
  oxide quality and sensor geometry)
• In silicon strip sensors, surface damage effects
  (oxide charge) saturate at a few hundred kRad

39
Bulk Damage
Vacancy/Oxygen Center

• Bulk damage is mainly from hadrons displacing
  primary lattice atoms (for E gt 25 eV)
• Results in silicon interstitial, vacancy, and
  typically a large disordered region
• 1 MeV neutron transfers 60-70 keV to recoiling
  silicon atom, which in turn displaces 1000
  additional atoms
• Defects can recombine or migrate through the
  lattice to form more complex and stable defects
• Annealing can be beneficial, but
• Defects can be stable or unstable
• Displacement damage is directly related to the
  non-ionizing energy loss (NIEL) of the
  interaction
• Varies by incident particle type and energy
• Normalize fluence to 1 MeV n-equivalent

O
Vacancy
Disordered region
Interstitial
C
Carbon Interstitial
Carbon-Carbon Pair
C
C
Di-vacancy
Phosphorous dopant
Carbon-Oxygen pair
P
O
C
40
Microscopic defects

• Damage to the silicon crystal Displacement of
  lattice atoms

EKgt25 eV
Vacancy Interstitial
point defects, mobile in silicon,can react
with impurities (O,C,..)
EK gt 5 keV
point defects and clusters of defects
Distribution of vacancies created by a 50 keV
Si-ion in silicon (typical recoil energy for 1
MeV neutrons) SchematicVan Lint 1980
SimulationM.Huhtinen 2001

80 nm

• Defects can be electrically active (levels in the
  band gap) - capture and release electrons and
  holes from conduction and valence band
• ? can be charged - can be generation/recombination
  centers - can be trapping centers

41
Impact of Defects on Detector properties
Inter-center charge transfer model (inside
clusters only)
Shockley-Read-Hall statistics (standard
theory)
Trapping (e and h)? CCEshallow defects do not
contribute at room temperature due to fast
detrapping
charged defects ? Neff , Vdepe.g. donors in
upper and acceptors in lower half of band gap
generation ? leakage currentLevels close to
midgap most effective
enhanced generation ? leakage current ?
space charge
Impact on detector properties can be calculated
if all defect parameters are known?n,p  cross
sections ?E ionization energy
Nt concentration
42
Radiation Damage Effect on Neff

• Change of Depletion Voltage Vdep (Neff)
  . with particle fluence
  

Type inversion Neff changes from positive to
negative (Space Charge Sign Inversion)
before inversion
n
n
p
p
after inversion
43
Depletion Voltage Death of SVX Layer 0

Steve Worm, Vertex 2003
Central Prediction
1s Prediction
300
1s Prediction
Data Extrapolation
Depletion Voltage (V)
200
100
0
0
4
6
2
8
Integrated Luminosity (fb1)
SVXII L0 lifetime prediction based on Hamburg
Model (M.Moll) - Will SVXII L0 survive Run II ?
-gt Antonios Talk
44
Radiation Damage Leakage Current

• Change of Leakage Current (after hadron
  irradiation) . with particle
  fluence

80 min 60?C

• Damage parameter ? (slope in figure)
  Leakage current
  per unit volume
  and particle fluence
• ? is constant over several orders of fluenceand
  independent of impurity concentration in Si ?
  can be used for fluence measurement

45
Recent Bias Current (Re-) Analysis
L0
P.Dong et. al, upcoming CDF note
46
Radiation Damage Trapping

• Deterioration of Charge Collection Efficiency
  (CCE) by trapping

Trapping is characterized by an effective
trapping time ?eff for electrons and holes
where
Increase of inverse trapping time (1/?) with
fluence
47
Decrease of Charge Collection Efficiency

• Two basic mechanisms reduce collectable charge
• trapping of electrons and holes ? (depending on
  drift and shaping time !)
• under-depletion ?
  (depending on detector design and geometry !)
• Example ATLAS microstrip detectors fast
  electronics (25ns)

• n-in-n versus p-in-n - same material, same
  fluence- over-depletion needed

• p-in-n oxygenated versus standard FZ- beta
  source- 20 charge loss after 5x1014 p/cm2 (23
  GeV)

48
Oxygenation Benefits

• oxygenation increases radiation hardness
• sometimes, standard FZ exhibits similar
  radiation hardness - reasons unclear
• Concentrate RD on CZ and EPI silicon

Michael Moll, IWORID Glasgow 2004
49
Summary

• Have fun in California - You deserve it!

50

• Backup

51
Sensor Materials Diamond, SiC and GaN

• Wide bandgap (3.3eV)
• lower leakage current than silicon
• SignalDiamond 36 e/mmSiC
  51 e/mmSi 89 e/mm
• more charge than diamond
• Higher displacement threshold than silicon
• radiation harder than silicon (?)

RD on diamond detectorsRD42
Collaborationhttp//cern.ch/rd42/
Recent review P.J.Sellin and J.Vaitkus on behalf
of RD50 New materials for radiation hard
semiconductor detectors, submitted to NIMA
52
Microscopic defects

• Damage to the silicon crystal Displacement of
  lattice atoms

EKgt25 eV
Vacancy Interstitial
point defects, mobile in silicon,can react
with impurities (O,C,..)
EK gt 5 keV
point defects and clusters of defects
Distribution of vacancies created by a 50 keV
Si-ion in silicon (typical recoil energy for 1
MeV neutrons) SchematicVan Lint 1980
SimulationM.Huhtinen 2001

80 nm

• Defects can be electrically active (levels in the
  band gap) - capture and release electrons and
  holes from conduction and valence band
• ? can be charged - can be generation/recombination
  centers - can be trapping centers

53
Radiation Damage in Silicon
Close proximity to the interaction region means
the sensors are subject to high doses of radiation

• Two general types of radiation damage
• Bulk damage due to physical impact within the
  crystal
• Surface damage in the oxide or Si/SiO2
  interface
• Cumulative effects
• Increased leakage current (increased Shot noise)
• Silicon bulk type inversion (n-type to p-type)
• Increased depletion voltage
• Increased capacitance
• Sensors can fail from radiation damage by virtue
  of
• Noise too high to effectively operate
• Depletion voltage too high to deplete
• Loss of inter-strip isolation (charge spreading)
• Ratio of signal/noise is the important quantity
  to watch

54
Bulk Damage Depletion Voltage

• Depletion voltage is often parameterized
• in three parts (Hamburg model)
• DNeff(T,t,F) NA NC NY
• Short term annealing (NA)
• NA Feq iga,iexp(-ka,i(T)t)
• Reduces NY (beneficial)
• Time constant is a few days at 20 C
• Stable component (Nc)
• Nc Nc0(1-exp(-cFeq))gcFeq
• Does not anneal (does not depend on time or
  temperature)
• Partial donor removal (exponential or limited
  exponential)
• Creation of acceptor sites (linear)
• Long term reverse annealing (NY)
• NY NY,81-1/(1 NY,8kY(T)t), NY,8 gYFeq
• Strong temperature dependence
• 1 year at T20 C is the same as lt1 day at T60 C
  or 100 years at T -7 C (ATLAS)
• Can be significant long term must cool Si

S
Figure 9 Dependence of Neff on the accumulated 1
MeV neutron equivalent fluence for standard and
oxygen enriched FZ silicon irradiated with
reactor neutrons (Ljubljana), 23 GeV protons
(CERN PS) and 192 MeV pions (PSI).  
Fig.13 Annealing behaviour of the radiation
induced change in the effective doping
concentration ?Neff at 60?C.
55
Bulk Damage Leakage Current

• Defects created by bulk damage provide
  intermediate states within the band gap
• intermediate states act as stepping stones of
  thermal generation of electron/hole pairs
• Some of these states anneal away the bulk
  current reduces with time (and temperature) after
  irradiation
• Annealing function a(t)
• Parameterized by the sum of several exponentials
  aiexp(-t/ti)
• Full annealing (for the example below) reached
  after 1 year at 20ºC
• At low temperatures, annealing effectively stops

56
Bulk Damage Effects (Simple View)

• Leakage Current
• DI a(t)FV
• Current depends on a(t) (annealing function), V
  (volume), and F (fluence).
• Annealing reduces the current
• Independent of particle type
• Depletion Voltage
• Vdep qNeffd2/2ee0
• Depends on effective dopant concentration (Neff
  Ndonors Nacceptors), sensor thickness (d),
  permitivity (ee0).
• Depletion voltage is often parameterized in three
  parts
• Short term annealing (Na)
• A stable component (Nc)
• Long term reverse annealing (NY)

57
Bulk Damage Leakage Results

• Measured values of a(t)
• Typically one quotes measured values of a(t)
  after complete annealing at T20ºC a8 a(t8)
• Some typical world averages for a8 are
• 2.2 x 10-17 A/cm3 for protons, pions
• 2.9 x 10-17 A/cm3 for neutrons
• Recent results show a(t80min,T60ºC) 4.0 x
  10-17 A/cm3 for all types of silicon, levels of
  impurities, and incident particle types (NIM A426
  (1999)86).

58
Silicon Detectors How They Work

• Rainer Wallny
• Silicon Detector Workshop at UCSB
• May 11th, 2006
• Slides ruthlessly stolen from
• Paula Collins, CERN
• Alan Honma, CERN
• Christian Joram, CERN
• Michael Moll, CERN
• Steve Worm, RAL