Jaap Velthuis (University of Bristol)1 Radiation damage in silicon sensors Overview radiation damage due to heavy particles State-of-the-Art sensors In

Jaap Velthuis (University of Bristol) 1

Radiation damage in silicon sensors

• Overview radiation damage due to heavy particles

• State-of-the-Art sensors

In case of any questions:

[email protected]

mailto:[email protected]


sLHC radiation dose

• 5 year radiation dose close to beam pipe ~1016 neq/cm2

– too high for state-of-the-art standard silicon sensors


Radiation with protons/neutrons

• Energetic radiation knocks atoms out of lattice: similar to doping

• Energy needed to displace atom from lattice=15eV

• This damage is called Non-Ionizing Energy Loss (NIEL)

• Displacement changes band structure– Donor removal– Acceptor generation


Radiation damage: Leakage current

• I = Volume• Material

independent– linked to defect

clusters

• Scales with NIEL• Temp dependence

– Thermal runaway1011 1012 1013 1014 1015

eq [cm-2]

10-6

10-5

10-4

10-3

10-2

10-1

I /

V

[A/c

m3 ]

n-type FZ - 7 to 25 Kcmn-type FZ - 7 to 25 Kcmn-type FZ - 7 Kcmn-type FZ - 7 Kcmn-type FZ - 4 Kcmn-type FZ - 4 Kcmn-type FZ - 3 Kcmn-type FZ - 3 Kcm

n-type FZ - 780 cmn-type FZ - 780 cmn-type FZ - 410 cmn-type FZ - 410 cmn-type FZ - 130 cmn-type FZ - 130 cmn-type FZ - 110 cmn-type FZ - 110 cmn-type CZ - 140 cmn-type CZ - 140 cm

p-type EPI - 2 and 4 Kcmp-type EPI - 2 and 4 Kcm

p-type EPI - 380 cmp-type EPI - 380 cm

kT

ETTI g

2exp2 = 3.99 0.03 x 10-17Acm-1

after 80minutes annealing at 60C


Type inversion• Dopants may be

captured into defect complexes.

• Donor removal and acceptor generation– type inversion: n p– depletion width

grows from n+ contact

• Increase in full depletion voltage

biasbi

DA

DAd VV

NqN

NNx

2

0 0.5 1 1.5 2eq [1014cm-2]

1

2

3

4

5

|Nef

f| [1

012 c

m-3

]

50

100

150

200

250

300

Vde

p [V

] (

300

m)

1.8 Kcm Wacker 1.8 Kcm Wacker 2.6 Kcm Polovodice2.6 Kcm Polovodice3.1 Kcm Wacker 3.1 Kcm Wacker 4.2 Kcm Topsil 4.2 Kcm Topsil

Neutron irradiationNeutron irradiation

cNN effeff exp0

= 0.025cm-1 measured afterbeneficial anneal

P-strips in p-bulk


Partially depleted detectors

• Depletion zone grows from p-n-junction. Need depleted area around strips for isolation– Signal proportional depleted

area• Undepleted region like high-

ohmic resistor• If detector partially depleted

– from strip side only charge in depleted region

contributes smaller signal, similar spatial resolution

– from backplane carriers travel towards strips, but

don’t reach it signal spread over many strips poor spatial resolution

undepleted

undepleted


Solutions radiation damage

• N+-on-n detectors (LHCb)– Need full depletion before type

inversion– After radiation p-type bulk

• P-type bulk sensors– Just becomes more p-type

• Cool to cryogenic temperatures – No trapping, no leakage current

• Diamond sensors– Very high bandgap: no background– Very though lattice


Solutions radiation damage

• Oxygenation– Oxygen binds and neutralises

vacancies• Czochralski silicon

– “cheap” Si, contains loads of oxygen– no type inversion donor generation

overcompensates acceptor generation• 3D sensors

– Spacing electrodes so small: full depletion at very low voltages

– “Edgeless”


State-of-the-Art devices

• DEPFETs– Very thin p-n sensor with in-pixel

amplification• ISIS

– Pixel sensor with short CCD each pixel• MAPS• 3d integrated devices

• Trend is towards thin, fast and integration


DEPFET Principle

• Was developed towards ILC, so needs to be very thin and fast• Will be used at SuperBelle • A p-FET transistor (=amplification!) integrated in every pixel.• By sidewards depletion potential minimum created below

internal gate.• Electrons, collected at internal gate, modulate transistor

current

~1µm

p+

p+ n+

rear contact

drain bulksource

p

sym

met

ry a

xis

n+

ninternal gate

top gate clear

n -

n+p+

--

++

++

- 50

µm

------

MIP


DEPFET Principle (II)• Advantages:

– Fast signal collection due to fully depleted bulk

– Low noise due to small capacitance and amplification in pixel

– Transistor can be switched off by external gate – charge collection is then still active !

– Non-destructive readout

• Disadvantages:– Need to clear internal gate.


Ladder proposal

• Detectors 50µm thick, with 300µm thick frame yields 0.11% X0

• SWITCHER & CURO chips connected by bump bonding

• Radiation hardness not an issue: can change operating voltage to correct Vthres shift

SWITCHER

CURO


Thinning

sensor wafer

handle wafer

1. implant backside on sensor wafer

2. bond wafers with SiO2 in between

3. thin sensor side to desired thick.

4. process DEPFETs on top side

5. etch backside up to oxide/implant

first ‘dummy’ samples:50µm silicon with 350µm frame

thinned diode structures:leakage current: <1nA /cm2


Testbeam results• Placed 450µm thick

DEPFET in testbeam • Cluster signal (5σ seed,

2σ neighbour cut) • S/N=112.0±0.3 for 450

μm S/N≈12 for 50 μm

• Position resolution 1.82µm (incl telescope error) for 22µm pitch• Intrinsic resolution 1.25μm• Second best ever

measured!


ISIS• Operational Principles:– Every pixel has mini CCD to store charge: burst camera with

multiframes– Charge collected at photogate– Transferred to storage pixel during bunch train– 20 transfers per 1ms bunch train– Readout during 200ms quiet period after bunch train


ISIS• It works! • Here you see Fe55

spectrum• Results of a laser

scan• And position

resolution in a beam test

• ISIS2 is currently available

preliminarypreliminaryUsing η


MAPS operation principle

• Epitaxial layer forms sensitive volume (2-20m)

• Charge collection by diffusion (no field!)

• Charge collected by N-well• Build amplifiers in P-well

(Intrinsic amplification)– Only NMOS possible

• Small signals (~800e-), but small noise (~15e-)

• Developed for:– SuperBelle, STAR vertex

detector, replacing CCDs in camera’s & satellites

Vreset Vdd

Out

Select

Reset


MAPS (dis)advantages• Advantages:

– Integrated detector and electronics• High S/N (first amplification in pixel)• Possible in-pixel or on-chip intelligence (System on chip)

– Low power consumption– Radiation hardness (w.r.t. CCDs)– Small pixel size (10-20 m)– Thin

• can be less than 20 m• 50µm in industry

– Standard CMOS “cheap”– Room temperature operation– Excellent position resolution

• Disadvantages:– Thin active volume low signals (80 eh pairs /µm)– Smaller CMOS sizes usually yield thinner epilayer thickness


On chip data processing:MIMOSA VIII

• On chip data processing:– TSMC 0.25 µm (8 µm

epitaxial layer)– 32//columns of 128

pixels– 25x25 µm2 pixels– On-pixel CDS– Discriminator on each

column

55Fe


MAPS with storage: FAPS• Active pixel with memory

cells: sample and store charge during bunchtrain can be read out in 200 ms in between trains no high speed readout required!

• FAPS:– 10 memory cells in each pixel– First step of incorporating in-

pixel intelligence– S/Ncell between 14.7±0.4 and

17.0±0.3

• Issue: need 20 C’s per pixel. Small pixels small C’s. Then spread in actual C-values large. Bad for S/N.

FAPS

Column Output

Write amplifier

RST_W

SELA

1

Memory Cell #0

Memory Cell #1

Memory Cell #9

Ibias

Seed 3x3 5x5


MAPS• Problem with MAPS: charge

collection by n-well. So can only make p-mos transistors and electronics.

• Now trying to do proper CMOS in-pixel using deep p-well

• Plan to incorporate signal processing logic inside the pixels!– Store X and Y location (14μm

res. in X and Y)– Digitize seed signal with 5bit

ADC– Get 13 bit time stamp– Sum lower signals for total

cluster charge– Use higher threshold for hit flag– Per strixel only one 32 bit

output word/train


3D integrated devices• New development in electronics

industry• Put memory directly on processor

– Reduces R, L, C– Improves speed– Can optimize technology for each layer

• Problems:– Dies must be same size– Precise alignment is essential


Mechanical bonding techniques

• Direct silicon fusion bonding– Mechanical bond only– High temperature and pressure– Need very flat surface

• Adhesive bonding– Glue– Low temperature


Electrical+mechanical

• Copper to copper fusion bonding– Press copper surfaces

together– Need 400oC

• Copper-tin eutectic bonding– Soldering– Need 250oC


Processing

• Thinning– Wafers can be

easily thinned to 50μm, much thinner (6μm) done

• Making contact– Drill hole– Fill with Cu


3D sensors (example I)

• 3 Layer Infrared camera– HgCdTe sensor– 0.25μm CMOS (analog)– 0.18μm CMOS (digital)


3D (example II)• 3D Laser rader imager

– 64x64 array, 30µm pixels– 3 tiers

• 0.18µm SOI• 0.35 µm SOI• High resistivity substrate

diodes

• Oxide to oxide wafer bonding

• 1.5µm vias• dry etch• 6 3D vias/pixel


Summary radiation hardness

• Radiation damage in sensors mainly bulk damage– Atoms knocked out of their lattice

position extra levels in band gap • Effectively donor removal (type inversion)• High leakage currents

– High noise– Thermal runaway

• Problems to get full depletion


Summary radiation hardness (II)

• Solutions:– n+-on-n or even better n-on-p detectors– Material engineering (oxygenated

Si/Cz)– Cool to cryogenic temperatures

(Lazarus effect)– Use different materials like diamond– Use different detector type like 3D


Summary

• Trend towards more integration– Sensor and electronics in same device– In-device, or in-pixel signal processing– Faster, smaller feature sizes– Less material


Summary• Tried to show that Particle Physics is more

than hunting for Higgs and CP violation• Forefront of

– Engineering (stiff light weight support structures, cooling, tunnel building)

– High speed and radiation hard electronics– Computing (web, grid, online)– Accelerators (e.g. cancer therapy, diffraction)– Imaging sensors (e.g. nth generation light

source, medical imaging)

• If you find these things interesting, why don’t you join us?

Particle Physics

Thanks for attention

Hope you found it interestingenough to stay awake

Documents

Jaap Velthuis (University of Bristol)1 Radiation damage in silicon sensors Overview radiation damage due to heavy particles State-of-the-Art sensors In