1

Introduction to Silicon Detectors Marc Weber, Rutherford Appleton Laboratory

RAL Graduate Lectures, October 2008

•Where are silicon detectors used?

• How do they work?

• Why silicon?

• Electronics for silicon detectors

• Silicon detectors for the ATLAS experiment

• Radiation-hardness

• Future

2

Where are silicon detectors used?

in your digital Cameras to detect visible light

A basic 10 Megapixel camera is less than $150 …

3

in particle physics experiments to detect charged particles

Example: ATLAS Semiconductor Tracker (SCT); 4088 modules; 6 million channels

1 billion collisions/sec

Up to 1000 tracks

4

in astrophysics satellites to detect X-rays

Example: EPIC p-n CCD of XMM Newton New picture of a supernova observed

in 185 AD by Chinese astronomers

5

in astrophysics satellites to detect gamma rays

11,500 sensors350 trays18 towers

~106 channels83 m2 Si surface

INFN, Pisa

6

Silicon detectors are used at many other places

• in astrophysics satellites and telescopes to detect visible and infrared light, X ray and gamma rays

• in synchrotrons to detect X-ray and synchrotron radiation

• in nuclear physics to measure the energy of gamma rays

• in heavy ion and particle physics experiments to detect charged particles

• in medical imaging

• in homeland security applications

What makes silicon detectors so popular and powerful?

7

1. Incident particle deposits energy in detector medium positive and negative charge pairs

(amount of charge can vary wildly from ~100 – 100 M e, typical is 24,000 e = 4 fC)

2. Charges move in electrical field electrical current in external circuit

Most semiconductor detectors are ionization chambers

How to chose the detection medium ?

Operation principle ionization chamber

8

Desirable properties of ionization chambers

Always desirable: signal should be big; signal collection should be fast

for particle energy measurements: particle should be fully absorbed

high density; high atomic number Z; thick detector

Example: Liquid Argon

for particle position measurements: particle should not be scattered

low density; low atomic number; thin detector

Example: Gas-filled detector; semiconductor detector

Typical ionization energies for gases 30 eV

for semiconductor 1-5 eV

You get (much) more charge per deposited energy in semiconductors

9

Semiconductor properties depend on band gap

Small band gap conductor

Very large charge per energy, but

electric field causes large DC current >> signal current

Charged particle signal is “Drop of water in the ocean”

This is no good. Cannot use a piece of metal as a detector

Large band gap insulator (e.g. Diamond)

Little charge per energysmall DC current; high electric fields.

This is better. Can build detectors out of e.g. diamond

Medium band gap semiconductor (e.g. Si, Ge, GaAs)

large charge per energy

What about DC current ?

10

Semiconductor basics

When isolated atoms are brought together to form a crystal lattice, their wave functions overlap

The discrete atomic energy states shift and form energy bands

Properties of semiconductors depend on band gap

11

Semiconductor basics

Intrinsic semiconductors are semiconductors with no (few) impurities

At 0K, all electrons are in the valence band; no current can flow if an electric field is applied

At room temperature, electrons are excited to the conduction band

There are too many free electrons to build detectors from intrinsic semiconductors other than diamond

Si Ge GaAs Diamond

Eg[eV] 1.12 0.67 1.35 5.5

ni (300K) [cm-3] 1.45 x 1010 2.4 x 1013 1.8 x 106 < 103

12

How to detect a drop of water in the ocean ? remove ocean by blocking the DC current

Most semiconductor detectors are

diode structures

The diodes are reversely biased

only a very small leakage current

will flow across it

~ 150V

Streifen- oder Pixel-Elektroden

Operation sequence

Charged particle crosses detector

+charged particle electrodes

Positive voltage

Ground

~ 150V

Streifen- oder Pixel-Elektroden

Operation sequence

Creates electron hole pairs

-

-+

+

+

+ -

-

-

-

+

+

+

~ 150V

Streifen- oder Pixel-Elektroden

Operation sequence

these drift to nearest electrodes position determination

-

-+

+

+

+ -

-

-

-

+

+

+

16

Components of a silicon detector

- Silicon sensor with the reversely biased pn junctions

- Readout chips

- Multi-chip-carrier (MCM) or hybrid

- Support frame (frequently carbon fibre)

- Cables

- Cooling system

+ power supplies and data acquisition system (PC)

Let’s look at a few examples now before moving on with the talk

17

Detector readout electronics

Typically the readout electronics sits very close to the sensor or on the sensor

Basic functions of the electronics:

• Amplify charge signal

typical gains are 15 mV/fC• Digitize the signal

in some detectors analog signals are used• Store the signal

sometimes the analog signal is stored• Send the signal to the data acquisition system

The chips are highly specialized custom integrated circuits (ASICs)

18

• Noise performance

output noise is expressed as equivalent noise charge [ENC]

ENC ranges from 1 e- to 1000 e-;

for strip detectors need S/N ratios > 10

• Power consumption

typical power of strip detectors is 2-4 mW/channel; for pixels at LHC 40-100 W/pixel; elsewhere can achieve << 1 W/pixel

• Speed requirements range from 10 ns to ms

• Chip size smaller and thinner is usually best• Radiation hardness needed in space, particle physics and elsewhere

These requirements are partially conflicting; compromise will depend on specific application

Critical parameters for electronics

19

Number of transistors per chip increases exponentially due to shrinking size of transistors

Unfortunately the fixed costs (NRE) increase for modern technology;

bad for small-scale users like detector community

Moore’s Law

20

• ATLAS SLHC silicon area: >150 m2; CMS LHC: 200 m2 today; GLAST: 80 m2; variants of CALICE (MAPS): 2000 m2

• Industry is achieving incredible performance for sensors

However there are not many vendors and SLHC is tougher

Silicon strip sensors

p-in-n; 6 inch wafers;

300 m thick; AC- coupling;

RO strip pitch 80 m;

Area: 4x9.6 cm2;

Depl. voltage: 100-250 V

K. Hara; IEEE NSS Portland 2004

21

The SVX readout chip family

SVX’1990

SVX21996 SVX3

1998

SVX42002

• Increasing feature size makes chips smaller

• Adding new features (e.g. analog-to digital conversion; deadtime-less readout) makes them bigger

The SVX2 was a crucial ingredient to the top quark discovery at the Tevatron collider at FNAL near Chicago

22

Multi-chip-carrier/hybrid

• carries readout chips and passive components (resistors and capacitors

• distributes power and control signals to chips; routes data signals out

• filters sensor bias voltage

Typically have 4 conductor layers separated by dielectric/insulation layers

Size: 38 mm x 20 mm x 0.38 mm

Example: ceramic BeO hybrid for the CDF detector

23

4-chip hybrid: top layer

Package efficiency: 31%; 30 passive components; material: 0.18% rad. length; no technical problems; yield on 117 hybrids: 90% (after burn-in)

24

Critical parameters for hybrids

• want low-Z material and small feature size and thickness (minimize multiple scattering)

• good heat conduction to cooling tubes

• reliability/ high yield

• good electrical performance

25

“Packaging is what makes your cell phone small”

How to stack sensors; MCMs; chips; CF support; cables and cooling while connecting them electrically, thermally and mechanically ?

Packaging

Cell phone,

Digital camera,

PDA, Web access,

Outlook

3D packaging

26

Technological challenges: Pixel detector

• innovative packaging of sensor/chips/support structure/cooling

- sophisticated, crowded flex-hybrid

- carbon-carbon support structures

- bump-bonding of chips to sensors

- direct cooling of chips

• Global and local support structures: stiff; lightweight; precise; “zero” thermal expansion

27

Technological challenges: Pixel detector

• Bump-bonding of chips to sensors:

pitch of only 50 μm (commercial pitches 200 μm)

28

Packaging solution for SCTStill very compact

- flex-hybrid with connectors

- separate optical readout for each module

- separate power for each module

- cooling pipes not integrated to structure

29

Radiation-hard sensors1. Radiation induced leakage current independent of impurities; every 7C

of temperature reduction halves current cool sensors to -25C (SCT = -7C)

2. “type inversion” from n to p-bulk

increased depletion voltageoxygenated silicon helps (for protons);

n+-in-n-bulk or n+-in-p-bulk helps

3. Charge trappingthe most dangerous effect at high fluences

collect electrons rather than holes

reduce drift distances

30

Strong candidate for inner layer: 3D pixels• 3D pixel proposed by Sherwood Parker in 1985

• vertical electrodes; lateral drift; shorter drift times; much smaller depletion voltage

• Difficulty was non-standard via process; meanwhile much progress in hole etching; many groups; simplified designs

see talk of Sabina R. (ITC-irst)

3D planar

31

0

20

40

60

80

100

0 5 1015 1 1016 1.5 1016 2 1016

Sig

nal

eff

icie

ncy

[%

]

Fluence [n/cm2]

0 8 1015 1.6 1016 2.4 1016 3.2 1016

Fluence [p/cm2]

3D silicon C. DaVia et a. March 06

Diamond W. Adam et al. NIMA 565 (2006) 278-283

n-on-p strips P. Allport et al.IEEE TNS 52 (2005) 1903

n-on-n pixels CMS T. Rohe et al. NIMA 552(2005)232-238

C. Da Via'/ Aug.06

Signal loss vs. fluence see C. da Via’s talk at STD6 “Hiroshima” conference

3D pixels perform by far the best

Large Hadron Collider: the world’s most powerful accelerator

7 TeV protons vs. 7 TeV protons; 27 km circumference

7 x the energy and 100 x the luminosity of the Tevatron

ATLAS detector

ATLAS detector

• Huge multi-purpose detector; 46 m long; diameter 22 m; weight 7000 t

• Tracking system much smaller; 7 m long; diameter 2.3 m; 2 T field

ATLAS Silicon Tracker

17 thousand silicon sensors (60 m2 )

6 M silicon strips (80 m x 12.8 cm)

80 M pixels (50 m x 400 m)40 MHz event rate; > 50 kW power

2 m 5.6 m

1 m

1.6 m

What’s charged particle tracking ?

1. Measure (many) space points/hits of charged particles

2. Sort out the mess and reconstruct particle tracks

Difficulty is:

- not to get confused

- achieve track position

resolution of 5-10 m

…it’s not easy !Up to 1000 tracks

1 billion collisions/sec

Status as of October 2006

37

How does it look in real life ? SCT Detector• 4 barrel layers at 30, 37, 45, 52 cm radius and 9 discs (each end)

• 60 m2 of silicon; 6 M strips; typical power consumption 50 kW

• Precision carbon fiber support cylinder carries modules, cables, optical fiber, and cooling tubes

• Evaporative cooling system based on C3F8 (same for pixel detector)

Barrel 6 at CERN

38

Why tracking at LHC is tough ?• Too many particles in too short a time

- 1000 particles / bunch collision

- too short: collisions every 25 ns

• Too short need fast detectors and electronics; power!

• Too many particles

- need high resolution detectors with millions of channels

- detectors suffer from radiation damage

to date this requires silicon detectors

39

Example

Need many channels to resolve multi-track patterns

Expect 30-60 M strips and >100 M pixels

40

Extreme radiation levels !• Radiation levels vary from 1 to 50 MRad in tracker volume

- less radiation at larger radii; more close to beam pipe

- more radiation in forward regions

• Fluences vary from to 1013 to 1015 particles/cm2

• Vicious circle: need silicon sensors for resolution and radiation hardness cooling (sensors and electronics) more material even more secondary particles etc.

Don’t win a beauty contest in this environment, but detectors are still very good !

41

Extreme radiation levels !Plots show radiation dose and fluence per high luminosity LHC year for

ATLAS (assuming 107 s of collisions; source: ATL-Gen-2005-001)

Fluence [1 MeV eq. neutrons/cm2] Radiation dose [Gray/year]

“Uniform thermal neutron gas” Put your cell phone into ATLAS ! It stops working after 1 s to 1 min.

• Neutrons are everywhere and cannot easily be suppressed

42

The Boring masks the Interesting HZZ ee + minimum bias events (MH= 300 GeV)

LHC in 2008 ?? : 1032 cm-2s-1 LHC first years: 1033 cm-2s-1

LHC: 1034 cm-2s-1 SLHC: 1035 cm-2s-1

43

Why are silicon detectors so popular ?

• Start from a large signal

good resolution; big enough for electronics• Signal formation is fast

• Radiation-hardness

• SiO2 is a good dielectric

• Ride on technological progress of Microelectronics industry

extreme control over impurities; very small feature size; packaging technology

• Scientist and engineers developed many new concepts over the last two decades

44

Technologies come and go

Random examples are

• Bubble chamber

45

Technologies come and go

Steam engines

46

Silicon detectors are not yet going!Future detectors are being designed and will be

• Larger: 200-2000 m2 • More channels: Giga pixels• Thinner: 20 m• Less noise• Better resolution

Your next digital camera will be better and cheaper as well

47

Appendix

48

Missing Material for future talksThis would require >2h talk rather than 1h

• What is doping• What is a pn junction• Show HS fig 1.14; 2.19; 2.20• Say what a MIP is• Cover photon and x-ray stuff • More semiconductor fundamentals poisson equation, limits

of approximation• What is a FET• Explain radiation damage for FETs• Wafer processing and wire-bonding etc.

49

• Semiconductor band structure ’ energy gap• Asymmetric diode junction: example p(+) into contact with n (Na>>Nd)• Space charge region formed by diffusion of free charges, can be increased with "reverse bias“

Silicon Detectors

p+ nV=0 VRB>0

W

bias reverse applied V) 0.8(~ potentialin built

101e

1 material n type ofy resistivit

11.9 mobility,electron

5.02 :idthjunction w

0

RBBI

D

RBBIRBBI

VV

cmkN

VVmVVW

50

Point Resolution

• Discrete sensing elements (binary response, hit or no hit), on a pitch p, measuring a coordinate x

• Discrete sensing elements (analog response with signal to noise ratio S/N) on a pitch p, where f is a factor depending on pitch, threshold, cluster width

12

px

)(~S

Nfpx

p

x

51

Position Sensitive StructureSingle sided configurationPixel readout

Diagram courtesy of Z.Li and V.Radeka

Charge sensitive Preamp + shaping,signal processing, pipelines, digitization

52

x

xya

x

yby

y

by

y

bb

xbyx

xxyyyya

x

yy

xx

yyb

bxay

yyy

xx

81

2

2

21

2121

2

121

2

1intercept

21

21

21slope

errors with points, measured 2,1

planest measuremen 2,1

22

22

2

for good resolution on angles ( and ) and intercepts (d, z0 )Precision track point measurementsMaximize separation between planes for good resolution on interceptsMinimize extrapolation - first point close to interaction

Vertex Resolution

x1 x2

y1

y2

a

53

x

xya

yyy

xx

81

2

errors with points, measured 2,1

planest measuremen 2,1

for good resolution on angles ( and ) and intercepts (d, z0 )Precision track point measurementsMaximize separation between planes for good resolution on interceptsMinimize extrapolation - first point close to interactionMaterial inside 1st layer should be at minimum radius (multiple scattering)

Vertex Resolution: Material

x1 x2

y1y2

a

54

Signal Processing Issues

• Signal: expressed as input charge, typically 25,000 pairs (4 fC)

• Leakage current: blocked DC by oxide, but RB CB >>TM

– Before (after) radiation damage ~ 1nA(1ma)

– AC component is seen by pre-amp

• Noise fluctuations ~ Gaussian N – Leakage Current– Preamp “input noise charge”, white

noise, decreases with pre-amp current, increases with faster risetime where a,b are constants and CD is the detector capacitance

– Bias resistor: source of thermal noise

MLEAKN TI

DN bCa

BIASN R

1

V bias

R bias

integrator shaper

t rise TM

V out

n+

np+

oxideCb

G

55

pre-amp leakage

(ignore scale, just an example)

56

Silicon Collider Detectors• 1st generation: LEP and CDF vertexers, L=1029

– 2-4 layers, single sided DC coupled silicon or early double sided, ~50K channels– Charge integration + S/H, analog readout, 3 m radsoft CMOS and NMOS– Rad soft components (~25 KRad)

• 2nd generation: LEP and CDF , L=1030 (~100 KRad)– AC coupled detectors, improved double sided structures– Rad hard components, 1.8 m radhard CMOS– Early pixel implementations

• 3rd generation: CDF2a, D0, and B factories , L=1031 (few MRads, 1012-1013/cm2)– Early examples of trackers– Complex double sided constructions, ~500K channels– On chip storage pipelines, ADC’s, digital readout, 0.8 m radhard CMOS

• 4th generation: ATLAS, CMS trackers, CDF2b , L=1032-34 (~10 MRads, 1014-1015/cm2)– Large scale systems (5-10M channels), uniform designs, mass construction methods– Return to single sided detectors (radiation hardness and HV operation: SSC/LHC R&D)– New IC processes (Maxim, DMILL, 0.25 m), fast front ends, deep pipelines– Engineered, large pixel systems for vertexing

• 5th generation: New trackers for L=1035 (~100 MRad, 1014-1016/cm2)– Very large scale systems, simplifications– New rad hard sensor structures and materials– Lower mass supports and services– Increased azimuthal AND longitudinal segmentation, pixel structures move to larger radii– Further evolution of IC (0.13,0.09 m, heterostructures…) technology

First Cosmic Particle Tracks (Summer 2006)

Excellent build precision! Resolution will improve after alignment and for higher momentum tracks