development of backside illuminated silicon photomultipliers at the mpi semiconductor laboratory

H-G MoserMax-Planck-Institute for

PhysicsSemiconductor

Lab

PD07Kobe, JapanJune 27-29

Development of backside illuminated Silicon

Photomultipliers at the MPI Semiconductor Laboratory

Outline:

-Motivation for a backside illuminated SiPM (BID-SiPM)-Layout of BID-SiPM-R&D Program-Results and Simulations (preliminary)-Applications in Particle Physics

on behalf of the MPI SiPM group

MPEMax-Planck Institut

für extraterrestrische Physik

WHIMax-Planck Institut für

Physik(Werner-Heisenberg-

Institut)

PNSensor GmbH



Lab


Motivation for backside illumination

In general SiPM offer great advantages compared to photomultiplier tubes:

Simple, robust devicePhoton counting capabilityEasy calibration (counting)Insensitive to magnetic fieldsFast response (< 1 ns)Large signal (only simple amplifier needed)competitive quantum efficiency (~ 40% at 400-800 nm)No damage by accidental lightCheapLow operation voltage (40 – 70 V)

R&D goal:

Increase Quantum efficiency to the physical limit



Lab


QE & Fill Factor

What limits the QE?

QE = surface transmission x Geiger efficiency x geometrical fill factor

Front illuminated devices:

Large area blinded by structures– Al-contacts– Bias-resistor– Guard rings/Gap between HF implants

For 42 x 42 m2 device: 15% fill factor

Solutions: – larger pixel size (80% reached for 100 m pitch device)– back-illumination light enters through homogeneous back side, not covered by any structure

3 m light spot scanned across device



Lab


Geiger Efficiency of electrons and holes

Avalanche Efficiency (1 m high field region)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

250000 350000 450000 550000 650000 750000

Field (V/cm)Ef

ficie

ncy

Electrons

Holes

Electrons have a higher probability to trigger an avalanche breakdown then holes

Efficiency depends on depth of photon conversion and hence on the wavelength

Solutions:-Increase overvoltage

Or: - Ensure that only electrons trigger an avalanche

holes

p-substrate

p- epip+

n+

el.

n-substrate

n- epin+

p+el.

holes



Lab


Sensitivity at different wavelengths

light absorption in Silicon

0.001

0.01

0.1

1

10

100

1000

10000

250 450 650 850 1050

Wavelength (nm)

Ab

sorp

tio

n le

ng

th ( m

)

Thin entrance window needed

holes electrons

Holes triggeravalanche

Electrons triggeravalanche

p-substrate: photons < 450 nm: only holes contributephotons > 700 nm: lost in insensitive bulk

n-substrate: ok for short wavelengths,hole efficiency dominates for > 500 nm

Back illumination: whole thick (> 50 m) bulk absorbs photonsdesign for electron collection

Example:p-substrate



Lab


Concept of a back illuminated SiPM

Combine SiPM and drift (photo) diode:

Each Pixel of a SiPM array is drift diode with a geiger APD as amplifying element in the center

By drift rings the electrons from photon conversions are focused into a small HF region

Homogeneous sensitivity, no dead regions

Back Illuminated Drift SiPMBID-SiPM



Lab


Design of the avalanche cell

The HF region is created between a n+ contact at the surface and a deep p well underneath. By modulating the depth of the p-implant and/or the n-contact the HF region can be confined to a small area of a few m diameter

-> Small HF region: Low capacitance, low gain (important to fight cross talk)



Lab


Engineering of Entrance Window

(Calculation: R. Hartmann)

Homogeneous, unprocessed thin entrance window at backside- minimal UV absorption in surface layer (important for < 350 nm)- possibility of antireflective coating



Lab


Disadvantages

• Large volume for thermal generated currents (increased dark rate)

Maintain low leakage currentsCoolingThinning ( < 50 m instead of 450 m)

• Large volume for internal photon conversion (increases x-talk)

Lower gain (small diode capacitance helps)

Possible show stopper!

• Electron drift increases time jitter

Small pixels, Increased mobility at low temperature<2 ns possible



Lab


Design of Devices

Hexagonal Cells

100-200 m diameterUp to 3 three drift rings

Central HF region with <8 m diameter

Capacitance ~ 5 fFGain: O(105)~ 1 m depth

95% Geiger efficiency @ 8V overvoltage (electrons)

Drift field extends into bulk



Lab


Test structure production in 2005

-> fix parameters of avalanche cell (radius, depth, resistor values…)-> no backside illumination yet

Single pixel structuresSmall arraysLarge arrays (20 x 25 pixel180 m pitch)HF diameter: 5-25 m

Successfully tested



Lab


Results: Test Structures

LowMediumHigh

Results with light pulses from a laser (< 1 ns):Photoelectron peaks clearly resolved up to large n(photon) RMS of single photoelectron signal ~ 5%



Lab


Results: Test Structures

T = 0oC T = 10oC T = 20oC

Gain proportional to overvoltageBreakdown voltage in good agreement with device simulations



Lab


Test Structures

Dark rate mainly from highly doped HF region: For a 5 x 5 mm2 matrix with 500 pixels: ~0.2 MHz @ 20oC (8V)



Lab


Leakage Currents and Dark Rates

Back side illuminated: bulk leakage current dominates:

For devices thinned to 50 m: ~10MHz @ 20oCCooling needed: ~ 1 MHz @ 0oC



Lab


Top Wafer

a) oxidation and back side implant of top wafer

b) wafer bonding and grinding/polishing of top wafer

c) process passivation

open backside passivation

d) deep etching opens "windows" in handle wafer

Processing thin detectors (50 m)

Successfully tested with MOS diodes (keep low leakage current ~ 100 pA/cm2)



Lab


Cross Talk Studies

Dark spectrum of 25 m arrays

x-talk heavily suppressed due to small HF region and large pitchBackground due to pile up (suppressed by cooling to -20 C)

2pe signal clearly visible:

Probability for x-talk ~ 10-4 (@*V U)

For backside illumination: Bulk is sensitive to cross-talk photons

Use MC to extrapolated to full structure

1 pe~ 2x106

2 pe< 200

V

en

trie

s



Lab


Monte Carlo Simulation of cross talk

0

0.05

0.1

0.15

0.2

0.25

1 2 3 4 5 6 7 8 9 10 11

n(photon)

p(n

)

2.9

Generate photons

Propagate through device

Photon Converts1. In pixel of origin2. In neighbour pixels

1. Active region2. Apply geiger efficiency

e e

drift drift



Lab



Generate photons




Number of photons poisson distributedLacaita, IEEE TED, 40 (1993): 2.9 photons/105 e- (E > 1.14 eV)

Use black body spectrum with T=4300K

-However: el/holes not in thermal equilibrium

-Band structure not taken into account(2.9 ph > 1.14 eV => 8.5 ph total)



Lab



Generate photons




Calculate photon absorption lengthFrom Photon energy

1 mm

100 m

e e

Surface reflection given by n=3.57



Lab



Generate photons




drift drift

Electrons: drift in HF region (if bulk depleted)Apply local Geiger efficiency (as function of overvoltage)



Lab



Generate photons




Cross talk spectrum



Lab


Contribution from photons with range from O(pitch) – O(mm)

Cross talk spectrum



Lab


larger pitch -> x-talk spectrum narrowLarger pitch -> less x-talk

Dependence on Pitch



Lab


Results

Cross talk measured with test structures implies:

~54 photons (E>1.14 eV) per avalanche (@*V U)

For a backside illuminated device with 100 m pitch:

cross-talk probability: 99.99%

Due to large capacitance (47 fF of HF region + coupling capacitances), the gain is very high: ~4 x 106

Scaling to the expected gain of 105:

Cross talk 20-30 %

Still high but could be manageable

Extrapolation has large systematic error!



Lab


Further Program

2nd step: production of fully functional backside illuminated SiPMs:

-> including drift rings-> double sided processing, deplete bulk

Finished: End 2007

Various test structures(single pixel, small arrays)

Arrays:30 x 31 pixel Diameter HF region: < 8 mPitch: 100, 120, 150, 200 mArea: 3x3 mm2 – 6x6 mm2

In addition: some front illuminated arrays



Lab


Applications in (Astro-) Particle Physics

Main advantages: Drawbacks:High QE Dark rateGood sensitivity for Cross talk 300nm < < 1000nm Costs

(double sided processing)

Can be used where QE and spectral response are important but high dark rate can be tolerated

•Air cerenkov telescopes - peak wavelength 300-350 nm (like MAGIC) - considerable LONS background



Lab


Applications

• Cerenkov detectors in particle physics detectors

• Compact high density calorimeters

low light yield,

direct coupling to blue scintillator

• In colliding beam experiments: dark rate can be suppressed

using coincidence with beam

no self trigger needed (cross

talk!)

• Large scale applications (big calorimeters) probably excluded (costs)

Practical advantage:

Direct coupling of entrance window to scintillator (no wire bonds)

Scintillator

SiPMReadout board



Lab


Smart SiPMs

clock

x, y, t

SiPM

BiCMOS analogue

CMOS digital

Connect ASIC chip to back-illuminated SIPMFor each pixel: signal detection & active quenching

-Fast timing (no stray capacitances)-Low threshold -> low gain-Active quenching -> low gain-Minimal cross-talk-Single pixel position resolution-Veto of noisy pixels

Could be made using

•Bump bonding•3D integration techniques



Lab


Summary

Backside illuminated Silicon Photomultipliers are developed at the MPI Semiconductor LaboratoryApplication: Upgrade of Magic Camera

Aim for highest quantum efficiency in a large spectral range (only limited be quality of entrance window)

First test structures (not yet back illuminated) produced and tested

Production of full devices ongoing

Design goals:

QE: 80% for 300 – 950 nm, peaking at > 90%Dark rate: <1MHz for 0.25 cm2 device @ 0oCGain: 105

Cross talk: 20-30 % ?

High cross talk is of major concern!

development of backside illuminated silicon photomultipliers at the mpi semiconductor laboratory

Documents