a. nepomuk otte mpi für physik, munich / humboldt universität, berlin

The Geiger-APD

a novel photon detector and its application in

astrophysics experiments and

positron emission tomography

A. Nepomuk OtteMPI für Physik, Munich / Humboldt Universität,

Berlin

Max-Planck-Institut für Physik / Humboldt Universität Berlin 2A. Nepomuk Otte

Outline

• why new photon detectors for experiments in astroparticle physics

• the G-APD and some of its characteristics

• application of G-APD in:– positron emission tomography (PET)– air Cherenkov telescopes


Many future experiments will use >> 100,000 photon detectors

• robust and stable• easy to calibrate• blue sensitive• low cost (+ low peripheral costs)• compact• low power consumption• …• highest possible photon detection efficiency

Astroparticle experiments that will use this photon detector

Requirements to be fulfilled by the photon detector candidate:


Cosmic Ray Physics from Space

30°

400 k

m

ČerenkovFluorescence

EECR

230 km

Earth

Atmosphere

M.C.M. ‘02

30°

400 k

m

ČerenkovFluorescence

EECR

230 km

Earth

Atmosphere

M.C.M. ‘02

Atmospheric SoundingAtmospheric Sounding

http://www.euso-mission.org/

• Highest energy cosmic rays > 1020 eV• GZK mechanism• sources of CR• …


Ground based Gamma Ray Astrophysics

http://wwwmagic.mppmu.mpg.de/

Gamma Ray induces electromagnetic cascade

relativistic particle shower in atmosphere

Cherenkov light

fast light flash (nanoseconds)100 photons per m² (1 TeV Gamma Ray)

MAGIC: world largest air Cherenkov telescope


VHE gamma-ray sources: status ICRC 2007

Rowell

71 known sources

very successful above 100GeV

factor of 6 increase within 4 years


pushing to lower energies

entering an unexplored energy window between 10 GeV and 100 GeV

• extragalactic background light studies• gamma ray bursts• dark matter• tests of quantum gravity• pulsars• …

requires:

• larger light collectors• high efficiency photon detectorshigh efficiency photon detectors

currently used: classical photomultiplier tubes with ~20% QE


The G-APDa promising photon detector concept invented in Russia in the 80’s

P. Buzhan et al. http://www.slac-stanford.edu/pubs/icfa/fall01.html

• sensors with ~60% efficiency become available• internal gain ~105 -106

• compact and robust• …

advantages

disadvantages

• small sizes (<5x5mm²)• optical crosstalk (10%)• high dark count rate (~MHz)• …

Otte et al., IEEE TNS. 53 (2006) 636.

SNIC-2006-0018, Apr 2006


3x3 mm² G-APD


A look into basic operations of

semiconductor photon detectorswith

internal amplification


• Bias: (10%-20%) ABOVE

breakdown voltage

• Geiger-mode: it’s a BINARY

device!!

• Count rate limited

• Gain: “infinite” !!

• Bias: slightly BELOW breakdown

• Linear-mode: it’s an AMPLIFIER

• Gain: limited < 300 (1000)

• High temperature/bias dependence

• No single photo electron resolution

Working modes of avalanche photodiodes

Geiger ModeLinear/Proportional Mode


Advantages of APDs in Geiger Modeor

Single Photon Avalanche Diodes (SPADs)

• Large standardized output signalhigh immunity against pickup

• High sensitivity for single photons

• Excellent timing even for single photo electrons (<<1ns)

• Good temperature stability

• Low sensitivity to bias voltage drifts

• Devices operate in general < 100 V

• Complete insensitive to magnetic fields

• No nuclear counter effect (due to standardized output)


The principal disadvantage for many applications:

It is a binary device

One knows: There was at least one electron/hole initiating the breakdown

but not

how many of them

solved in the G-APD concept


Basic unit in a G-APD is a Single Photon Avalanche Diode (SPAD)

Breakdown in SPAD is quenched by individual polysilicon resistor (passive quenching)

from B. Dolgoshein (ICFA 2001)http://www.slac.stanford.edu/pubs/icfa/


The G-APD

typically 100…2000 small SPADs / mm²

All SPADs connected in parallel

30µm

1mm

Bias andOutput

Only one common signal line

small signal replacement circuit

quenching resistor

SPAD


SiPM output is the analog sum of all SPADs

Well defined output signal per SPAD multi pixel resolution


Dynamic Range

Dynamic range naturally limited by number of available SPADs

working condition:Number of photo electrons < SPAD cells

From probability considerations:

1 10 100 1000 100001

10

100

1000

Number of pixels fired

Number of photoelectrons

576 1024 4096

from B. Dolgoshein Light06

wor

king

rang

e

20% deviation from linearity if 50% of cells respond


Photon Detection Efficiency (PDE)or

Effective Quantum EfficiencyMost important parameter of a photon detector!!

limiting factors:

• Fraction of sensitive area (20% - 80%)

• Intrinsic quantum efficiency

• Surface reflection losses

• Probability for Geiger breakdown (depends on electric field)

W.Oldham, P.Samuelson, P.Antognetti, IEEE Trans. ED (1972)

In total: Currently claimed best PDE values are ~60%

• SPAD recovery time (passive quenching)

• Active volume / absorption length


Measurement of the Photon Detection Efficiency

PDE measurements are not an easy task– optical crosstalk– dependency on bias voltage

often a photomultiplier with unknown photoelectron collection efficiency is used as reference

Overestimation of the PDE


A method to measure the PDE

Otte et al., NIM A 567 360–363, 2006

use calibrated PiN-diode as reference

use integrating sphere with two exit ports (splitting ratio of several thousand)

flash PiN-diode and G-APD with pulsed monochromatic light source


Problems:

Optical Crosstalk

High Dark Count Rate


• SPADs not only detect photons they also emit photons during breakdown

Emission microscopy picture of a prototype SiPM

Optical Crosstalk


Optical crosstalk

Artificial increase in signal

Excess Noise Factor of SiPM

Photons can trigger additional cells

Sketch from Cova et al. NIST 2003Workshop on single photon detectors

can be quite significant


Using optical crosstalk to learn more about the photons emitted in

avalanches


Light Emission in Avalanches

• measured spectra do not show similar behavior• emission mechanisms not well known• very few absolute measurements

W. J. Kindt


optical crosstalk spectrum from dark noise

Try to reproduce this distribution with Monte

Carlo simulations


SiSi: The SiPM Simulator

*Elisabeth ”Sisi” von Wittelsbach was the empress consort of Emperor Franz Joseph of Austria. She was born 1837 in Munich, Bavaria and murdered 1898 in Geneva, Switzerland


SiSi

SiSi is an “almost” complete simulator of a SiPM

photoelectrons in non-depleted bulk are subject to simple diffusion model;lifetime of electrons is a free parameter

simulation of avalanche photons:

• black body radiation with free parameters:

- temperature- intensity

• isotropic emission


Example of a good match

residuals

Residuals can be explained by dark counts which are not simulated in SiSi

no unique set of model parameters (Temperature and Intensity of photon spectrum)


Characteristics of Photons that cause Optical Crosstalk

Intensity (1.15 … 1.40 eV): ~3*10-5 photons / electron (systematic uncertainty of ~2)

FWHM: ~0.2 eVPeak: ~1.26eV

2 photon spectra that reproduce the measured crosstalk distributions

energy distribution of

photons causing optical crosstalk

Crosstalk is only caused by photons within a narrow

range of energies


Reason for narrow range of photon energies

strong dependence of absorption lengths on photon energy

absorption in same cell

not absorbed in G-APD


Two possible applications for G-APDs:

1. positron emission tomography (PET)

2. air Cherenkov telescopes


Positron Emission Tomography (PET)Basic principle

PET : image distribution of a radio-labeled tracer inside the body

Tracer Molecules : Labeled by positron emitting isotopes ( 11C, 13N, 15O, 18F)

Positron annihilation : two back-to-back 511 keV photons

The emitted photons are detected by two opposing detectors in coincidence

γ

γObject containing some quantity of radio-labeled tracer

(positron-source)

detectors


what you are looking for!

This is...

The Reconstruction Problem


G-APDs in PET: the first studies

Advantage: very compact, no sophisticated amplifier needed, …

Otte, et al. NIM A 545 (2005)

•direct coupling of SiPM to crystal

•no cooling

•Factor 4 area miss match between SiPM and crystal

wrapped crystals

G-APDs

signal readout


G-APDs in PET: the first studies

Advantage: very compact, no sophisticated amplifier needed, …

Otte, et al. NIM A 545 (2005)

•direct coupling of SiPM to crystal

•no cooling

•Factor 4 area miss match between SiPM and crystal

• Energy resolution 22% FWHM on 22Na coincidence spectrum

• Time resolution 1.5 nsec FWHM

Things have quite improved since then

first evermeasurement


Result of measurements with MW-3 (3x3 mm2) Geiger- mode APDs from Dubna (Z. Sadygov) + LYSO crystals (2x2x10 mm3)

0 200 400 6000

500

1000

1500

2000

150 200 2500

500

1000

1500

2000

Amplitude (pC)

Co

un

ts

511 keV : A/A = 12.7% (FWHM)1275 keV : A/A = 7.7% A

1275 / A

511 = 2.60

22Na + LSO (2x2x10 mm3; reflector = teflon)

MW-3 (3x3 mm2, n.1): RT, U = 138.0V, I = 1.05A

Alexey Stoykov,Dieter Renker (PSI)

Energy Resolution:12% FWHM

Time Resolution:540ps

(limited by crystal)

MRS diode used

(2006)


Imaging Air Cherenkov Technique

~ 10 kmParticleshower

~ 1o

Ch

eren

kov

ligh

t

~ 120 m

Gammaray

Cherenkov light image of particle shower in telescope camera

• fast light flash (nanoseconds)• 100 photons per m² (1 TeV Gamma

Ray)

reconstruct: arrival direction, energyreject hadron background


Figure of Merit

Cherenkov spectrum folded with photon detector

response

Cherenkov spectrum on ground photon detector response


Application of G-APDs in air Cherenkov telescopes

Cherenkov spectrum folded with photondetector

response

Figure of Merit

MPPC from Hamamatsu with highest PDE

(from data sheet MPPC with 100x100 µm² cells)

photomultiplier tubes

factor >2 increase in sensitivity compared to

bialkali PMTs

can hope for


Test on La Palma with MAGIC

4 MPPC-33-050C from Hamamatsu:

sensor size: 3x3mm²single cell size: 50x50µm²nominal bias: 70.4Vdark rate at nominal bias: ~2MHzgain at nominal bias: 7.5*105

crosstalk at nominal bias: 10%

peak photon detection efficiency 55%

needs to be confirmed

Array of 4 MPPCs:light catchers with factor 4 concentration; 6x6mm² onto 3x3mm²

MAGIC Pixel Size


array mounted next to the MAGIC camera for 3 nights for fine tuning and tests

• array not removed or protected during day

• it was raining for one day!

G-APDs signals recorded by the MAGIC DAQ with each trigger


Array mounted onto the MAGIC camera entrance window for two nights


Light recorded from Calibration Runs

UV-LEDs 375nm

Pedestal

1 phe

2 phe

3phe

…

single phe-resolution degraded because of light from night sky

easy calibration

some recorded showers

reduced systematics


location of MPPC array

1 phe

MPPCs

PMTs

2 phe 4 phe 1 phe

70 phe 35 phe 35 phe 15 phe


Shower Signals: MPPC vs PMT

event selection:two PMTs next to MPPCs with more than 15 photoelectrons in each tube

signals are correlated

cou

nts

~300 events from ~30 min data

on average MPPC record a larger signal


ratio of signals MPPC/ (scaled) PMT

event by event

on average 1.6 times more light detected with MPPCs (crosstalk corrected)

100% efficiency assumed for the light catcher in front of the

MPPCs

in reality higher due to non perfect light concentrator


Summary

The silicon photomultiplier is a real breakthrough in photon detection!!

CMOS like technology prospects for cheap mass production <10$ per mm²

It can not be damaged by exposure to strong source of light

Offers high internal amplification (>105)

Fast timing (<nsec)

No aging

Low power consumption (1…100µW/mm²)

High photon detection efficiency (>60%)

…


Summary

In PET G-APDs already outperform other photon detectors

+ they allow to build very compact and robust scanners insensitive to magnetic fields combination of MRT and PET

PMTs in Cherenkov telescopes could be replaced by G-APDs

• at least 1.6 times more light recorded• G-APD intrinsic noise is 10 times lower than the “noise”

coming from the night sky• optical crosstalk is an issue for IACTs and should be reduced to

below 5%


The End


Ongoing Development:SiPM exploiting Backillumination

predicted characteristics:

• PDE > 80%• Single photo electron time jitter ~ nsec• Cooling is mandatory

By the Semiconductor Laboratory affiliated to the MPIs for Physics and Extraterrestrial Physics

Si

photon

depleted bulkavalanche regions

path of the photo electron

output

50µm … 450µm

Blow up of one “cell”


drift rings p+

shallow p+

avalanche region

photondrift path of the photo electron

µm100

µmµm 450...50

quenching resistor

output line

deep n

n type depleted bulk

• test structures of novel avalanche structure will be finished next month• After successful evaluation prototypes end 2007

Crosstalk problem can be a showstopper!!will be evaluated by dedicated structuressmall cell capacitance is of advantage

a. nepomuk otte mpi für physik, munich / humboldt universität, berlin

Documents

photon detectorrequirements

nepomuk otteground

nepomuk ottepushing

photon detectors robust

photon detector candidate

nepomuk ottecosmic ray

gamma ray astrophysics

degamma ray