Aug 14, 2009 NGC/CSTC, Hamilton, ON 1

Semiconductor Scintillators and Three-Dimensional Integration

• Isotope identification spectroscopic energy resolution

• Direction to sourceangular resolution

critical needs:

Aug 14, 2009 NGC/CSTC, Hamilton, ON 2

X-ray (γ-ray) attenuation

1 mm

1 cm

1 dm

absorption length

Aug 14, 2009 NGC/CSTC, Hamilton, ON 3

1E-3 0.01 0.1 1 10

0.1

1

10

100

1000

10000

100000

1 mm

SiGe

CdTeInP

Atte

nuat

ion

Coe

ffici

ents

(cm

-1)

hν (MeV)

Element Z

Si 14×2

Ga/As 31/33

Ge 32×2

In/P 49/15

Cd/Te 48/52

X-ray (γ-ray) attenuation by materials

Aug 14, 2009 NGC/CSTC, Hamilton, ON 4

Semiconductors and scintillators

ThalliumEC

EV

hν

= 3 eVEG

= 7 eV

NaI scintillator

> 200 nS

38,000 ph/MeV

Si or Ge pin

diode

cm

> 100 nS

77K

> 10 kV

up to 300,000 e-h/MeV

p

n

Aug 14, 2009 NGC/CSTC, Hamilton, ON 5

Gamma spectroscopy

One measures the number N of electrons and holes produced by incident gamma particle

N ~ Eγ

That number fluctuates.

If the statistics of N

were Poisson,

var

(N) = N

but due to correlations (in semiconductors)

var

(N) =

F N

the Fano factor, F ≈

0.1

Aug 14, 2009 NGC/CSTC, Hamilton, ON 6

Non-proportionality in scintillators

Luminescence in dielectric scintillators is controlled by reactions nonlinear in N (exciton formation, Auger, etc.)

This is one of the reasons γ

spectroscopy with scintillators is not as accurate as it is with semiconductor (diodes).

In semiconductor scintillators, every reaction on the way to luminescence is linear

in the N

of minority carriers

Expect

no non-proportionality effects!

S. A. Payne et al (LLNL group), preprint, 2009

Need: N ~ Eγ

but ...

Aug 14, 2009 NGC/CSTC, Hamilton, ON 7

3D scintillator array

Semiconductor scintillators, each endowed with its own photoreceiver

10 × 10 × 10 array contemplated

Enables both

isotope discrimination and determination of the direction to source

A different way of determining Eγ

(unlike γ

spectroscopy)

Aug 14, 2009 NGC/CSTC, Hamilton, ON 8

3D pixellation of response to a single γ

photon

Upon analog-to-digital conversion each unit reports not a 1 ns pulse but an information-carrying signal:

• where ionization occurred

• time of the event

• amplitude of the event

Photosensitive Layer

γ

- Sensing

Semiconductor

Scintillator

Stack of Scintillator

Slabs

2D pixel

γ

Aug 14, 2009 NGC/CSTC, Hamilton, ON 9

Compton “telescope”

⎟⎟⎠

⎞⎜⎜⎝

⎛−+=

101

111cosEE

θ

iii

iii

EEEE

−=Δ−+=

−

−−−

1

1111cosθ

keV) 5112 =cme(in units of

The energy E0

of the incident γ-photon

Compton kinematics:two equations at each interaction site

i

The

incident

direction

⎟⎟⎠

⎞⎜⎜⎝

⎛−Δ

+Δ+Δ

+Δ=2

222

210 cos1

421

2 θE

θ3

θ1

Δ1

θ2

E0

1n̂

0n̂Δ2

γ

Δ3

Aug 14, 2009 NGC/CSTC, Hamilton, ON 10

What is needed

• Semiconductor scintillator “transparent”

to its own luminescence

• Integrated (optically tight) surface photoreceiver systemof slightly smaller bandgap

• Readout ASICcustomized to the photoreceiver system

The triad:

I will focus on the first two legs of the triad

Aug 14, 2009 NGC/CSTC, Hamilton, ON 11

Semiconductor scintillator: transparency

• Moss-Burstein shift

• Photon re-absorption suppressed

• Radiative decay time ≈

10−9

s

• Need material transparent to its own fundamental light emission

• Photons must be delivered to the surface

EC

EV

EF

hνInP

“Conventional”

transparency

Aug 14, 2009 NGC/CSTC, Hamilton, ON 12

Need for optically-tight photoreceiver

θ0InP

Air

Total Internal Reflection

escape cone

θ0 =17º

2% for free-space detectorsnn

/1sin3.3

0 ==

θ

023.02sinsin41 02

0

0

≈⎟⎠⎞⎜

⎝⎛=∫ θθϕθ

π

θ

dd

escaping fraction of photons

Optically-tight integrated detectors collect the entire scintillating radiation

Epitaxial InGaAsP diodes on InP

Aug 14, 2009 NGC/CSTC, Hamilton, ON 13

Epitaxially integrated pin diode

p+ n–

n+

i

n+

InP wafer

2 μmquaternary InGaAsP

Si3

N4

patterned resistp metal contact

n metal contact backside patterning

epitaxial layers

bare scintillator slab

Sarnoff – SBU collaboration

Aug 14, 2009 NGC/CSTC, Hamilton, ON 14

Characteristics of quaternary epi diodes

(Q) GE

(InP) GE

AEFE

)cm10(for eV 0.23

(designed) eV 1.24300K) (Q;(expt) eV 1)300(

315F

G

A

−≈=

=≈

nE

EKE

as determined by CV

profiling both estimate jibe !

IV Characteristics of Diodes at 300 and 77 K

200

400

600

800

10001200

1400

1600

1800

2000

‐2 ‐1.5 ‐1 ‐0.5 0 0.5 1 1.5

Voltage (V)

Current (pA) 295 K

77 K

0.56 V

⎟⎟⎠

⎞⎜⎜⎝

⎛ +−∝

kTmeVTEI

2)(exp A

< 10 pA

at 300 K (1 pA

in best diodes)

Light collection efficiency ≈

85%

Aug 14, 2009 NGC/CSTC, Hamilton, ON 15

Read-Out Circuits

Output digital readout data

ASIC

IPD

CDET

Peak detector

Preamplifier

Pulse-shapingPhotodetector

ADC

Vss

( ) LKshMOSm

th2MOSDET

2 238ENC qIa

fCKa

gkTaCC ff τ

τ+⎥

⎦

⎤⎢⎣

⎡⋅

+⋅+=

shot noisethermal noiseENC –

currently: 3 ×

103

Next generation: < 103 ☺

low

ILK

≈

10 pAhigh CDET≈ 50 pF

Aug 14, 2009 NGC/CSTC, Hamilton, ON 16

Single-quanta response

Train of scintillator pulses recorded as voltage waveform in the read-out circuit

α-particles from 241Am

α-particle from 241Am

γ-photon from 137Cs

Aug 14, 2009 NGC/CSTC, Hamilton, ON 17

Making semiconductor transparent

• Moss-Burstein shiftSuccess “mixed”

• Photon recyclingNearly ideal non-transparent scintillator

• Subband luminescence centersE.g., Yb3+ luminescent ion in InP ( 1μm emission)

• Impregnated “guest-host”

structuresNon-layered two-phase random systems

... to its own fundamental luminescence

... new ideas are welcome

Aug 14, 2009 NGC/CSTC, Hamilton, ON 18

Transparency: theory vs. reality

EF

EG

Conduction band

Valence band

hν

EC

EV

Momentum non-conservation in heavily-doped InP

“Average”

(over the emission spectrum) photon mean free path is about 0.1 mm

1.3 1.4 1.5Photon energy (eV)

Emission spectrum

Log Transparency

EF

no free lunch

Aug 14, 2009 NGC/CSTC, Hamilton, ON 19

Luminescence Experiments

Excitation beam

Reflection Luminescence

Transmission Luminescence

Monochromator InP wafer

Monochromator 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.600.0

0.2

0.4

0.6

0.8

1.0

TransmissionLuminescence

ReflectionLuminescence

Transm'nSpectrum

hv, eV

300 K

InP, ND

=6.3×1018

cm-3

(S)

Heavily doped n-type InP wafers from Nikko materials (ACROTEC)

1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.600.0

0.2

0.4

0.6

0.8

1.0

Trans'n Luminesc.

ReflectionLuminescence

Transmission spectrum

hv, eV

77 K

Aug 14, 2009 NGC/CSTC, Hamilton, ON 20

Transport of Holes/Photons

In a direct-gap n-type semiconductor, an event of interband absorption does not finish off the photon:

It emerges as a new hole and then again a new photon (spectrum)

Holes and photons inside are “interchangeable” entitiesAs a photon, its lifetime is limited by free-carrier absorption (rate ~ n). Key parameter — residual absorption coefficient μr

As a hole, its lifetime is limited by non-radiative processes, (rate ~ n2). Key parameter — radiative efficiency η

Photon “recycling”a.k.a. photon-assisted diffusion of holes

Aug 14, 2009 NGC/CSTC, Hamilton, ON 21

Photon recycling

( )[ ])(

)(1)( det

0det zP

zpPpzDn

n

ηξηηη+

=−×= ∑=z

d

D(z)

where

η

≈

0.9 is the radiative efficiency and ξ=1−η

h S hη η

ξ=1−η ξP(z)=pdet

+pfca

1−P 1−P

P(z)

Setc.

Aug 14, 2009 NGC/CSTC, Hamilton, ON 22

Problem with InP and possible remedy

The signal dependence on the position

D(z) presents a fundamental problem for the intended application:

How do we distinguish large faraway event from a smaller event closer to the detector surface?

Can we correct for the z dependence?

☺ Double-sided epi detector

zd

z

d − zd

if we know

z, we can correct for the attenuation J. H. Abeles & S. Luryi

US Pat. Prov. Appl. (May 2009)

Aug 14, 2009 NGC/CSTC, Hamilton, ON 23

Photon recycling with 2-sided detector

)()()(

)()()(

22

11

zPzpzD

zPzpzD

ηξηηξ

η

+=

+= z

d − zd

D1

(z)

D2

(z)

h S h Sη η

ξ=1−η ξP(z)

1−P 1−P

P(z)=p1

+p2

+pfca

etc.

Aug 14, 2009 NGC/CSTC, Hamilton, ON 24

Calculated attenuation ratio

0 20 40 60 80 100 120 140 160 1800.0

0.2

0.4

0.6

0.8

1.0

Atte

nuat

ion

Rat

io, ρ

Distance to nearest surface, μm

zd

D1

D2

)()()(

1

2

zDzDz ≡ρ

)()()(

)()()(

22

11

zPzpzD

zPzpzD

ηξηηξ

η

+=

+=

)()(

)()(

2

1

2

1

zpzp

zDzD

=

ratio of single-pass probabilities

Aug 14, 2009 NGC/CSTC, Hamilton, ON 25

Lightly doped InP: bright luminescence

Reflection geometry Transmission geometry

Aug 14, 2009 NGC/CSTC, Hamilton, ON 26

Lightly doped InP: low free-carrier abs’n

For ND

= 2×1017

cm-3, the measured αFCA

< 0.1 cm-1

0.95 1.00 1.05 1.10 1.15 1.20 1.25 1.3002468

101214161820

cm-3 (Doping level)A

bsor

ptio

n co

effc

ient

, cm

-1

hν, eV

2E+17 2E+18 3.7E+18 4.3E+18 6.3E+18 8E+18

Aug 14, 2009 NGC/CSTC, Hamilton, ON 27

Ideal non-transparent scintillator

)()()(

)()()(

22

11

zPzpzD

zPzpzD

ηξηηξ

η

+=

+=

[ ]

[ ][ ]FCApzpzp

zpzpzP

zpzpzDzD

+++−+

=

++

=+

)()()1()()( PCE

)()()()()(

21

21

2121

ηηηηξ

η

zd

D1

D2

Based on photon recycling in low-doped (2×1017

cm-3) InP, where:

η

≈

98-99% and

FCA negligible

for η

≈

98%, PCE =91%

for η

≈

99%, PCE =95%

Unusual situation:

semiconductor is opaque, in the conventional sense, but photon collection efficiency approaches unity. No losses!

Aug 14, 2009 NGC/CSTC, Hamilton, ON 28

Pause for reflection

Two-sided detection pinpoints the z-coordinate precisely because of distance-dependent attenuation of single-pass photons

Would not work in a transparent

scintillator (except for the factor of 2)

Suppose we find semiconductor with much higher PCE?Is there advantage in 2-sided detection (apart from the factor of 2)?

Precision of z determination

is no longer ±

d, hence can make thick voxels!

Search for better PCEMust be accompanied by due regard to leg 2 of the triad (integrated PD)

Going beyond InP needs ideas for optically tight integration

Aug 14, 2009 NGC/CSTC, Hamilton, ON 29

Layered superlattice, e.g. InP/InGaAsPTransparency enhancement by the duty cycle factor, (d/a) ≤

100

Difficult to make

Need rapid growth (100 μm per hour)

HVPE ?

Hard to maintain lattice coherency in a thick free-standing structure: even a small mismatch...

Still worth trying !Any volunteers ?

holes migrate to the nearest well

diffusion time must much shorter than lifetime

d2/2D <<

τ

limits the duty cycle to about 100

Two-phase semiconductor structures

Emitted light

EG1 EG2

Host Guest

holes still migrate to the nearest inclusion

capture time must much shorter than lifetime

the duty cycle is practically unlimited

Random non-planar inclusions: “impregnations”

• Transparency gain scales as ratio of volumes

(d/a)3

• Lattice-matching requirement is removed

x'

d

x

a

Aug 14, 2009 NGC/CSTC, Hamilton, ON 30

Making a bulk impregnated structure ...

... any other ideas ?

• Direct growth with interruptions and self-organizatione.g. InGaAs quantum dots on GaAs substrate

• Phase separation via spinodal decomposition e.g. VPE grown InGaN shows subband emission (Shur)

• Transparent ceramics techniques (LLNL)Consolidation of nano-dot powders (e.g. ZnSe) by HIP

Two-phase ceramic, e.g. InP guest in CdTe matrix

Aug 14, 2009 NGC/CSTC, Hamilton, ON 31

Two-phase transparent ceramics

Mater. EG

(eV) Tm

(ºC)ZnS 3.68 1,850

ZnSe 2.822 1,100

ZnTe 2.394 1,240

CdS 2.50 1,750

CdSe 1.714 1,350

CdTe 1.474 1,041

InP 1.344 1,060

GaAs 1.424 1,240

Dots MatrixCdSe ZnTe

CdSe ZnSe

ZnTe ZnSe

CdS ZnSe

InP CdTe

impregnations must have smaller bandgap but higher melting point

Vacuum Sintering Hot Isostatic Pressing (HIP)

T ≈

0.9 Tm

Nano‐dots

matrix

Core/Shell

Nano‐mixture

InP impregnations

CdTe

Aug 14, 2009 NGC/CSTC, Hamilton, ON 32

Summary

Semiconductor scintillators – enabling technologyCompact and portable Compton telescope3D array of units, each endowed with integrated photoreceiverOptically tight integration, two-sided photodiode

Semiconductors “transparent” to own luminescenceDirect-gap semiconductors with photon recyclingSemiconductors activated with luminescent centersComposite “impregnated” systems

Aug 14, 2009 NGC/CSTC, Hamilton, ON 33

Sponsors

Energy and Spatial Correlation Effects in the Energy Resolution of Semiconductor Radiation Detectors (DTRA)

Semiconductor high-energy radiation detector with excellent isotope identification and directional capability (DHS)

Three-Dimensionally Pixellated Semiconductor Scintillator (DHS/NSF)

Aug 14, 2009 NGC/CSTC, Hamilton, ON 34

Going beyond InP... e.g., GaAs

GaAs pros• Higher radiative recombination coefficient, while nonradiative

similar

• Higher bandgap (good for low-noise photodetection at room temperature)

• May be cleaner, mature technology

GaAs cons• No proven epitaxial

detector

Wafer fusion technology may come to rescue

InPGaAs 7BB ≈ InPGaAs CC ≈

GaAs may be better than InP, but

Aug 14, 2009 NGC/CSTC, Hamilton, ON 35

no intermediate layer (wafer fusion)

Sacrificial (InP) substrate

Wafer bonding versus wafer fusion

Scintillator (GaAs) wafer

Photodiode (InP) heterostructure etchstop

intermediate bonding layer

Aug 14, 2009 NGC/CSTC, Hamilton, ON 36

What’s wrong with an intermediate layer ?

NOTHING

if it is really thin

ϕ

n3

n2d

BUT

n1GaAs

SiO2

InP

0 50 100 150 200 250 3000.0

0.2

0.4

0.6

0.8

1.0

Oxide Layer Thickness, nm

Mea

n Tr

ansm

issi

on C

oeffi

cien

t

0 2 4 6 8 10

0

2

4

6

8

10

“Frustrated”

Total Internal Reflection

λ

≈ 900 nm

naive comparison of the optical thickness with wavelength does not work for lower-index layers

even for 60 nm thick oxide layer, over 70% of GaAs scintillation will not go through

Aug 14, 2009 NGC/CSTC, Hamilton, ON 37

no intermediate layer (wafer fusion @ 600°C)

Sacrificial (InP) substrate Substrate removed (etched away using etchstop)

Photodiode integration by wafer fusion

Scintillator (GaAs) wafer

Photodiode (InP) heterostructure etchstop

Lithographic processing begins then the etchstop layer is removed too

Aug 14, 2009 NGC/CSTC, Hamilton, ON 38

InP

4. Fabrication of pin diode1. HIP on prepared InP substrate5. Flipping the picture upside down...

Optically tight photodiode integration

InGaAs etchstop layer

CdTe/InP transparent ceramic

InP sliver pin

diode

2. Etching away the InP substrate3. Etchstop layer goes away too

Aug 14, 2009 NGC/CSTC, Hamilton, ON 39

Optically tight photodiode integration

6. ... and looking at it again

p+

(Zn diffusion)n–

n+

i quaternary InGaAsP, 2 μm

Si3

N4

n+

InP slivern+

p metal contact

CdTe/InP transparent ceramic

n metal contact

Aug 14, 2009 NGC/CSTC, Hamilton, ON 40

Epitaxial pin diode fabrication

p+ n–

n+

i

n+

InP wafer

2 μmquaternary InGaAsP

Si3

N4

patterned resist

window etchingp metal contactZn diffusion

n metal contact backside patterning

epitaxial layers

bare scintillator slab

1 mm

Sarnoff – SBU collaboration

Aug 14, 2009 NGC/CSTC, Hamilton, ON 41

Compton Scattering

θ3

θ1

L3

L1

L2

θ2

E0

0.0

0.5

1.0

1.5

2.0

0

30

60

90

120

150

180

210

240

270

300

330

0.0

0.5

1.0

1.5

2.0

σ(θ)

θ γ ′

γ

e

γγ

θ′

−+=E

cmE

cm ee22

1cos

kinematics (Compton):

⎟⎟⎠

⎞⎜⎜⎝

⎛−+⎟

⎟⎠

⎞⎜⎜⎝

⎛=

′

′′ )(sin)( 2

2

0 θσθσγ

γ

γ

γ

γ

γ

EE

EE

EE

dynamics (Klein-Nishina):

γγ ′−= EEL)(Cs KeV 662

KeV 511137

2

=

=

γE

cme

Aug 14, 2009 NGC/CSTC, Hamilton, ON 42

Direction to source

( )n

n

N

j

j

nn N

npN

n δρρρr

vrr+−=≡ ∑

=

111

)(11

⎟⎟⎠

⎞⎜⎜⎝

⎛−+⎟⎟

⎠

⎞⎜⎜⎝

⎛= −

−−

)(sin)( 21

1

2

10 i

i

i

i

i

i

ii E

EEE

EE θσθσ

anisotropic scattering cross-section

dynamics (Klein-Nishina formula):

662 keV:

1ρr

Center of mass

cluster of n

interactions

0.0

0.5

1.0

1.5

2.0

0

30

60

90

120

150

180

210

240

270

300

330

0.0

0.5

1.0

1.5

2.0

σ(θ)