Wei Shi - Université Lavalwcours.gel.ulaval.ca/2016/a/GEL4203/default/5notes/PD.pdf · Junction pn: photodiodes, APD, phototransistor 3. Dr. Wei Shi GEL-4203 OPTOÉLECTRONIQUE Light

Dr. Wei Shi GEL-4203 OPTOÉLECTRONIQUE

Photodétecteurs

Département de génie électrique et de génie informatique Laval Université

Octobre 2016

Wei Shi

!1


Contenu

2

Optical absorption and PD materials Détection de lumière

Photoconductive detectors pn junction photodetector

Photovoltaic & Photodiode operations Efficacité quantique et responsivité Structures des détecteurs

PIN. Piles solaires. Avalanche. Phototransistors. Matrices CCD et CMOS.

Bruit


Qu’est un photodétecteur? Convertit la lumière en signal électrique: courant ou tension

La réponse est proportionnelle à la puissance optique incidente

Nous nous concentrons sur le photodétecteur de semiconducteur.

Photoconductive détecteur (photorésistance)

Junction pn: photodiodes, APD, phototransistor

3


Light absorption Photon energy

Must be larger than band gap

Carrier generation is a quantum process

1 photon → 1 electron-hole pair (EHP)

4


Transitions radiatives

Les trois processus se produisent simultanément à l'intérieur d'un laser.

Les taux de ces événements peuvent être quantifiés en utilisant des arguments de semi-classique (Einstein 1917).

Rsp et Rst dépendent de l'intensité optique Iv

E1

E2

spontaneousemission

stimulatedabsorption

stimulatedemission

contributes to noiseinside a laser

amplification mechanism

loss mechanism

24

h Eg


Direct bandgap vs. indirect bangap Direct bandgap

III-V materials (InP, InGaAsP, etc.)

Indirect bandgap

Phonon assisted (phonon: quantization of lattice vibration in a crystal): momentum conservation

Si, Ge

5


Popular materials Si: 1.107 eV @ 300 K; 2.3 × 10-4 eV/K

1.1 µm; +0.23 nm/K

Ge: 0.67 eV @ 300 K; 3.7 × 10-4 eV/K

1.85 µm; +1.0 nm/K

GaAs: 1.424 eV @ 300 K; 4.5 × 10-4 eV/K

0.87 µm; +0.27 nm/K

InSb: 0.225 eV @ 77 K; 1.4 × 10-4 eV/K

5.5 µm; +3.4 nm/K

6

[µm] =1.24

E[eV ]


Some photodetector materials

Band gap energy Eg at 300 K, cut-off wavelength λg and type of bandgap (D = Direct and I = Indirect) for someBand gap energy Eg at 300 K, cut-off wavelength λg and type of bandgap (D = Direct and I = Indirect) for some photodetector materials.

7

Semiconductor Eg (eV) λg (eV) TypeInP 1.35 0.91 DGaAs0.88Sb0.12 1.15 1.08 DSi 1.12 1.11 I In0.7Ga0.3As0.64P0.36 0.89 1.4 DIn0.53Ga0.47As 0.75 1.65 DGe 0.66 1.87 IInAs 0.35 3.5 DInSb 0.18 7 D

Kassap Table 5.1


Absorption Coefficient α

δ = 1/α : penetration or absorption depth

Optical loss: e.g., surface reflection R; scattering in optical waveguides

8

I(z) = Ie–↵z


Photoconductive detectors (Photoresistors) “Sole” semiconductor block

Conductance/resistance change induced by light absorption

Collection – detection: electric field – current (gain)

Ohmic contact is desired for optical gain (more than one electrons per photon for the photocurrent external)

9

PbS photoconductive detector for IR up to 2.9 µm.


Steady state illumination hcdd

hvAdA i

ii λη

ηη II

g =⎟⎠⎞

⎜⎝⎛

=Φ

= phph

Photon flux = Φph

0ph =Δ

−=Δ

τn

dtnd g

ηi = Internal quantum efficiency

Δσ = eµeΔn + eµhΔp = eΔn(µe + µh) Photoconductivity

hcde hei )( µµλτη

σ+

=ΔI Eσσ Δ=Δ=

ℓVJ ph

Photogeneration rate Photoconductivity Δσ and Photocurrent Density Jph

10

)n = gph

=il

hcd


Photoconductive Gain

11

Photon flux = Φph

hcw

ewdJ

eI hei EI )(flow electron of Rate phph µµλτη +

===

hcwwd i λη Igg ℓℓ === phph )()Volume(generation electron of Rate

Photoconductive gain G

ℓE)(

absorptionlight by generationelectron of Ratecircuit externalin flowelectron of Rate heG µµτ +

==


Photoconductive Gain

12

Photon flux = Φph ℓ

E)(absorptionlight by generationelectron of Ratecircuit externalin flowelectron of Rate heG µµτ +

==

Electron and hole transit times (time to cross the semiconductor) are

te = ℓ / (µeE)

th = ℓ / (µhE)

⎟⎟⎠

⎞⎜⎜⎝

⎛+=+=

e

h

ehe tttG

µµτττ 1

Electron

Hole

Photoconductive gain G


Contributions of electrons and holes The gain comes from the increase in conductivity, NOT from the

requirement of neutrality.

Electrons contribute more to the gain due to the faster transit time (higher velocity).

Shockley-Ramo theorem (reading: Section 5.2 Kassap)

13

i(t) =qvdt(t)

lInstantaneous current

Drift velocity


Example An n-type Si photoconductor has a length L = 100 µm and a cross

sectional area A = 10-4 mm2. The applied bias voltage to the photoconductor is 10 V. The recombination time is roughly 1 µs.

Electron and hole drift mobilities

µe = 1450 cm2 V-1 s-1; µh = 490 cm2 V-1 s-1

Q: What are the transit times, te and th, of an electron and a hole across L? What is the photoconductive gain? Which process limits the speed of on-off switching of the device?

14


Solution E = V/L= 10 V/(100×10-6 m)= 105 V/m.

Electron and hole transit times

15

Solutions Manual (Preliminary) Chapter 5 5.31 11 December 2012

(a) What are the transit times, te and th, of an electron and a hole across L? What is the photoconductive gain?

(b) It should be apparent that as electrons are much faster than holes, a photogenerated electron leaves the photoconductor very quickly. This leaves behind a drifting hole and therefore a positive charge in the semiconductor. Secondary (i.e. additional electrons) then flow into the photoconductor to maintain neutrality in the sample and the current continues to flow. These events will continue until the hole has disappeared by recombination, which takes on average a time τ. Thus, more charges flow through the contact per unit time than charges actually photogenerated per unit time. What will happen if the contacts are not ohmic, i.e. they are not injecting ?

(c) What can you say about the product ∆σ and the speed of response which is proportional to 1/τ.

Solution

(a) We are given, length L = 100 µm, and applied bias, V = 10 V. The electric field,

E = V/L= 10 V/(100×10-6 m)= 105 V/m.

From Table 3.1, electron and hole drift nobilities are: µe = 1450 cm2 V-1 s-1, and µh = 490 cm2 V-1 s-1 (see inside front cover of textbook).

Electron and hole transit times are

( ))V/m10)(sVm101450(

m1010051-124

6

−−

−

××

==E

Lte

e µ= 6.9 ns.

and ( ))V/m10)(sVm10490(

m1010051-124

6

−−

−

××

==E

Lth

h µ= 22.4 ns.

hole lifetime τ = 1 µs The photoconductive gain,

( ) ( )( )( )m10100

)V/m10(sVm10490sVm101450s1016

5-1124-11246

−

−−−−−

××+××

=+

=L

EG he µµτ = 194

(b) If the contacts are not ohmic, secondary electrons cannot flow into the photoconductor to maintain neutrality. So, only the photogenerated charges can flow through the external circuit; no excess charge can flow and we will not get photoconductive gain. If the contacts cannot inject carriers, then there will be no photocurrent gain, G = 1. However, there will still be a photocurrent as photogereated carriers will be drifting. The situation is similar to photogeneration inside the i-layer of a reverse biased pin detector.

(c) The expression for ∆σ is given by,

( )hcd

e he µµλτησ +=∆

I

The speed of response is proportional to 1/τ. For example, if we suddenly switch off the light, it will take, on average, τ seconds for the excess (photogenerated) carriers to disappear by recombination. Therefore, the product of ∆σ and the speed of response is,

Solutions Manual (Preliminary) Chapter 5 5.31 11 December 2012

(a) What are the transit times, te and th, of an electron and a hole across L? What is the photoconductive gain?

(b) It should be apparent that as electrons are much faster than holes, a photogenerated electron leaves the photoconductor very quickly. This leaves behind a drifting hole and therefore a positive charge in the semiconductor. Secondary (i.e. additional electrons) then flow into the photoconductor to maintain neutrality in the sample and the current continues to flow. These events will continue until the hole has disappeared by recombination, which takes on average a time τ. Thus, more charges flow through the contact per unit time than charges actually photogenerated per unit time. What will happen if the contacts are not ohmic, i.e. they are not injecting ?

(c) What can you say about the product ∆σ and the speed of response which is proportional to 1/τ.

Solution

(a) We are given, length L = 100 µm, and applied bias, V = 10 V. The electric field,

E = V/L= 10 V/(100×10-6 m)= 105 V/m.

From Table 3.1, electron and hole drift nobilities are: µe = 1450 cm2 V-1 s-1, and µh = 490 cm2 V-1 s-1 (see inside front cover of textbook).

Electron and hole transit times are

( ))V/m10)(sVm101450(

m1010051-124

6

−−

−

××

==E

Lte

e µ= 6.9 ns.

and ( ))V/m10)(sVm10490(

m1010051-124

6

−−

−

××

==E

Lth

h µ= 22.4 ns.

hole lifetime τ = 1 µs The photoconductive gain,

( ) ( )( )( )m10100

)V/m10(sVm10490sVm101450s1016

5-1124-11246

−

−−−−−

××+××

=+

=L

EG he µµτ = 194

(b) If the contacts are not ohmic, secondary electrons cannot flow into the photoconductor to maintain neutrality. So, only the photogenerated charges can flow through the external circuit; no excess charge can flow and we will not get photoconductive gain. If the contacts cannot inject carriers, then there will be no photocurrent gain, G = 1. However, there will still be a photocurrent as photogereated carriers will be drifting. The situation is similar to photogeneration inside the i-layer of a reverse biased pin detector.

(c) The expression for ∆σ is given by,

( )hcd

e he µµλτησ +=∆

I

The speed of response is proportional to 1/τ. For example, if we suddenly switch off the light, it will take, on average, τ seconds for the excess (photogenerated) carriers to disappear by recombination. Therefore, the product of ∆σ and the speed of response is,


Principe: photodiode Structure

Jonction pn

Exemple: p+n jonction

Reverse biased or short-circuit

AR coating (multi-layer dielectric thin films)

Loaded line for I-V conversion

Génération des EHPs

Transition des electrons à cause de l’absorption des photons incidents

Transport des porteurs

Dérive dans la zone de déplétion

Diffusion dehors

16


Détection de lumière – photodiode Deux types en termes d'incidence de la lumière

Surface éclairée et guide d’onde

Surface éclairée

Courant en parallèle avec le flux des photons;

Grande zone de détection;

Pratique pour la détection de l'espace libre.

17


Photodiode de guide d’onde Courant perpendiculaire au flux des photons

Responsivité élevé: un long guide d’onde

Défi pour le couplage entre le guide d'onde d'entrée et le matériau d'absorption

Convient aux circuits intégrés

18

Feng, Dazeng, et al. Applied Physics Letters 95.26 (2009): 261105.


Absorption dans SCL: courant de dérive

19

• Une jonction p-n polarisée inversée.

• La photogénération à l'intérieur du SCL génère un électron et un trou (SHP).

• Les deux tombent des leurs collines énergétiques respectives (électron le long de Ec et trou le long de Ev) i.e. c'est-à-dire qu'elles dérivent et provoquent un photocourant Iph dans le circuit externe.

• Courant de dérive: très rapide


Absorption dans la région neutre: courant de diffusion

20

• La photogénération se produit dans la région neutre.

• L'électron doit diffuser vers la SCL puis rouler vers le bas de la colline d'énergie, c'est-à-dire dériver à travers la SCL.

• Considérer: Que se passe-t-il si un EHP est généré dans une région neutre (p ou n) trop éloignée de SCL (>> la longueur de diffusion)?

• Ne peut pas atteindre SCL avant de se recombiner avec un porteur majoritaire;

• Pas de contribution au photocourant


Caractéristiques I-V avec éclairage Convention: direction positive de p vers n

V: La tension de la région p par rapport à la région n (la flèche indique une augmentation par défaut)

21

I = I0e( eVkBT 1) Ip IP: Photocurrent

I0: Saturation current


Modes de fonctionnement Deux modes pour une photodiode: photodiode (photoconducteur) &

photovoltaïque

Déterminé par la condition de polarisation et les circuits externes.

22


Mode Photodiode (photoconducteur) Dans le troisième quadrant des caractéristiques I-V, y compris la

condition de court-circuit sur l'axe vertical pour V = 0..

Fonction comme une source de courant

23


Mode Photodiode (Photoconducteur) Court-circuit: dérive due au champ “build-in” (build-in voltage); Bande-

diagramme est comme “zero bias”; EFp = EFn

Polarisée en inverse: les bandes sont inclinée; un champ électrique plus fort -> dérive plus rapide; EFp > EFn, EFp - EFn = eVr

24


Mode photovoltaïque Dans le quatrième quadrant, y compris la condition de circuit ouvert

sur l'axe horizontal pour I = 0.

Fonction comme une source de tension.

25


Mode photovoltaïque: tension en circuit ouvert Nous regardons le diagramme de bande pour comprendre la tension

en circuit ouvert

26

• La jonction pn en circuit ouvert.

• L'électron photogénéré et le trou roulent en bas de leurs collines d'énergie (processus dérive).

• Il y a une tension Voc (forward) à travers la diode qui les fait diffuser vers l'arrière de sorte que le courant net soit nul.

• EFp < EFn, EFp - EFn = -eVoc


Efficacité quantique Efficacité quantique interne (IQE): génération des EHPs

Efficacité quantique externe (EQE, ou QE): Collection des EHPs

27

ηe =I ph / ePo / hυ

R =Photocurrent (A)

Incident Optical Power (W)=I phPo

Responsivité

i =NEHP,gen

Np

i =NEHP,col

Np

R =ηeehυ

=ηeeλhc

e[µm]

1.24


Spectre de responsivité Dépendant de la longueur

d'onde

La forme du spectre de responsabilité est déterminé par

La longue d’onde

QE

Matériel: le coefficient d’absorption

Structure: le transport et la collection des porteurs

Le photocourant généré dans un certain spectre.

28

X: maximum QE Iph =

Z max

min

R()P0()d


Le transport des porteurs La dépendance de longueur

d'onde liée à la structure.

Profils de photogénération correspondant à des longueurs d'onde courtes, moyennes et longues.

Différentes contributions au photocourant Iph.

Les EHPs ont des processus de transport différents (dérive et diffusion)

La vitesse dépend également de la longueur d’onde

Dérive est rapide et diffusion est lente

29


PIN photodiode Une zone non-dopée, dite intrinsèque (i),

intercalée entre deux zones dopées P et N

Trous et électrons diffusent des zones P et N, respectivement, aux la zone intrinsèque;

Se recombinant dans la zone intrinsèque;

Des charges aux frontières p-i et i-n;

Champ électrique intégré (build-in) qui empêche la diffusion pour l’equilibrium.

EHPs générés par l’absorption optique

Polarisation inverse pour balayer les EHPs hors de la zone intrinsèque pour Iph.

p-i-n vs. p-n

Un plus grand espace pour un transport rapide (dérive)

Capacitance inférieure –> Response plus rapid

30

Build-in field


Photocourant dans PIN Photodiode Considérons une PD éclairée en surface.

En supposant que lp est très mince et en supposant W >> Lh : tout le courant est généré dans la région intrinsèque

31

I ph ≈eηiTPo(0)

hυ[1− exp(−αW )]

T = Transmission coefficient of AR coating

α = Absorption coefficient

(Reading: Example 5.5.4, Kasap)


Vitesse d'une photodiode High-speed photodiodes are the most widely used photodetectors in

applications requiring high-speed or broadband photodetection.

The speed of a photodiode is determined by two factors:

The response time of the photocurrent The RC time constant of its equivalent circuit

Most photodiodes use the photoconductive (photodiode) mode for high-speed applications.

32

Because a photodiode operating in photovoltaic mode has a large RC time constant due to the large internal diffusion capacitance upon internal forward bias in this mode of operation.

Q: Why NOT photovoltaic?

fc =1

2RC


Limite de vitesse intrinsèque La vitesse est limitée intrinsèquement à la vitesse à laquelle l'EHP peut

être balayé (vitesse de dérive), tandis que le circuit peut être optimisé séparément.

33

E =Eo +VrW

≈VrW

Cdep =εoεrAW

Vitesse-efficacité trade-off

Augmenter W –> vitesse élevé (tdrift réduit), mais résponsivité (R) réduit

tdrift =Wvd Drift velocity


Exemple: 100 GHz PD

34

Ultra-Fast 100 GHz Photodetector – April 2014

Confidential and Proprietary © 2014 Finisar Corporation. All rights reserved Rev. A1 Page 3

IV. Environmental Conditions Parameter Symbol Condition Min. Typ. Max. UnitOperating Case Temperature TCase 0 75 °C Relative Humidity RH non condensing 5 85 % Storage Temperature Tsto -40 85 °C

V. Operating Conditions

Parameter Symbol Condition Min. Typ. Max. UnitOperating Wavelength Range λ 1480 1620 nm Average Optical Input Power Range POPT -20 10 dBmPhotodiode Bias Voltage VPD 1.5 2.0 2.8 V

VI. Electro-Optical Specifications1

Parameter Symbol Condition Min. Typ. Max. UnitPhotodiode DC Responsivity R optimum polarization 0.5 A/W

Polarization Dependent Loss PDL 0.5 dB Optical Return Loss ORL 27 dB

3dB Cut-off Frequency2 f3dB XPDV4121R 100 110 GHz XPDV4120R 90 95 GHz

Output Reflection Coefficient S22 0.05 - 50 GHZ -10

-8 dB 50 - 100 GHZ -5

Overload POVERL 10 dBmPhotodiode Dark Current I dark 5 200 nA

Pulse Width3 7.5 8 Ps Notes:

1. VPD = 2.8 V, Tcase = 25 °C, 1550 nm 2. measured using a heterodyne measurement system 3. measured utilizing Tektronix Scope with 70 GHz sampling head


Product Specification Ultra-Fast 100 GHz Photodetector

XPDV412xR

PRODUCT FEATURES

x 100 GHz electrical 3 dB bandwidth x Flat response of up to 100 GHz

x Excellent pulse behavior

x Well matched 50 Ω output

APPLICATIONS

x High-speed lightwave characterization

x 100 Gb/s communication systems

x Microwave photonics

The XPDV412xR comprises an optimized 100 GHz waveguide-integrated photodiode, which shows an extremely flat frequency response in both, power and phase. The on-chip integrated bias network with an optimized RF design in particular, ensures an undisturbed frequency response from DC to the 3 dB cut-off frequency and saves costs for internal bias-tees. The module is especially designed for optimal RF performance; therefore the pulse response reveals virtually no ringing. A further advantage of the waveguide structure is the unbeatable high-power behavior. The photodetector shows a linear response up to an optical input power of 10dBm. An output voltage swing of more than 0.5 Vpp can be achieved for short pulses without any degradation of the pulse response. Each photodetector module is characterized in the frequency domain by using a heterodyne technique. In the time domain, a femto-second pulse source and a 70 GHz sampling oscilloscope are used to measure the pulse response.

ORDERING INFORMATION

x: 1 = minimum 100 GHz 0 = minimum 90 GHz

zz: FP = FC/PC (standard) Customized connectorization available upon request

XPDV412xR-WF-zz


Product Specification Ultra-Fast 100 GHz Photodetector

XPDV412xR

PRODUCT FEATURES

x 100 GHz electrical 3 dB bandwidth x Flat response of up to 100 GHz

x Excellent pulse behavior

x Well matched 50 Ω output

APPLICATIONS

x High-speed lightwave characterization

x 100 Gb/s communication systems

x Microwave photonics

The XPDV412xR comprises an optimized 100 GHz waveguide-integrated photodiode, which shows an extremely flat frequency response in both, power and phase. The on-chip integrated bias network with an optimized RF design in particular, ensures an undisturbed frequency response from DC to the 3 dB cut-off frequency and saves costs for internal bias-tees. The module is especially designed for optimal RF performance; therefore the pulse response reveals virtually no ringing. A further advantage of the waveguide structure is the unbeatable high-power behavior. The photodetector shows a linear response up to an optical input power of 10dBm. An output voltage swing of more than 0.5 Vpp can be achieved for short pulses without any degradation of the pulse response. Each photodetector module is characterized in the frequency domain by using a heterodyne technique. In the time domain, a femto-second pulse source and a 70 GHz sampling oscilloscope are used to measure the pulse response.

ORDERING INFORMATION

x: 1 = minimum 100 GHz 0 = minimum 90 GHz

zz: FP = FC/PC (standard) Customized connectorization available upon request

XPDV412xR-WF-zz

R = e[µm]

1.24Idéalement R>1 @ 1550 nm


Vitesse de dérive Linéaire lorsque le champ électrique est faible

Saturates à champ électrique élevé

Le temps de dérive à travers la région intrinsèque (W) suit la même tendance

35

tdrift =Wvd

Vitesse de dérive par rapport au champ électrique pour les trous et les électrons dans Si.

Q: 2V across 1 µm intrinsic region in silicon. What’s the transit time for hole and electron? Which one dominates?

A: E = 106 V/m Vd,e = 105 m/s; Vd,h = 4 x 104 m/s

te = 10 ps; th = 25 ps

th dominates the speed as it’s a slower process.


PIN Photodiode: diffusion A short wavelength light pulse is absorbed very near the surface.

The photogenerated electron has to diffuse to the depletion region

Then it is swept into the i-layer and drifted across.

Diffusion velocity is orders of magnitude lower than the drift velocity.

De : diffusion coefficient

The doped region should be minimized for high-speed operation.

36

Q: Assume l = W = 1 µm, Vr = 2V, calculate the diffusion time, drift time, and the total transit time for an electron generated in p+ region. Use De = 3 x 10-4 m2/s.

A: tdiff = l2/(2De) = 1.67 ns tdrift = W/ve = 10 ps = 0.01 ns ttotal = tdiff+tdrift = 1.678 ns

tdiff = l2/(2De)


Waveguide Ge photodiode On the SOI platform

1.58 A/W; BW > 20 GHz

37

1

SiEPIC Active

Photodiode

Jonathan St-Yves

I. Introduction

Photodiodes are a critical component of opticalcommunication networks. The photodiode currentlyprovided in the SiEPIC Library isn’t up to par withthe current state of the art, as described in an articleby Yi Zhang et al. [4].

The main figures of merit of a photodiode are itsresponsivity (in A/W) and its 3dB bandwidth (inGHz). The results of Zhang et al. are reportedly of1.44 A/W and 36 GHz. Meanwhile, the photodiodepresent in the library doesn’t have any documentationand is reported to have a much lower bandwidth,about 20 GHz. Other important figures are therequired reverse bias and the dark current.

II. Goal

The goal of this project is to:• Reproduce the results described by Zhang and al.

in [4].• Make it available for the SiEPIC program.• Scrutinize the various design decisions of the

previous work and optimize them further.

III. Device description

The photodiode consists of a lateral silicon P-Njunction with a germanium layer on top, as seen infigure 1. While the P and N conductor aren’t aroundthe germanium, the charges created by the absorptionof light are still recovered by the fringe field presentover the P-N junction. The light is mainly guided bythe intrinsic silicium and the germanium, reducing thelosses caused by dopant absorption.

The electric field in the germanium is weaker inthis configuration than with a vertical P-N junctionwhere the field directly traverses it. However, it isstill possible to achieve maximal drift velocity witha reasonable voltage. The advantage of the presentconfiguration is that the germanium lattice isn’t dis-turbed by dopants or metal contacts. This highquality lattice is what explains the good performance.[4] This is a fundamental dierence from photodiodedesigns where the P-N junction is built using doped

Figure 1. Cross-section of the photodiode.

germanium with metals in contact with the lattice, ineither a vertical[1] or horizontal [3] arrangement.

Figure 2. Schematic of the photodiode seen from the top.

Figure 2 shows the device from the top, with thelight input being on the bottom. Labels indicate thevariable length of multiple features. A longer totallength of the device (L1) means a higher absorptionlength and thus higher responsivity. The length of thesilicon and germanium taper (St and Gt) aect themode-mismatch reflection and thus, the responsivity.

A new technique attempted with this design is to ta-

3

The electrical bandwidth is the inverse of the prod-uct of the capacitance and resistance. Capacitanceis proportional to the length of the device whileresistance is inversely proportional to it, meaning theycancel out and length is not a factor. Capacitancecan be lowered a bit by widening the P-N junction.However, it is very straightforward to reduce the re-sistance of the junction by putting the dopants closerto the waveguide. The Zhang article points out thatby doping the slab with high density (++) dopants,the resistance will be low enough that the bandwidthwon’t be limited by the RC constant. We propose togo a little further and taper the dopants to reduce theresistance while aecting the optical absorption evenless.

The carrier transit time aects the bandwidth be-cause the charges created near the electrode are notcollected simultaneously with the ones further away.Dierence of time between the arrival of these dierentcharges causes a transient response that can limitthe bandwidth if the next pulse arrives before thecarriers are swept out. This eect is limited by thediameter of the mode in the absorption region and thespeed at which carriers move across it. A strongerelectric field sweeps out the charges faster, up to asaturation velocity of 6.5 ú 10≠4 m/s in germanium,which cannot be exceeded. [2] Thus, the only ways toimprove this aspect is to either change to a materialwith an higher saturation velocity, make sure thevoltage is high enough or to make the mode smaller.Germanium is the only material readily available andvoltage can be adjusted as needed in the experiment.To increase the bandwidth, devices with a smallersection of germanium are fabricated, reducing themode size.

Figure 3. Fundamental mode of the germanium absorptionzone. The size of the mode governs the carrier transit timelimited bandwidth.

V. Layout and Variations

The full layout is included in appendix.In order to be able to test the device, the electrical

connections must be far enough from the optical inputto leave space for the fiber array (about 500 um). Thedevices are interleaved to use up that space eciently.

All devices were fabricated with a default plusPlus-Penetration of 0.7 micron and length of 17.

The following variations were fabricated:• All permutations of GeWidth = 1,1.2,1.5,2 with

length = 12,17,26,41.• taperGe = 2,3,4,5.• pnPenetrationMidPercent = 15,20,30,35.• All permutations of GeWidth = 1.2,1.5,1.7 with

plusPlusPenetration = 0,0.5,1,1.3.• All permutations of taperDopants = 1,5 with

plusPlusPenetration = 0,0.5,1,1.3.• All permutations of GeWidth = 1,1.2,1.5 with

length = 12,17. For around at 1310 nm with theappropriate routing width and grating couplers.

VI. Fabrication

In order to reduce the resistance of the electricalcircuit without adding much optical loss, it is prefer-able to use an advanced doping profile with 3 levels ofdoping for both P and N. The junction will be madewith the structure P++, P+, P, I, N, N+ and N++.

The design doesn’t include any small features,meaning lithography resolution should not be a con-cern.

The IME PDK has a design rule stating that thegermanium cannot be right on the edge of the trench,it must be included within 1 micron of full silicon.However, it appears to be a soft rule and only somedesigns respect it as a backup.

VII. Experimental results

The devices have been tested at Laval Universityusing fiber grating couplers as the optical inputs andoutputs. The current was measured using RF probesand a multimeter, with a current source applying areverse voltage.

Figures 4 and 5 show the current for various volt-ages and optical powers for a photodiode with thedefault parameters, which is close to the design of theprevious demonstration, but with dopant tapers.

The dark current and sensitivity accurately repro-duce what is announced in the previous paper. Notethat all optical power measurements have a 20%uncertainty due to variation between grating couplers.As such, it is not possible to determine whether the

4

Figure 4. Current in function of reverse bias voltage, fordierent optical powers. Photocurrents are calculated by sub-tracting dark current from the measurements.

Figure 5. Current in function of optical power, for dierentbias voltages.

results are better or worse, but both measurementsindicate a quantum eciency around 90% between 2and 5 volts, before avalanche behavior begins.

We observe that for voltages over around 5 volts,the current steadily increases due to increasingavalanche gain. The eect is even accelerating be-tween 10 and 17 volts. A maximum photo-current isgenerated around 18 volts, beyond which the currentincreases sharply with or without illumination due tojunction breakdown.

On an other chip, we measured the absorption andresponsivity of the photodetector using a smaller ger-manium cross-section with a width of 1.0 µm insteadof 1.5 µm. This should help increase the bandwidthwhich will be measured with a high speed VNA later,but is expected to reduce the eciency. Figure 6

shows the response for dierent voltages comparingthe smaller width to the usual one. We observe thatthe responsivity is similar, within the unknown vari-ance caused by fiber grating coupler. We measuredthrough port losses of 12 dB versus the usual 15 dB,indicating double the through port losses, from 3%to 6%. The main dierence is the breakdown voltage,situated at -18 V instead of -22 V. It makes sense thatsince the P-N junction is thinner, the field is strongerfor the same voltage. This is desirable since reachinghigher voltage for avalanche gain is dicult in someapplications.

Figure 6. Photocurrents in function of reverse bias voltages,for dierent widths of germanium absorber.

VIII. Conclusion

From what we measured so far, we successfullyreproduced the results published by Zhang and al.[4].The sensitivity values agree within the uncertaintyrange. Measurements are underway to evaluate theRF response, but preliminary results agree so far.In order to determine if the variations in geometryproduce better results as expected, we will have torefine our evaluation of the optical power at the inputof the photodiode. We also need to measure the devicewith an high speed network analyzer to evaluate thebandwidth accurately.

Further improvement to photodiodes would be todecrease the mode size further to improve the band-width, although only some improvement is possibleon that front and pushing the bandwidth further willrequire the use of other materials. The quantumeciency is already very high due to the low amount ofrecombination due to dislocations. Better sensitivity

1

SiEPIC Active

Photodiode

Jonathan St-Yves

I. Introduction

Photodiodes are a critical component of opticalcommunication networks. The photodiode currentlyprovided in the SiEPIC Library isn’t up to par withthe current state of the art, as described in an articleby Yi Zhang et al. [4].

The main figures of merit of a photodiode are itsresponsivity (in A/W) and its 3dB bandwidth (inGHz). The results of Zhang et al. are reportedly of1.44 A/W and 36 GHz. Meanwhile, the photodiodepresent in the library doesn’t have any documentationand is reported to have a much lower bandwidth,about 20 GHz. Other important figures are therequired reverse bias and the dark current.

II. Goal

The goal of this project is to:• Reproduce the results described by Zhang and al.

in [4].• Make it available for the SiEPIC program.• Scrutinize the various design decisions of the

previous work and optimize them further.

III. Device description

The photodiode consists of a lateral silicon P-Njunction with a germanium layer on top, as seen infigure 1. While the P and N conductor aren’t aroundthe germanium, the charges created by the absorptionof light are still recovered by the fringe field presentover the P-N junction. The light is mainly guided bythe intrinsic silicium and the germanium, reducing thelosses caused by dopant absorption.

The electric field in the germanium is weaker inthis configuration than with a vertical P-N junctionwhere the field directly traverses it. However, it isstill possible to achieve maximal drift velocity witha reasonable voltage. The advantage of the presentconfiguration is that the germanium lattice isn’t dis-turbed by dopants or metal contacts. This highquality lattice is what explains the good performance.[4] This is a fundamental dierence from photodiodedesigns where the P-N junction is built using doped

Figure 1. Cross-section of the photodiode.

germanium with metals in contact with the lattice, ineither a vertical[1] or horizontal [3] arrangement.

Figure 2. Schematic of the photodiode seen from the top.

Figure 2 shows the device from the top, with thelight input being on the bottom. Labels indicate thevariable length of multiple features. A longer totallength of the device (L1) means a higher absorptionlength and thus higher responsivity. The length of thesilicon and germanium taper (St and Gt) aect themode-mismatch reflection and thus, the responsivity.

A new technique attempted with this design is to ta-

Jonathan St-Yves


Basic photodiode circuits Consider steady-state operation at a certain incident optical power

Iph = - Isc Load RL converts current to voltage and is usually large (e.g., MΩ)

The load line represents the behavior of the load R

I = − (V + Vr) / RL= − VL / RL VL: Voltage drop on RL

38

Example: Vr = 6V; RL=1MΩ


Photodiode Equivalent Circuit (small signal) Depletion region: Cdep

Termal/packaging capacitance

Total capacitance: Ct

Series resistance (Rs): p and n regions

Shunt/parallel resistance (Rp): parasitics, e.g., surface leakage

39

Ideal photodiode


Cut-off frequency Small signal modulation (O-E): inverse process to optical modulation

(E-O)

40

V(t)

fc =1

2πReqCt

≈1

2π (Rs + RL )Ct

≈1

2πRLCt

Req is equivalent resistance and represents (Rs + RL) in parallel with Rp


A commercial photoreceiver A photoreceiver that has an InGaAs APD and peripheral electronics

(ICs) to achieve high gain and high sensitivity.

APD: avalanche photodiode (Section 5.7, Kasap)

There is also a thermoelectric cooler (TEC) and a temperature sensor (TSense). Courtesy of Voxtel Inc (www.voxtel-inc.com)

41

APD

Thermoelectric (TEC) cooler

Op amp

Output

APD bias Temperature sensor (Tsense)

TEC Current in direction

TEC Current out direction

Base/Collector

Emitterr

Op amp bias


Pulsed Excitation

42

Po(t)

t Short light pulse Large resistor to

bias the PD

Bias or shorting capacitor to short RB and the battery for the transient

photocurrent. It is a short for ac/transient signals

Reverse biases the PD

Very fast buffer or amplifier that does not load RL.

Coupling capacitor that allows ac/transient signal coupling

Load resistor for developing a voltage signal


Pulsed Excitation No limit from the drift/diffusion

43

The Experiment

Rise time Fall timeAre these related to fc?


Rise and Fall Times, and Bandwidth τR, τF: 10% – 90%

44

Exponential Decay: V(t) ~ V100% exp(−t/τ)

τ = RLCt

Very roughly, τR ~ τF = 2.2τ

fc ≈1

2πRLCt

=12πτ

=0.35τF

=350MHzτF (ns)


Pulsed excitation: carrier transport Response due to the diffusion and drift of photogenerated carriers

Assume Rs + RL is very small so that (Rs + RL)Ct is negligible

45

Drift of carriers in the depletion

region

Diffusion of carriers in the neutral region

Slow

Fast

Fast

Slow t

Drift of carriers in the depletion region

Diffusion of carriers in the neutral region

Vout


Résumé Quelles sont les pensées que l'on devrait prendre pour concevoir un photodétecteur?

Application –> Conception

Matériel

Longueur d’onde –> Bandgap (directe vs. indirecte)

Efficacité énergétique -> Coefficient d’absorption

Disponibilité dans le contexte

Exemple: Ge est compatible avec la plate-forme de Si (CMOS & la photonique sur silicium)

Conception de structure et fonctionnement

Matériel –> structure de diode et mode de fonctionnement

p-i-n diode en mode photoconducteur est populaire pour la detection optique; mais, InAs et InSb détecteurs utilisez habituellement la mode photovoltaïque afin de supprimer le “dark current.”

Performance (matériel et structure) –> conception physique et circuits

La vitesse: response intrinsèque (dérive et diffusion) et RC constante

Responsivité: conception optique (AR & WG); efficacité-vitesse trade-off

Le bruit

46

Documents

Wei Shi - Université Lavalwcours.gel.ulaval.ca/2016/a/GEL4203/default/5notes/PD.pdf · Junction pn: photodiodes, APD, phototransistor 3. Dr. Wei Shi GEL-4203 OPTOÉLECTRONIQUE Light