Professional Documents
Culture Documents
Stimulated Emission
Stimulated emission and photon amplification
Stimulated emission rate and Einstein coefficients
Light amplification
Rate equations for amplifiers
Laser Oscillation Condition
Optical threshold gain
Output characteristics
Semiconductor laser diode
Principles
Gain spectrum under forward bias
Homojunction laser diode
Heterojunction laser diode
Laser diode layout and lateral mode confinement
Laser diode characteristics
Steady state semiconductor rate equations
Optimal out-coupling
Distributed bragg reflection for single mode lasers
Gain spectrum in QW lasers
Low dimensional structures & threshold current reduction
Vertical cavity surface emitting lasers
Late News: Ge electrically pumped laser
TFH
SS 2012
92
TFH
Shown later: Light amplification requires more electrons in the excited state than in
the ground state (population inversion).
SS 2012
93
2
1
2
1
2
1
2
1
2
1
spont
stim
De-excitation rate R21 = R21
+ R21
I
I
stim
R21
analog to absorption:
= B21 N2 (h21 )
spont
R21
= A21 N2 , independent of number of photons, only
prop. to N2 .
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
Thermal equilibrium:
a) R12 = R21 .
b) Boltzmann statistics: N2 /N1 = e(E2 E1 )/kB T = eh21 /kB T .
c) Photon energy density in BB given by Plancks law: eq (h) =
TFH
SS 2012
8h 3
c3
eh/kB T
94
Interpretation:
I
I
I
TFH
3
8h21
A21
(= h Photon DOS)
=
B21
c3
Stimulated emission and absorption are inverse processes and occur with equal
probability per available initial state (2 for stim. emission, 1 for absorption)
population inversion (N2 > N1 ) can never be achieved by intense optical pumping a
two level sytem, at most transparancy (N1 = N2 ) is achieved.
Spontaneous emission prop. and stimulated emission prop. are related, ratio
increases with 3 , harder to construct x-ray lasers.
stim
R21
spon
R21
B
(h)
A
(h)
1
h Photon DOS
95
Light amplification
= I I (x) = I0 ex exponential decay of intensity I .
Absorption law:
For expressing via electron transition rates, we convert intensity I into photon
flux , photon density nPh and : I = h = nPh ch = c(h).
I dI
dx
I
I
TFH
dI
dx
dI
dx
= h d
= ch
dx
dnph
dx
= h
dnph
dt
hB
= (N1 N2 ) c
I
SS 2012
96
1
1
1
1
1
example: spontaneous lifetime of level 2: 21 = 21
+ 20
= 21,sp
+ 21,nr
+ + 20
dN3
N3
=R
dt
32
dN2
N3
N2
=
N2 Wi + N1 Wi
dt
32
2
dN1
N1
N2
=
+ N2 Wi N1 Wi
dt
21
1
dN0
N1
N2
= R +
+
dt
1
20
TFH
I 32 3x N3 = R32
I stationary solution
N2
dN2
=R
N2 Wi + N1 Wi
dt
2
d
=0:
dt
R2 (1 1 /21 )
N 0
1 + Wi [2 + 1 (1 2 /21 )]
1 + Wi s
N 0 : pop. inv. in the absence of amplifier radiation.
s : saturation time.
I N N2 N1 =
97
abs : absorption: impurities in laser medium, non-inverted allowed transitions with similar
transition energy, free carriers (semiconductors !!)
sc : scattering out of resonator mode: defects and inhomogeneities.
Ri : Out-coupling losses, power reflection coefficient of
. cavity mirrors R1 , R2 / 1
!
stationarity conditions: Pf = Pi
Pf = Pi R1 R2 e2Lgth e2L(abs +sc )
z }| { 1
gth = abs + sc + 2L
ln( R11R2 ) threshold gain.
c0
threshold pop. inv.
B21 nref h0
0
Increase R (i.e N ): g remains clamped at gth
after stationary is re-established N = Nth
Nth = gth
TFH
SS 2012
N/N 0 = g/g0
1.0
laser turn-on
time
0.8
0.6
0.4
steady state
loss=gth
ph. flux
density
0.0
102 101 100 101
Wi s = /s
0.2
102
98
1 + Wi s
Wi (h12 )
1 N 0
1
Wi =
s Nth
0
s N 1 , N 0 > N th
th
N
=
0,
N 0 N th
Nth =
steady-state laser-internal
photon-flux density
TFH
SS 2012
99
PrinciplesoftheLaserDiode
pn junctioninaDegenerate
Semiconductor
Eg
Appliedbiasdiminishesthebuildin
potentialbarrier
Depletionregionisnolongerdepleted
Therearenowmoreelectronsinthe
conductionbandthaninthevalance
bandnearEv Populationinversion
Junction
n+
Ev
EF p
Ho les in V B
Electro ns
Electro ns in C B
n+
p+
eVo
Ec
EF n
Ec
EF n
In v ersio n
reg io n
Ec
Eg
eV
EF p
ChangeintheFermilevelisthework
donebytheappliedvoltage,eV
Ifthejunctionisforwardbiased
suchthatEfn Efp =eV >Eg
TFH
p+
Ec
Whenvoltageisapplied
FermilevelinthepregionisbelowEv
FemilevelinthenregionisaboveEc
Withnoappliedvoltage,Efn=Efp yields
averynarrowdepletionregion
Thereisapotentialenergybarrier,eVo
thatpreventsnsideelectronsfrom
diffusingtothepsideandviceversa
(a)
Ev
(b)
V
The energy band diagram of a degenerately dopedp-n with no bias. (b) Band
diagram with a sufficiently large forward bias to cause population inversion and
hence stimulated emission.
1999 S.O. Kasap, Optoelectronics (Prentice Hall)
SS 2012
100
PrinciplesoftheLaserDiode
Populationinversionregionisa
layeralongthejunctioncalledthe
inversionlayer(activeregion)
Anincomingphotonwithenergy
EcEv cannotexciteanelectronin
Ev toEc astherearehardlyany
presentinthevalanceband
withintheactiveregion
Hencethereismorestimulated
emissionthanabsorption
Theopticalgainpresentinthe
activeregionduetolackof
probabilityofvalanceelectron
absorption
Energy
Optical gain
EF n EF p
CB
EF n
Ec
Electrons
in CB
eV
Eg
Holes in VB
= Empty states
Ev
EF p
At T > 0
VB
At T = 0
Optical absorption
Density of states
(a)
(b)
(a) The density of states and energy distribution of electrons and holes in
the conduction and valence bands respectively at T 0 in the SCL
under forward bias such that E Fn E Fp > E g. Holes in the VB are empty
states. (b) Gain vs. photon energy.
1999 S.O. Kasap,Optoelectronics (Prentice Hall)
TFH
SS 2012
101
SS 2012
102
From A. Yariv, Quantum Electronics, 3rd edition, John Wiley & Sons, (1989)
TFH
SS 2012
103
ElectroopticalPerformanceofIII/VDiodes
Current
injectionpumping:Opticalpumpingisachievedbyforward
diodecurrentandthepumpingenergyisanexternalbattery
Forlaserwealsoneedanopticalresonatorcavity.Thisis
achievedthroughtheuseofaslabwaveguidewithahigh
n
1
indexcontrastattheemissionend R = ( nGaAs +1 )2 35%
GaAs
Wavelengthoftheradiationthatcanbuildupinthecavity
dependsonthelength(L)inhalfmultiples
2n
Electrode
GaAs
L
n+
GaAs
Electrode
Active region
(stimulated emission region)
Optical P ower
Laser
LED
Stimulated
emission
Optical P ower
Laser
Spontaneous
emission
I
It h
Typical output optical power vs. diode current ( I) characteristics and the corresponding
output spectrum of a laser diode.
1999 S.O. Kasap, Optoelectronics (Prentice Hall)
TFH
L
p+
=L
Optical Power
Optical P ower
2criticalcurrentidentifiers
Transparency current:Currentabove
whichnonetphotonabsorption
occurs
Threshold current:currentabove
whichopticalgainovercomesall
photonlossesinthecavity
typically: Jth = 500 Amm2 for
homojunction LD !!
SS 2012
104
Heterojunction LaserDiodes
Mainissuewithhomojunction diodesis
thatthelaserthresholdcurrentdensityis
toohighforpracticaluses.
TFH
Heterostructured diodesreducethese
currentdensitiesbyordersofmagnitude
Thisisachievedthroughacombinationof
carrierconfinement(mismatched
materials),andphotonconfinement
(geometricshapeofthewaveguide)
Doubleheterojunction (DH)deviceswith
npp layersallowfordesigned
confinementoftheactiveregion
LowerrefractiveindexoftheAlGaAs
enhancesthemodeconfinementin
comparisontoahomoorsimple
heterojunction device
Significantlyreducesthresholdcurrent
density
n
AlGaAs
(a)
GaAs
AlGaAs
(a) A double
heterostructure diode has
two junctions which are
between two different
bandgap semiconductors
(GaAs and AlGaAs).
(~0.1 m)
Electrons in CB
Ec
Ec
Ec
2 eV
1.4 eV
2 eV
(b)
Ev
Ev
Holes in VB
Refractive
index
(c)
Photon
density
Active
region
n ~ 5%
(d)
SS 2012
105
Figure 15.10 A typical double heterostructure GaAs/GaAlAs laser. Electrons and holes are injected into the active GaAs layer from the
n and p GaAlAs. Photons with frequencies near =Eg/h are amplified by stimulating electronhole recombination.
Cc2007Photonics,6theditionYariv andYeh (OxfordUniversityPress
Photonics,6theditionYariv andYeh
TFH
SS 2012
106
Figure 15.12 The magnitude of the energy gap in Ga1x Alx As as a function of the molar fraction x. For x > 0.37 the bandgap is indirect.
(After Reference [11].)
Photonics,6theditionYariv
andYeh
TFH
11
SS 2012
107
Photonics,6theditionYariv
andYeh
TFH
12
SS 2012
108
Heterojunctions Laser
Heterojunction diode: different materials for n & p
Different materials: significantly different index n
Also different lattice constants
Important point: want the lattice matched at layer boundary
Use mixed alloy: eg GaAs and AlAs
AlxGa1-1As
x = mole fraction of Aluminum
1-x = mole fraction of Gallium
Heterojunctions Laser
Single Heterojunctions: one sided confinement
p-GaAlAs: p-GaAs: n-GaAs
Better confinement means lower threshold current for lasing
Thus operates in pulsed mode at room temperature
Double Heterojunction lasers: confines both top & bottom
p-GaAlAs: GaAs: n-GaAlAs: n-GaAs
EdgeEmittingLasers
VerysimilartoELEDdevicespresentedinchapter
3
AdditionalcontactinglayerofpGaAs nexttothe
p=AlGaAs providesbettercontactingandavoids
Schottky junctionwhichwouldlimitthecurrent
inthedevice.
p andn AlGaAs layersprovidecarrierand
opticalconfinementintheverticaldirection
LaseremissionintheactivepGaAs(oradifferent
AlGaAs constitution)regionisbetween870900
nmdependingondoping.
Stripe electrode
TFH
Oxide insulator
p -GaAs (Contacting layer)
p -Al x Ga 1- x As (Confining layer)
p -GaAs (Active layer)
n -Al x Ga 1- x As (Confining layer)
n -GaAs (Substrate)
2
1
Current
paths
Substrate
Elliptical
laser
beam
Substrate
Electrode
J > J t h.
SS 2012
Optoelectronics
(Prentice Hall)
115
BuriedHeterostructure LDs
Althoughthestripeelectrodegeometryprovides
somegeometricconfinement,itismore
advantageoustorestrictlateralgeometryphysically
throughtheuseofconfininglayersalongtheside
ofthediode
Creationofaopticalwaveguideinbothverticaland
horizontaldirectionsaidsinreducingopticalcavity
modesandpromotesconfinement
Significantlyreducescurrentdensityrequiredfor
stimulatedemission
SS 2012
116
ElementaryLaserDiode(LD)Characteristics
Longitudinalmode:lengthdetermined
Lateralmode:widthdetermined
Emissioniseithermultimodeorsingle
modedependingontheopticalresonating
structureandthepumpingcurrentlevel
Po = 5 mW
Dielectric mirror
P o = 3 mW
Fabry-Perot cavity
P o = 1 mW
(nm)
Length, L
Height, H
TFH
Width W
778
Diffraction
limited laser
beam
780
782
SS 2012
117
LaserPerformanceasaFunctionofTemperature
Outputspectrumandmodepropertiesare
temperaturedependant
SinglemodeLDsexhibitamodehopatcertain
temperaturescorrespondingtoachangein
peakemissionwavelength
Nominalwavelengthofthelaserincreases
slowlybetweenhopsduetochangein
refractiveindex,n,withtemperature
Slopeefficiencydetermineslaserefficiency
andisnotthesameasLEDconversion
efficiencystatedinchapter3.
slope =
TFH
P o (mW)
10
0 C
25 C
50 C
6
4
2
I (mA)
0
0
20
40
60
80
Output optical power vs. diode current as three different temperatures. The
threshold current shifts to higher temperatures.
Po
I I th
SS 2012
118
ph
, where ph is
Photon loss rate: dtph = ph
average time for a photon to be lost from
lasing cavity mode due to transmission at end facets,
scattering and absorption in the semiconductor
Use rate equ. results (p. 97 f):
dN
assume a) 2 = 21 = 21,sp ; b) 1 0 N1 0
N0
1+Wi s
R21,sp
1+Wi 21,sp
N =
I
N2
st
Nph =
I
( RRth 1) =
Jth ph J
( Jth
ed
1) =
= R th 21,sp
ph
(J
ed
Jth )
Pout =
TFH
1
C 21,sp
R21,sp
1+CNph 21,sp
Rmirror ) =
SS 2012
hc02 ph W (1Rmirror )
2enref
(J Jth )
119
Optimal out-coupling I
I
Tm = 1 Rm transmission coefficient
+
=
dt
dt
dt
loss
gain
1/ph (Tm )
dNph
dt
loss,out
dN
+ dtph
loss,else
dN
+ dtph
gain
}|
{
1
Tm c0
= [ (
+
) +N2 C ]Nph
2nL
loss
[Tm (R C 21,sp
1)] 0
dTm
dTm
Tm c0 loss + 2nL
v
u 2Ln
p
2Ln
2Ln
Tmopt =
+u
Floss,int + g0 Floss,int
C R21,sp
t
| {z } c0 loss
c0 loss
c0
N 0 N20
TFH
120
Optimal out-coupling II
From A. Yariv, Quantum Electronics, 3rd edition, John Wiley & Sons, (1989)
TFH
SS 2012
121
DistributedBraggReflection
forSingleModeLasers
Ensuresinglemoderadiationinthelasercavityistousefrequencyselectivemirrorsatthe
cleavedsurfaces
DistributedBraggreflectorisamirrorthathasbeendesignedareflectiveBragggrating
Reflectedwaveoccursonlywhenthewavelengthcorrespondstotwicethecorrugation
periodicity,.
Thediffractionorderofthereflectorisinteger,q=0,1,2,
B
n
N=refractiveindexofthemirror
BraggwavelengthofthemirroroutputisB
Corrugated
dielectric structure
= 2
q(B /2n) =
A
B
Active layer
Distributed Bragg
reflector
(a)
(b)
(a) Distributed Bragg reflection (DBR) laser principle. (b) Partially reflected waves
at the corrugations can only constitute a reflected wave when the wavelength
satisfies the Bragg condition. Reflected waves A and B interfere constructive when
q(B/2n) = .
1999 S.O. Kasap, Optoelectronics (Prentice Hall)
TFH
SS 2012
122
DistributedFeedbackLaser(DFB)
InaDFBlaser,thecorrugatinggratingiscalledtheguidinglayerandrestontopofthe
activelayer.
ThepitchofthecorrugationprovidesopticalgainattheBraggwavelength,B.
Travelingwavesareexcitedbytheactivelayerandcoupletotheguidinglayerasthey
reflectbackandforthacrossthegratingtogenerateallowedDFBmodesthatarenot
exactlymatchedtotheBraggwavelength,butareplacedsymmetricallyjustofftheideal
modeoftheguidinglayeratm.
m = B
Corrugated grating
B 2
2nL
(m + 1)
Ideal lasing emission
Optical power
Guiding layer
Active layer
0.1 nm
(a)
(b)
(nm)
(c)
(a) Distributed feedback (DFB) laser structure. (b) Ideal lasing emission output. (c)
Typical output spectrum from a DFB laser.
1999 S.O. Kasap, Optoelectronics (Prentice Hall)
TFH
SS 2012
123
CleavedCoupledCavityLaser
DevicehastwodifferentopticalcavitiesoflengthLandD.
Eachlasercavityispumpedbyadifferentcurrent
Onlymodesresonantinbothcavitiesareallowedtoresonatethroughtheentiredevice,
allowingtheengineertotuneoutcertainmodesfromoneorbothindependentlaser
diodes
Whypumpboththecavities?Ans.Allowedmodesinanunpumped cavitywillundergo
recombinationifthedeviceisnotdriven.
Cavity Modes
In L
Active
layer
D
(a)
In D
In both
L and D
(b)
TFH
SS 2012
124
QuantumWell(QW)Devices
Devicewithanultrathin(50nm)narrowbandgap activeregionbetweentwowiderbandgap
semiconductors
AssumethatinQWdevicesthatthelatticematchsothatallthesemiconductorshavethesame
latticeconstantasothatcrystallinedefectsareminimized
Badgap changesattheinterfacearethereforeonlyduetodiscontinuitiesbetweenEc andEv of
thedifferingmaterialsyieldingdiscreteallowablequantumstatesthatcanbesolvedasparticle
inaboxtypeproblems.
E
Dy
E
AlGaAs
n=1
Eg2
QW
Ev
2 2me
[E V ( x ) ] = 0
+
x 2
h
Eg1
y
x
Energyinaquantumwell
Bulk
E3
E2
E1
Ec n = 2
d
D z AlGaAs
QW
Ec
Ev
(a)
Bulk
g(E)
Density of states
GaAs
(b)
(c)
A quantum well (QW) device. (a) Schematic illustration of a quantum well (QW) structure in which a
thin layer of GaAs is sandwiched between two wider bandgap semiconductors (AlGaAs). (b) The
conduction electrons in the GaAs layer are confined (by Ec) in the x-direction to a small length d so
that their energy is quantized. (c) The density of states of a two-dimensional QW. The density of states
is constant at each quantized energy level.
E = Ec +
h 2 n 2y
h 2 n 2z
h2n2
+
+
8 m e* d 2 8 m e* D y2 8 m e* D z2
n , n y , n z = 1, 2 ,3,...
Note:potentialenergybarrierofthe
conductionbandisdefinedbyw.r.t.Ec
TFH
SS 2012
125
EnergySpectruminaQuantumWell(SQW)
Ec
E
E1
h = E1 E1
E1
Ev
TFH
SS 2012
126
Example:AGaAs QW
GaAs QW
Effectiveelectronmassisme*=0.07me
WhatarethefirsttwoelectronenergylevelsforaQW
ofthickness10nm?
WhatistheholeenergybelowEv iftheeffective
electronmassofthehole,mh*=.5me?
Whatistheemissionwavelengthw.r.t.bulkGaAs
whichasanenergybandgap of1.42eV?
n =
h2n2
= 0 .0537 eV
8 m e* d 2
n' =
h 2n 2
= 0 .0075 eV
8 m h* d 2
g =
hc 1240 eV nm
=
= 874 nm
Eg
1 .42 eV
QW
TFH
1240 eV nm
=
=
= 839 nm
(1 .42 + 0 .0527 + 0 .0075 )eV
E g + n + n'
hc
Differenceinemissionwavelengthbetweenabulk
GaAs LDandaQWLDis35nm
SS 2012
127
DensityofStatesfromourParticleina
BoxSolution(QWDevice)
TFH
SS 2012
128
TFH
SS 2012
129
DOSDifferentQuantumWellsTypes
for 0,1,2 dimensional quantum
BasedonGeometry
confined structures
TFH
SS 2012
130
TFH
SS 2012
131
VerticalCavitySurfaceEmittingLasers(VCSELs)
Alternatinglayersoflowandhighindexaboveandbelow
theQWregioncreatesadistributedBraggreflectorof
dielectricmirrors
Themirrorsareneededtomatchtheopticalgainlostby
theshortcavitylength.Thuswiththemirrorsthelight
passesthroughthecavitysome2030timestoobtaina
desiredreflectanceof99%
Thehighreflectanceincreasesthegeometriccomponent
ofthegainrequiredforlaseremission
Contact
/4n 2
/4n 1
Dielectric mirror
Active layer
Dielectric mirror
Substrate
Contact
Surface emission
Constructiveinterferenceofpartiallyreflectedwaves
Ofwavelength,,attheinterface
n1d1 + n2 d 2 =
TFH
SS 2012
132
VCSELAttributes
TFH
VCSELactivelayersaregenerallyverythin
0.1umandcomprisedofMQWforimprovedthresholdcurrent
Thedeviceiscomprisedofepitaxially depositedlayeronasuitablesubstratewhichistransparentinthe
emissionwavelength
Ex.980nmVCSELdevices
InGaAs istheactivelayer
GaAs isthesubstrate
AlGaAs withdifferentcompositionscomprisethedielectricmirrorstack
ThetopstackisthenetchedafterallthelayershavebeendepositedtocreatetheinvertedTshape
presentedinthepreviousslide
Inpractice,currentflowingthroughthedielectricmirrorsgivesrisetoanundesiredvoltagedropthat
makesthedeviceVERYsensitivetofailurefromelectrostaticdischarge.Infact,thisisthemostcommon
failuremodeduringVCSELoperationandinstallation.
Theverticalcavityandthustheemittedbeamisgenerallycircularincrosssection
Theheightoftheverticalcavityisseveralmicrons.Thusthelongitudinalmodeseparationissufficiently
largetoallowonlyonemodeofoperation.Howeverlateralmodesmaybepresentincertaincavity
geometries
InpracticeVCSELShaveseverallateralmodesbutthespectralwidthisonlynmwhichissubstantially
lessthanthelongitudinalmodesofaDFBorELD.
Also,VCSELShaveanaveragebeamdivergenceofabout812o dependingontheirfabricationand
materialsused
DualwavelengthVCSELemissionisobtainedbyoperatingathighcurrents.
SS 2012
133
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
1. Introduction
It has been long acknowledged that a monolithically integrated laser for silicon (Si) based
photonic circuits would be an enabling technology that could accelerate the implementation of
silicon photonics significantly [1]. Early attempts to integrate III-V semiconductor lasers on a
silicon platform had only limited success [2, 3]. More recently, germanium (Ge) has been
suggested as a gain medium for lasing on Si [4]. Using a combination of tensile strain and ntype doping, efficient direct bandgap emission of Ge can be achieved [5]. Optically pumped
TFH
- $15.00 USD
SS 2012#164840
(C) 2012 OSA
Received 19 Mar 2012; revised 24 Apr 2012; accepted 27 Apr 2012; published 2 May 2012
7 May 2012 / Vol. 20, No. 10 / OPTICS EXPRESS 11316
134
lasing in Ge was demonstrated using a Ge waveguide with polished facets [6]. Furthermore,
attempts in electrically injection have demonstrated pin and pnn Ge diodes emitting between
1590 and 1700nm [710]. Here we present an electrically pumped pnn Ge diode laser that can
be monolithically integrated into a CMOS process. These first laser devices produce more
than 1 mW of output power and exhibit a Ge gain spectrum of over 200nm.
2. Experiments and results
Initial estimates of gain in n-type Ge based on experimental results showed that an n-type
doping level of 1x1019cm3 would yield a gain of about 50 cm1 [11]. Such a gain can lead to
lasing when pumped optically because optical losses are mainly limited to facet losses and
free carrier losses in Ge. For electrical pumping, additional losses due to the electrical
contacts, free carrier losses in doped poly Si and losses due to the interaction with the contact
metal, have to be overcome. Modeling of mode propagation in Ge waveguides with electrical
contacts shows that these additional losses are >100 cm1.To overcome by these losses, the Ge
gain must be increased by increasing the n-type doping to a level of 3-5x1019 cm3 [2].
Recently, we achieved n-type doping levels of > 4x1019cm3 by using a delta-doping
technique during epitaxial growth of Ge [12]. By correlation of photoluminescence (PL)
intensity, n-type doping level, and measured material gain, we have determined that an n-type
doping level of 4x1019cm3 corresponds to a material gain of >400cm1, enough to overcome
the losses in an electrically pumped laser device.
Ge waveguides of 1m width were fabricated by selective growth of n-type Ge-on-Si in
silicon oxide trenches using Ultra-High Vacuum Chemical Vapor Deposition (UHV-CVD)
[3]. A delta-doped Ge layer was grown on top of the n-type Ge to serve as a phosphorous
diffusion source [12, 13]. The delta-doping technique inserts monolayers of P in the Ge film
at low temperatures by alternating the phosphine and germane gas flow in the CVD reactor.
After thermal annealing to drive the phosphorous into the n-type Ge layer, the delta-doped Ge
layer was removed during planarization using chemical mechanical polishing (CMP), to reach
a uniform doping concentration in the gain medium. The remaining thickness of the Ge
waveguide after CMP varied between 100 and 300nm depending on wafer and location on the
wafer. Due to severe dishing of the waveguides after CMP the supported optical modes in the
waveguides could not be determined exactly. Up to six cavity modes can be supported in the
largest waveguides. An 180nm thick amorphous-Si film was then deposited via a PlasmaEnhanced CVD process and subsequently phosphorus-implanted to a doping level of
1020cm3. After a dopant activation anneal at 750C, a metal stack, consisting of Ti and Al
was deposited for top and bottom contacts. The oxide trench provides excellent current
confinement. In order to assure even carrier injection into the n-type Ge, the top contact metal
was deposited on top of the waveguide. After dicing, the waveguides were cleaved to expose
the Ge waveguide facets. A thin oxide layer was deposited on the facets to protect against
contamination and catastrophic optical mirror damage which was observed in devices that did
not have oxide protection.
TFH
Received 19 Mar 2012; revised 24 Apr 2012; accepted 27 Apr 2012; published 2 May 2012
7 May 2012 / Vol. 20, No. 10 / OPTICS EXPRESS 11317
SS 2012
The waveguide emission was measured using a Horiba Micro PL system equipped with a
cooled InGaAs detector with lock-in detection. The emission power measurement was
calibrated using light from a commercial 1550nm laser that was coupled into a single mode
optical fiber with the fiber end at the sample location. In the calibration we verified that the
detection was linear with input power. The electrical pumping was supplied by a pulse
generator with current pulse widths in the range of 20 s to 100 ms. The duty cycle was
varied between 2 and 50%, typically 4% to reduce electrical current heating effects. The laser
was contacted with metal probes and the current was measured using an inductive sensor
placed directly in the biasing circuit. The experimental set-up is shown in Fig. 1.
Fig. 2. Ge laser emission spectrum before (a) and after (b) threshold. The cavity length of the
waveguide is 333m and the waveguide height about 100nm. Current injection employed pulse
widths of 50s at 800Hz and 15C. The detector spectral resolution was 1.2nm.
Figure 2 shows the spectrum of an electrically pumped Ge laser below and above
threshold. The broad, direct band gap related electroluminescence spectrum, observed for
highly doped n-type Ge LEDs, has been reported earlier [14]. The spectra in Fig. 2 employed
short integration times to assure wide spectrum analyses. Measurement time for these large
laser devices is ultimately limited by metal contact breakdown due to the high current flow.
Figure 2(a) shows no spectral features above the noise floor. When the injection current
density is increased above threshold, sharp laser lines appear, as shown in Fig. 2(b). The
observed linewidth of the individual lines is below 1.2nm, the spectral resolution of the
measurement set-up. All measurements were performed with the samples mounted on a
thermo-electric cooler at 15C. Local device temperatures, however, are likely higher due to
the high current injection but could not be reliably determined.
Received 19 Mar 2012; revised 24 Apr 2012; accepted 27 Apr 2012; published 2 May 2012
7 May 2012 / Vol. 20, No. 10 / OPTICS EXPRESS 11318
135
Fig. 3. L-I curve for a 270m long waveguide device. 40s electrical pulses were used at
1000Hz. Measurement temperature was 15C.
Figure 3 shows the L-I spectrum for a typical electrically pumped Ge waveguide laser.
The lasing threshold at about 280kA/cm2 is clearly visible. This measurement was taken with
the set-up in Fig. 1 using a wide instrumental spectral resolution of 10nm, at a wavelength of
1650nm, monitoring a single laser line. The number of datapoints is limited by metal contact
breakdown at high current level. The optical emission power of about 1 mW corresponds to
Fig. 3. Occasionally we observed up to 7 mW. The spectrum in Fig. 2 shows two lines. The
estimate of the cavity free spectral range is 1nm, and the line spacing in Fig. 2, 3nm, is a
possible multiple of the FSR.
These lasers show a dependence of emission wavelength on threshold current density that
is consistent with the expected modal loss variation and that confirms the theoretical
conclusions that the gain spectrum of Ge for the given doping level and strain reaches over
more than 100nm spectral width [15]. For high doping levels of 4x1019cm3, and tensile strain
of ~0.2%, we observed lasing in the range from 1520nm to 1700nm. Figure 4 shows selected
laser lines between 1576nm and 1656nm for different Fabry-Perot cavities of the same gain
material.
devices, the Ge waveguide height is directly related to the modal loss. Since Ge has the
highest refractive index in our device structure, thinner Ge layers expel more of the resonant
mode into the highly-doped poly-Si cladding and into the lossy metal contacts. The
wavelength corresponding to the Ge gain peak (the threshold injection level) and the cavity
loss is the expected emission wavelength of the device. As the modal confinement decreases
with decreasing Ge layer thickness, modal loss and correspondingly threshold current
increases and the emission wavelength blue shifts. In Fig. 5, we show spectral threshold
conditions for two different modal cavity losses using a parabolic band model as described in
[8].
A Ge waveguide of 300nm thickness has a modal loss of about 90 cm1 due to losses in
the doped poly Si and the metal electrode (solid line). A Ge waveguide of 100nm thickness,
however, has a modal loss of about 1000cm1 due to the closer proximity of the mode to the
electrode (dashed line). To overcome the high losses of the thin Ge waveguide, a relatively
high carrier injection level is needed. Lasing is therefore expected at around 1520nm, close to
what we find in Fig. 2. For lower loss waveguides we expect lasing to occur at longer
wavelengths as shown in Fig. 4.
Fig. 5. Simulation of gain clamping condition for two different Ge waveguide thicknesses
(100nm: solid line; 300nm: dashed line). The axes plot the corresponding modal loss and gain
spectrum for the two different injection levels that are needed to overcome the respective
modal losses and to achieve lasing.
3. Conclusions
We have observed lasing from electrically pumped n-type Ge Fabry-Perot cavities. The
threshold current densities decrease with increased modal confinement. The emission
linewidth is less than the 1.2nm resolution of our measurement. Laser emission wavelengths
were observed between 1520nm and 1700nm with a variation consistent with the gain
clamping condition for each device. Measured output powers greater than 1 mW at room
temperature were measured. Improvements in the Ge growth, electrical contacts, and in modal
loss reduction will decrease the lasing threshold to values comparable with Fabry-Perot diode
lasers. The high power and observed gain spectrum of nearly 200nm indicate that the Ge laser
could be used for WDM applications. Since the laser can be monolithically integrated into any
CMOS process flow, novel device applications and systems can be developed.
Acknowledgments
Fig. 4. Spectra of Ge lasers with different Ge waveguide heights. The measured laser line
wavelengths are (a) 1576nm, (b) 1622nm, and (c) 1656nm.
TFH
The CMP-induced variation in cavity height provides self-consistent evidence of the wide
gain spectrum and the gain clamping condition by lasing. Under lasing action optical gain
(and population inversion) is clamped at exactly the value of the resonant cavity losses. In our
This work was supported by the Fully Laser Integrated Photonics (FLIP) program under APIC
Corporation, supervised by Dr. Raj Dutt, and sponsored by the Naval Air Warfare Center Aircraft Division (NAWC-AD) under OTA N00421-03-9-002. R.E.C.-A. was supported by a
NSF Graduate Research Fellowship award number 1122374.
Received 19 Mar 2012; revised 24 Apr 2012; accepted 27 Apr 2012; published 2 May 2012
7 May 2012 / Vol. 20, No. 10 / OPTICS EXPRESS 11319
SS 2012
Received 19 Mar 2012; revised 24 Apr 2012; accepted 27 Apr 2012; published 2 May 2012
7 May 2012 / Vol. 20, No. 10 / OPTICS EXPRESS 11320
136