Pro essing, Material properties and
Dispersion ee ts
AUDREYBERRIER
TRITA-ICT/MAPAVH Report2008:7
ISSN 1653-7610
ISRNKTH/ICT-MAP/AVH-2008:7-SE
ISBN 978-91-7178-969-3
Mi roele troni sandAppliedPhysi s
KungligaTekniskahögskolan
Ele trum229
SE-16440Kista
AkademiskavhandlingsommedtillståndavKunglTekniskahögskolanframlägges
till oentliggranskningföravläggandeavTeknologiedoktorsexamenfredagenden
30maj 2008kl.10:00iN1,Ele trum3,KunglTekniskahögskolan,Sto kholm.
Contents iii
A knowledgments vii
Listofpapers ix
A ronyms xiii
1 Introdu tion 1
1.1 Ba kgroundandMotivation . . . 1
1.2 AimandOverviewoftheoriginalwork. . . 4
1.3 Outlineofthethesis . . . 5
2 Photoni rystals: general on epts 7 2.1 Lightpropagationinperiodi media . . . 7
2.2 Two-dimensionalPhClatti e . . . 8
2.3 PlanarPhotoni Crystals . . . 8
Highindex ontrastsystem . . . 9
Lowindex ontrastsystem . . . 9
2.4 Lightpropagationinplanarphotoni rystals . . . 10
2.5 Computationalmethods . . . 13
PlaneWaveExpansion . . . 13
FiniteDieren eTimeDomain . . . 14
2.6 Leaky modes . . . 15
3 Fabri ation steps forphotoni rystals 17 3.1 Epitaxy . . . 17
3.2 Maskdeposition . . . 17
3.3 Ele tronBeamLithography . . . 18
Ebeamresists . . . 18
TheRaith150E-beamlithographysystem. . . 19
3.4 Maskopening-Rea tiveIonEt hing . . . 22
3.6 Post-et hingpro esssteps . . . 23
4 Chemi ally Assisted Ion Beam Et hing of InP-based photoni rystals 25 4.1 Chemi allyAssistedIonBeamEt hing . . . 25
4.2 Featuresize dependen e oftheet hing . . . 27
APhysi o- hemi almodelforInPet hing . . . 27
Experimental hara terizationofthelag-ee t. . . 29
Roughnessdevelopment . . . 32
Impa tontheopti alproperties . . . 33
4.3 Briefoverviewofsomedry-et hingpro esses. . . 33
4.4 Con lusion . . . 35
5 Opti alpropertiesofthe fabri atedphotoni rystal mirrorand avities 37 5.1 Internal LightSour e method . . . 37
Des riptionoftheinvestigatedPhCstru tures . . . 39
5.2 Opti allosses . . . 40
Intrinsi lossparameter . . . 41
Extrinsi lossparameter . . . 42
Optimization ofverti alheterostru turewaveguide . . . 43
5.3 Impa t ofthe featuresize dependen e oftheet hing ontheopti al propertiesofPhCs . . . 45
5.4 Con lusion . . . 47
6 Carrierlifetimesin et hedphotoni rystals 49 6.1 Surfa e/Interfa eStates . . . 49
6.2 Carrierdynami sinsemi ondu tormaterials . . . 50
6.3 Timeresolvedphotolumines en espe tros opy . . . 51
6.4 Modi ationof arrierlifetimes . . . 52
Sampledes ription . . . 52
Inuen eofthemaskmaterial . . . 52
Eviden efora umulatedsidewalldamage. . . 52
Non-radiativesidewallre ombinationvelo ity . . . 53
6.5 Modelingofthea umulateddamageinCAIBE . . . 54
Sputteringtheoryappliedto photoni rystalet hing . . . 54
Inuen e oftheholeshape. . . 56
6.6 Con lusion . . . 57
7 Lateral ele tron transport through photoni rystal elds 59 7.1 Ele trontransportthroughaphotoni rystaleld . . . 59
Modelingthe arriertransport . . . 59
Sample fabri ation . . . 64
Current-Voltagemeasurements . . . 65
Thermalee tsdue to arrierheating . . . 65
7.3 Et hing indu edmodi ationofsurfa epotential . . . 66
Inuen e of dry et hing on ele tri al properties of semi ondu tor surfa es . . . 67
7.4 Con lusion . . . 69
8 Some sele ted dispersion properties of photoni rystal based devi es 71 8.1 Blo hmodesintwo-dimensionalphotoni rystals . . . 71
8.2 Negativerefra tioninPhotoni rystals . . . 75
Negativerefra tionintwo-dimensionalInPbasedphotoni rystals . 76 8.3 End-re hara terization . . . 77
8.4 Experimentalinvestigationsofnegativerefra tion . . . 78
Lightfo using. . . 78
LightColle tionusingNegativeRefra tion. . . 79
8.5 FourierOpti s. . . 80
Experimentalset-up: prin ipleofmeasurement . . . 81
Auto- ollimation . . . 82
Visualizationoftheex itedBlo hmodesinthe2DPhCeld . . . . 83
8.6 Groupvelo itydispersioninphotoni rystalwaveguides. . . 83
PhC waveguides . . . 84
Slowlight . . . 85
Coupled avitywaveguides. . . 85
Phase-shift te hnique. . . 87
Measurementresults . . . 88
8.7 Con lusion . . . 89
9 Summary, on lusions and future work 91 9.1 A hievements . . . 91
9.2 Futurework . . . 92
10Guide to the papers 95
Abstra t
Photoni rystals (PhCs) are periodi diele tri stru tures that exhibit
aphotoni bandgap, i.e., arange ofwavelengthfor whi hlight propagation
is forbidden. The spe ial band stru turerelated dispersion properties oer
a realmof novel fun tionalities and interesting physi al phenomena. PhCs
havebeen manufa turedusing semi ondu torsand other material
te hnolo-gies. However, InP-basedmaterialsare themain hoi efora tivedevi esat
opti al ommuni ation wavelengths. Thisthesisfo uses ontwo-dimensional
PhCsintheInP/GaInAsP/InPmaterialsystemandaddressestheir
fabri a-tion te hnology and their physi al properties overing both material issues
andlight propagationaspe ts.
Ar/Cl
2
hemi allyassistedionbeamet hingwasusedtoet hthephotoni rystals. Theet hing hara teristi sin ludingfeaturesizedependentet hingphenomenawereexperimentallydeterminedandtheunderlyinget hing
me h-anismsareexplained. Fortheet hedPhCholes,aspe tratiosaround20were
a hieved,withamaximumet hdepthof5
µ
mforaholediameterof300nm. Opti al losses in photoni rystal devi es were addressed both in terms ofverti al onnementandholeshapeanddepth.Theworkalsodemonstrated
thatdryet hinghasamajorimpa tonthepropertiesofthephotoni rystal
material. Thesurfa eFermilevelat theet hedhole sidewallswasfound to
be pinned at 0.12eV below the ondu tion bandminimum. This is shown
to haveimportant onsequen es on arrier transport. It is alsofound that,
for anInGaAsPquantumwell, the surfa ere ombinationvelo ity in reases
(non-linearly)by morethanoneorder ofmagnitudeas the et hduration is
in reased,providingeviden efora umulationofsidewalldamage. A model
basedonsputteringtheoryisdevelopedtoqualitativelyexplainthe
develop-mentofdamage. Thephysi sofdispersivephenomenainPhC stru turesis
investigatedexperimentallyandtheoreti ally. Negativerefra tionwas
experi-mentallydemonstratedatopti alwavelengths,andappliedforlightfo using.
Fourier opti s was used to experimentally explore the issue of oupling to
Blo h modes insidethe PhCslaband to experimentallydetermine the
ur-vature of the band stru ture. Finally, dispersive phenomena were used in
oupled- avitywaveguidestoa hieveaslowlight regimewithagroupindex
ofmore than180 andagroupvelo ity dispersion upto 10
7
times that ofa
onventionalber.
Keywords:Photoni rystals,indiumphosphide, hemi allyassistedion
beamet hing,lagee t, avities,opti allosses, arriertransport, arrier
The a omplishment of a PhD thesis is a long journey. Many people made this
workpossibleandhelpedmegothrougheventhemostdi ultmoments. Iwould
liketothankthemall:
Gunnar Landgren, for a epting me as a PhD student at the department of
Semi ondu torMaterials,
SebastianLourdudoss,foralwayshavingtimeto listentomy on ernsand for
sharinghis onstantgoodmood aswellas forhis areandsupport,
Srinivasan Anand, my thesis supervisor, my main support through all those
years. It is animpossible taskto express the fullextent of mygratitude towards
him. Iwouldliketo thankhimherefortheimmensityofknowledgehetransfered
to me on the way to be ome a good s ientist and to grow as a person (what a
omplextask!). Alwaysfor ing me to think along dierent paths, he guided me
byindi atingthepresen eofdoors butallowingmetotakethene essarystepsto
openthem. Keepinganeyeonmyprogresses,hehasalwaysbeenreadytoredire t
mewhen ne essary. I would also liketo thank himfor alwaystaking the timeto
put meba k ontherighttra kwhen I myselfsaw onlydead-endsin front ofme.
His ontinuoussupport,boths ienti allyandmorally,ishighly a knowledged.
The ollaboratorsfromotherlabsIhadthepleasureto loselyworkwithwithin
theframeworkoftheEuropeanNetworkofEx ellen eonPhotoni Integrated
Com-ponents and Cir uits, and from whom I learned a lot: Anne Talneau from the
LaboratoryofPhotoni sandNanostru tures,Mar oussis,Fran e;RolandoFerrini,
Ni olas Le Thomas and Romuald Houdré from E ole Polyte hnique Fédérale de
Lausanne,Switzerland,
DavidHavilandandAndersLiljeborgfromtheNanophysi sLaboratory(KTH)
forprovidinga esstotheele tronbeamlithographysystem,
BozenaJaskorzynskaforprovidinga ess totheend-resetupat KTH,
My former olleagues, and nowadays lose friends: Olivier Douhéret for his
spe ial humor and his onstantpresen e when times were di ult; Mikaël Mulot
fortea hing me so mu h aboutpro essesand methods in my rstyearas aPhD
studentandforthenumberdis ussions,s ienti ornot,westillhave. Thememory
ofhiss ienti rigorandpre iseknowledgehasbeenamodelformeallalongthese
Allthepeoplehelping meoutin the leanroom. Inparti ularGunnar
Anders-son, Sven ValerioandPeterGoldmanfor theire ientte hni alsupport, and for
alwaystakingtimeto sharetheirknowledgewithme,
MarianneWidingforher onstant areande ientadministrativesupport,
JesperBerggrenforalwaysbeingreadytohelp,
WoutervanderWijngaartforintrodu ingmetoresear h,
The olleagues and friends I had the pleasure to work with: Yao heng Shi,
Mar inSwillo,Ri haTyagi,ShaguftaNaureen,NaeemShahid,JörgSiegert,Saulius
Mar inkevi ius,HenrikJohansson,JoakimStrandberg,GunnarMalm,
All my olleagues and friends from the Mi roele troni s and Applied Physi s
Department,A reoand ReplisaurusIhadthepleasure tosharemomentsand
dis- ussionswith. Inparti ular,ThomasAggerstam,NadeemAkram,SussiAlmqvist,
Ce ilia Aronsson,YohannesAssefaw-Redda,XavierBadel,JesperBerggren, Hans
Bergkvist,StefanoBonetti,MarekCha inski,SilviaCorlevi,MatteoDainese,
Beat-rizEspinoza,RezaFatehnia,Ming-HongGau,Os arGustafson,MattiasHammar,
Julius Hållstedt,Carl Junesand, StephaneJunique, Sören Kahl,Joo-HyungKim,
Magnus Lindberg, Kestutis Maknys, Ri kard Mar ks von Württemberg, Fredrik
Olsson, Min Qiu, Henry Radamson, IsabelSainz Abas al, Yan-Ting Sun, Petrus
Sundgren, IlyaSy hugov, Qin Wang, Ma iej Wolborski, Le h Wosinski, Xingang
Yu, Zhenzhong"Andy"Zhang,Zhi-BinZhang,
All my friendsin Sto kholm and in therest of theworld,for forgivingme for
mylongperiodsofsilen e andforoeringmetheirsupport,
Myparentsfor their onstant are andsupport, and forallowingmeto follow
myownpath.
Publi ations in luded in the thesis
A A.Berrier ,M.Mulot,A.Talneau,R.Ferrini,R.Houdré,andS.Anand
Cha-ra terizationofthefeaturesizedependen einAr/Cl
2
hemi allyassistedionbeamet hingof InP basedphotoni rystaldevi es, J. Va . S i. Te hnol.
B.25(1)1-10(2007).
B R.Ferrini, A.Berrier , L.A.Dunbar, R.Houdré, M.Mulot, S.Anand, S.de
Rossi, A.Talneau Minimization of out of plane losses in planar photoni
rystals by optimizing the verti al waveguide, Appl. Phys. Lett. 85(18),
3998-4000(2004).
C A.Berrier ,R.Ferrini,A.Talneau,R.Houdré,andS.Anand,Impa tof
fea-ture size dependen e on theopti al properties of two-dimensional photoni
rystaldevi es, J. Appl. Phys.,103(9)(2008).
D A.Berrier , Y.Shi, J.Siegert, S.Mar inkevi ius, S.He, and S.Anand,
Evi-den eofa umulatedsidewalldamageindry-et hedInP-basedphotoni
rys-tals, submittedto Appl. Phys. Lett. (2008).
E A.Berrier ,Y.Shi,S.Mar inkevi ius,S.He,andS.Anand,Developmentof
damageanditsimpa tonsurfa ere ombinationvelo itiesindry-et hed
InP-basedphotoni rystals,,manus ript(2008).
F A.Berrier ,M.Mulot,G.Malm,M.ÖstlingandS.AnandCarriertransport
throughadry-et hedInP-basedtwodimensionalphotoni rystal, J. Appl.
Phys,101,123101(2007).
G A.Berrier ,M.Mulot,M.Swillo,M.Qiu,L.Thylén,A.TalneauandS.Anand,
Negative refra tionat infra-red wavelengths in a two-dimensional photoni
rystal, Phys. Rev. Lett.,93(7),073902(2004).
H A.Berrier ,M.Swillo,N.LeThomas,R.Houdré,andS.Anand,Blo hMode
Ex itationinTwo-DimensionalPhotoni CrystalsImagedbyFourierOpti s,
I A.Talneau,G.Aubin,A.Uddhammar,A.Berrier ,M.Mulot,andS.Anand,Highly
dispersivePhotoni Crystal-basedguidingstru tures, Appl. Phys. Lett. 88,
201106(2006).
Other journal publi ationsand onferen e pro eedings not
in luded in the thesis
J1 P.El-Kallassi,R.Ferrini,L.Zuppiroli,N.LeThomas,R.Houdré,A.Berrier ,
S.Anand, A.Talneau Opti al tuning of planar photoni rystals inltrated
withorgani mole ules, J. Opt. So . Am. B 24(9), 2165-2171(2007).
J2 L.Prkna,A.Talneau,M.Mulot,A.Berrier ,S.AnandEviden eofthe
Pho-toni Gap ontributiontotheguidingme hanismforstrongly onnedmodes
inthe"refra tive-like"domain, Opt. Lett. 31,2139(2006).
J3 M.J.Martin,T.Benyattou,R.Orobt houk,A.Talneau,A.Berrier ,M.Mulot,
S.Anand,Eviden e ofBlo h wavepropagation within photoni rystal
wa-veguides, Appl. Phys. B-LasersandOpti s 82(1)(2006).
J4 T.Benyattou,M.Martin, R.Orobt houk,A.Talneau,A.Berrier ,M.Mulot,
S.AnandOpti al Blo h wavesstudied by near opti aleld mi ros opy, J.
ofthe KoreanPhys. So . 47S72-S75Suppl. 1,(2005).
J5 S.deRossi,I.Sagnes,L.Legratiet,A.Talneau,A.Berrier ,M.Mulot,S.Anand,
J.L.Gentner Longitudinal Mode Sele tion in Constri tedPhotoni Crystal
GuidesandEle tri allyInje tedLasers,IEEEJour. Lightwave Te h. 23(3),
1363(2005).
J6 A.Talneau,J.L.Gentner,A.Berrier , M.Mulot, S.Anand,S.Olivier,High
externale ien yinamonomodefullphotoni rystallaserunder ontinuous
waveele tri alinje tion, Appl. Phys. Lett. 85(11),1913-1915(2004).
J7 M.Qiu,S.Xiao,A.Berrier ,S.Anand,L.Thylén,M.Mulot,M.Swillo,Z.Ruan,
S.He,Negativerefra tionintwo-dimensionalphotoni rystals,Appl. Phys.
A00,1-6(2004).
J8 W.vanderWijngaart,A.Berrier ,G.StemmeAmi ropneumati -to-vibration
energy onverter on ept, Sensorsand A tuatorsA: Physi al 100(1) 77-83
(2002).
J9 A.Berrier ,Y.Shi,J.Siegert,S.Mar inkevi ius,S.HeandS.Anand"Impa t
ofdry-et hing indu eddamage in InP-basedphotoni rystals" Pro . SPIE
Vol.6989,Photoni CrystalsMaterials andDevi es,p.698930(2008).
J10 A.Berrier , M.Mulot,A.Talneau,R.Ferrini,R.HoudréandS.Anand
"Fea-turesizeee tsin hemi allyassistedionbeamet hingofInP-basedphotoni
J11 A.Berrier ,M.Mulot,G.Malm,M.ÖstlingandS.Anand"Ele tri al
ondu -tionthrougha2DInP-basedphotoni rystal",Pro . SPIEVol.6322,Tuning
the Opti al Response ofPhotoni Bandgap Stru turesIII p.63220J(2006).
J12 A.Berrier ,M.Mulot,A.TalneauandS.Anand"Featuresizeee tsinAr/Cl
2
hemi ally assisted ion beam et hing of InP-basedphotoni rystals" IEEE
Pro . InternationalConferen eonIndiumPhosphide andRelatedMaterials,
Vol.2006,pp.340-3(2006).
J13 G.Aubin,A.Talneau,A.Uddhammar,A.Berrier ,M.MulotandS.Anand,"
Highlydispersivephotoni rystal-based guidingstru tures"Quantum
Ele -troni sandLaserS ien e Conferen e,Vol.2,pp.972-4(2005).
J14 B.Jaskorzynska,M.Qiu,A.Berrier ,M.Dainese,S.Anand,L.Wosinski,L.Thylén,
Z.J.Zawistowski"Appli ations of wavelengthdispersionin 1D and2D
pho-toni rystals" Pro . SPIE Vol.5959, Photoni Crystals and Fibres, pp.1-9
(2005).
J15 M.Qiu,Z.C.Ruan,ABerrier ,S.S.Xiao,M.Mulot,S.Anand,S.He,L.Thylén
"Negativerefra tioninsemi ondu torphotoni rystals"Pro . SPIEVol.5624
345-A ronyms
1D,2D,3D One-,two-,three-dimensional
BZ Brillouinzone
CAIBE Chemi allyassistedionbeamet hing
CBM Condu tionband minimum
CCW Coupled- avitywaveguide
EFC Equi-frequen y ontour
FDTD Finitedieren etimedomain
FIB Fo usedionbeam
FP Fabry-Pérot
FSDE Featuresizedependentet hing
FT Fouriertransform
FWHM Fullwidthat halfmaximum
GVD Groupvelo itydispersion
ICP Indu tively oupledplasma
ILS Internallightsour e
IV CurrentVoltage
LED Lightemittingdiode
MOVPE Metalorgani vaporphaseepitaxy
PhC Photoni rystal
PECVD Plasma-enhan ed hemi al-vapor deposition
PL Photolumines en e
PMMA Polymethylmetha rylate
PWE Planewaveexpansion
RIE Rea tiveionet hing
RF Radiofrequen y
SEM S anningele tronmi ros ope
TE Transverseele tri
TM Transversemagneti
QW Quantum well
Notations
a
PhC latti eperiodc
Lightvelo ityin va uumd
Holediameter~
E
Ele tri eldf
Air-llingfa tor~
H
Magneti eldh(t)
Depth ofthestraightportionoftheholeJ
Iondensity~k
Waveve tork
x
Propagation onstantinaPhC waveguideL(λ)
MirrorlossL(t)
HoledepthL
decay
De aylengthoftheele tri eldin thesubstrateL
p
Opti alpenetrationdepthinsideaPhCmirrorn
Refra tiveindexn
core
Refra tiveindexinthe orelayern
clad
Refra tiveindexinthe laddinglayern
eff
Ee tiverefra tiveindexoftheplanarwaveguiden
g
GroupindexQ
Qualityfa torofaresonant avityq
Elementary hargeR
Mirrorree tion~r
Positionve tort
TransmissionofaPhCstru tureT
Mirrortransmissionu
Normalizedfrequen y(= a/λ
)u
~
k
Blo hfun tionW
Mirrorspa ingofa1D avityv
g
Groupvelo ityV
bias
Self-indu ed biasin RIEV
−
,V
+
Ele tri potentialsofthegridsoftheCAIBE systemY
SputteringyieldSymbols
θ
c
Coneangleof aPhCholeα, α
i
Opti alabsorption oe ientα
c
Clausing'sfa torδ
DepletionregionwidtharoundthePhCholesǫ
,ǫ
′
Diele tri onstant
ǫ
′′
Lossparameter(imaginarypartof omplex
ǫ
ǫ
′′
int
Intrinsi lossparameterǫ
′′
ext
Extrinsi lossparameterΓ
core
Connementfa torinthe orelayerλ
Opti alwavelengthµ
′
Magneti permeability
η
in
Couplinge ien yη
out
Colle tione ien yω
Opti alfrequen yφ
,φ
i
Phaseshiftσ
Ele tri ondu tivityζ
Verti aleldproleτ
P L
Photolumines en ede aytimeτ
nonrad
, τ
Carrierlifetime(non-radiative)ν
S
Surfa ere ombinationvelo ityattheholesidewallsθ
Cl
Surfa e overage(probability)forClattheet hplaneIntrodu tion
1.1 Ba kground and Motivation
Themajorappli ationareasforphotoni saredisplays,lighting,health are
in lud-ingmedi aldiagnosis/therapy,sensing,and ommuni ationtonameafew. The
Eu-ropeanroadmapforphotoni sandnanote hnologies,MergingOpti sand
Nanote h-nologies(MONA)aresultofdis ussionsamongEuropeana ademi andindustrial
organizationsandreleasedearly2008[1℄,identiesnineeldsinwhi h
nanophoton-i swill haveamajor impa tin the omingyears: sensors, data and tele om,
data storage, at panel displays, imaging, instrumentation, LEDs and
lighting, opti al inter onne ts and photovoltai s. Ontheroad towardsthe
developmentofnovelte hnologiesin these appli ationareas,astrongresear h
ef-fort is devoted to new promising materials su h as semi ondu tor quantum dots
and wires, metalli nanostru tures, arbon nanotubes, fun tionalized
nanoparti- les,metamaterialsand high index ontrast nanostru tures. Inthis last material
type,photoni rystalsarehighlypromising andidatesfortheiropti alproperties.
Photoni rystals(PhCs)arediele tri stru tureswithaperiodi ityinthe
die-le tri onstantoftheorderofthewavelengthoflight. One-dimensional(1D)PhCs
arehistori ally alledBragggratings/mirrorsandhavebeenstudiedformanyyears
[2, 3℄, e.g., as mirrors in laser designs [4, 5℄ or for their dispersiveproperties [6℄.
Later the on ept of periodi variation of the diele tri onstant was extended
to two and three dimensions. Subsequent to the rst proposals of PhCs for
in-hibition of spontaneous emission [7℄ and photon lo alization [8℄, several types of
two-dimensional(2D)andthree-dimensional(3D)PhCshavebeensu essfully
fab-ri atedand hara terized. Insu h stru tures,theperiodi diele tri potential an
giverisetoaphotoni bandgap,similartothesituationforele tronsina
semi on-du tor rystalinwhi h asetheperiodi potentialisthatoftheatoms onstituting
the rystal. ItisforthisreasonthatsometimesPhCsarealsoreferredtoas
photo-ni bandgapmaterials. Theexisten eofabandstru turemodiesthedispersionof
theowoflight. Inanotherwiseperfe t rystalthelo almodi ationofthe rystal
latti eorbasiswillintrodu e"defe ts"inthe rystal,givingrisetothepresen eof
new statesin the bandgap. Line or point defe tresonators an be reatedin the
rystal. Su hanengineeringofdefe ts inthePhClatti eoersnewwaysto guide
and onnelightbytheformationofwaveguides[911℄,or avities[12,13℄.
Figure 1.1: a) GaAs-based 1D PhC stru ture; Courtesy of F.Raineri (LPN); b)
GaAs-based 1D PhC mirror in GaAs; Reprinted bypermission from Elsevier [5℄,
Copyright2001; ) InP-based2DPhCstru turefabri atedatKTH;d)Si3DPhC
stru ture; Reprinted by permission from Ma millan Publishers Ltd:Nature [14℄,
Copyright1998;e)GaAs-based3DPhCstru ture;Reprintedwithpermissionfrom
[10℄,Ameri anInstitute ofPhysi s,Copyright2006.
Photoni rystalshave thepotentialto play animportant role in most of the
domains identiedbytheMONAroadmap. They an bring ompa tnessand
im-proved sensitivity to sensors ( hemi al [1517℄, biologi al[18℄). Quantum as ade
lasers withPhCsare sour esofmid-infraredradiation whi h willallowintegrated
terahertz spe tros opy for hemi al and biologi alsensing [1921℄. For data and
tele omappli ations,e ientand ompa twaveguides,lightsour esand
photode-te tors among others are required. Dispersion engineering an enable new
fun -tionalities su h as dispersion ompensation, slow light, et . PhC-based designs
allowimprovedperforman esforlightemittingdevi es: LEDs an makeuseofthe
bright-displays. Similarly,tele omwavelengthlasers anbemademore ompa t[24℄with
lowpower onsumption -lowthreshold-[25, 26℄. Photodete torsusephotoni
rys-talbaseddesignstoimprovetheirquantume ien yinthevisible[17℄,atinfra-red
wavelengths[2729℄.
PhCdesignsarepotentiallyinterestingforintegrationandfornewfun tionality
in omponentsused in opti al inter onne tsfor lo al areanetworks and Fiber To
TheHomes hemes[30℄,withwavelengthmultiplexingandre ongurabilityas
ad-vantages. SlowlightmodesinPhCs anbeawaytoobtainopti aldelaylines. The
signi antmodedispersion losetobandedgesor losetomode ut-osof
oupled- avitysuper-modes an slowdownlight, whi h is interesting foropti al buering
andpro essing [31℄ andalso forenhan inglight-matter intera tions[32, 33℄. The
steadyin reaseofthestoragedensityofdatawill soonrequiretobeatthe
dira -tion limit of light to perform memory read-out. New ways to improve imaging
apabilities,whi harenotdira tionlimitedwouldbevaluabletowardsthatgoal.
Even thoughplasmoni materialshaveverygood prospe ts, tailoredPhC designs
antoyieldsuper-resolution[34℄.
Figure1.2: S hemati drawingofatwo-dimensionalphotoni rystalindi atingthe
typeofverti allight onnementa)lowindex ontrast,verti alheterostru ture;b)
high index ontrast,membrane. The orientationof the
x
,y
andz
axisas used in thisthesisarealsoshown.The te hnology for manufa turingphotoni rystalsfor operationat infra-red
wavelength or in the visible spe trum is hallenging, requiringhigh pre ision
na-nofabri ationtools. Eventhoughsomesu essfulattemptswerereportedbasedon
top-down[35℄orbottom-upapproa hes[36℄,3DPhCstru turesarefarfrom
realis-ti appli ationsdue tothehighlevelof omplexityoftheirfabri ation. Therefore,
2DPhCswithin-planeperiodi ity ombinedwitha onventionalslab onnement
in the third dire tion are nowadays the dominant te hnologi al solution. Figure
1.2showsas hemati drawingofaplanar2DPhCeldwithlowindex ontrastor
highindex ontrastverti allight onnement.
Whenthisthesisworkwasinitiated,theinterestforphotoni rystalswasmainly
fo used ondefe t engineering anddevi es operatingin thebandgap; examples of
ap-interest, espe ially for un onventional light propagation phenomena, su h as
su-perprism ee t [37℄ or negativerefra tion [38℄. Re ently, it hasbeen shown that
modi ationsofthePhClatti e[39,40℄oralo alalterationofthebasisby,e.g.,a
variationoftheholeradius
r
[41℄of thephotoni rystal an enhan e greatlythe opti alpropertiesofPhClasersand avities.For the broadrange of appli ations mentioned earlier, the main driving for e
is thesteadily in reasingintegration levelof omponentsand fun tionalities. The
integration of a tive fun tionalities still relies on III-Vmaterials. InP-based
ma-terials are the main hoi e for operation in the 1.3 to 1.7
µ
m range. Resear h on onventional InP-based devi es has now rea hed a ertain degree of maturityand te hnologyplatforms/foundriesare beingstarted. However,nanophotoni
de-vi es in luding photoni rystals are still far from maturity. The hallenges are
even higher onsidering that most appli ations require devi es with high quality,
low loss, high ompa tness, robustness, and ompatibility with ele tri al a ess.
Fun tionaldevi eswouldideallybeopti allya tive,beele tri ally ontrollableand
enableintelligentpro essingoflight. Towardstheseobje tivesanin-depth
under-standing oftheopti al, ele tri al anddispersionpropertiesofphotoni rystalsis
indeedhighlyvaluable. Atthesametime,knowledgeand ontroloffabri ation
pro- essesare loselylinkedto theperforman esofnanophotoni sdevi es. Thusthere
is aneed to understand and ontrol fabri ation te hniques for a mature
te hnol-ogyyieldingdevi eswithhighreprodu ibilityandquality. Fabri ationofphotoni
rystalstru turesis hallengingintermsofpro essa ura yandreprodu ibility
af-fe tingtheopti alpropertiesofthedevi esin ludingpropagationloss. Con erning
waveguiding andother passivewaveguidefun tions,sili onnanowires haveshown
better performan es. However, for a tive fun tionalities (light emission, opti al
signal pro essing, non-linear pro esses, et ) III-V materials are parti ularly well
suited.
1.2 Aim and Overview of the original work
Theawarenessand ontroloffabri ation andofthematerialpropertiesare
ne es-saryforhighperforman ePhCdevi esandintegrationofthese,in ludingele tri al
a ess. It iste hnologi ally ru ial for devi e performan eto understand the
im-pa t of pro essing on opti al losses, on the ele tri al propertiesof thematerial
-relevant in the ontext of ele tri ally a tivedevi es -, and on the opti al
proper-ties of thematerial -relevant foropti ally a tivedevi es. Further, theknowledge
of the dispersion properties of PhCs and how to engineer them is important for
novel fun tional devi es. Therefore there is aneed forunderstanding the physi s
of fabri ation-relatedissues, andhow theyae tlossesand materialpropertiesin
order to ontrolandpredi tpro essout omes.
Thepresentthesisaddressestheabovementionedissuesinthe ontextofplanar
Theperforman esofthe hemi allyassistedionbeamet hing (CAIBE)should
be evaluated in terms of a hievable aspe t ratios, feature size dependen e of the
holeshapeanddepth. Knowledgeaboutthephysi o- hemistryoftheet hingand
amodelfortheet hing hara teristi sarene essarytounderstandthe apabilities
ofthepro essandoriginsofitslimitations. Thisisthesubje tmatterofPaperA
andthe obtainedresultsaredire tly relevantfor optimizingdevi e performan es.
The impa t of the et hing hara teristi s on the opti al properties of photoni
rystalsu haslosses,ree tivityandqualityfa torfordevi esoperatinginsidethe
bandgap,areaddressedinPaper B and C
Dry-et hing, a ne essarypro ess stepin PhC devi e manufa turing, is known
tointrodu edamageintheet hedmaterial [42,43℄. Pro essindu eddamageand
itsimpa t on the ele tri al and opti al properties of PhCs need to be evaluated.
Further, theextentandnature of the reateddamage depends ontheet hed
ma-terial,theet hing te hniqueandpro essparameters. Inlightofthis,amethodto
assesstheee tofsidewalldamageontheele tri al ondu tioninPhCsisproposed
( PaperF).Theopti alpropertiesofthePhCmaterialwillalsobeae tedbythe
typeofet hing. Paper DandE presenttheexperimental hara terizationofthe
variationof the arrierlifetime of InGaAsP quantum wellsin PhCs as afun tion
ofthe et h parameters. Theoriginfor material damageis identiedand amodel
isdevelopedtopredi tthepro essdependentdefe t reationandthendingshave
importantimpli ationsfora tivePhCdevi es.
ThepotentialofdesigningdispersioninPhCstru turesisexperimentally
demon-stratedwithexamplessu hasnegativerefra tion,atlensingandauto- ollimation.
Theunderstanding of the oupling to PhC Blo h modes is bene ial in order to
de idehowtomakethebestuseofPhCstru turesbothinthegapandinthe
trans-mission bands. Band stru ture related dispersionee ts in PhC are investigated
in Papers G, H and I.Paper G gives anexperimental demonstration of light
fo usingusingnegativerefra tion. Thispaperalso showsitspotentialappli ation
forlight olle tion. PaperHisatheoreti alandexperimentalanalysisofthe
ex i-tationofBlo hmodesinaPhCeldoperatingabovethebandgap. InPaperItwo
designsof PhC-based oupled- avitywaveguidesare opti ally hara terizedin the
slowlightregime. Groupvelo itydispersionandthegroupindexaredetermined.
1.3 Outline of the thesis
Therest of the thesisis organizedas follows. The following hapter givesa
gen-eral introdu tion to the on epts used to des ribe the opti al properties of
two-dimensionalphotoni rystalsin ludinganoverviewofsomerelevant omputational
methods. Chapter3presentsthepro essstepsusedinthisworkforthefabri ation
of2Dphotoni rystals, andin ludes abriefdes riptionof themethods involved.
Chapter4givesadetailed des riptionof hemi allyassistedionbeamet hingand
sults on the evolution of PhC losses as a fun tion of the verti al onnement or
the PhC hole shape and depth. The results onthe inuen e of the PhC et hing
onthe arrierlifetimemeasuredbytime-resolvedphotolumines en earepresented
in Chapter6. Chapter7detailstheresultsobtainedontheele tri al propertiesof
PhC elds whi hprovidesanin-depth understandingof theele trontransport as
wellasthepro essindu edmodi ationoftheele tri alpropertiesofthePhC
ma-terial. Chapter8des ribesthenature andbehaviorof Blo hmodesaswellas the
anomalousdispersionregimesin2DPhC elds. This hapter alsosummarizesthe
resultsonnegativerefra tion,theobservationofBlo hmodesbyFourieropti sand
experimental measurementsofslowlightmodes. Finally,Chapter9givesthemain
Photoni rystals : general on epts
Photoni rystals(PhCs)areele tro-magneti stru tures hara terizedbythe
spa-tialperiodi ityofthediele tri onstant
ǫ(~r)
. Theperiodi ityallowsa mathemat-i aldes riptionbyintermsofalatti ewhi hbringstheanalogywiththeperiodiarrangement of atoms in a rystalline material, hen e the name " rystal". The
periodofthelatti ein PhCsisoftheorderofmagnitudeofthewavelengthofthe
"light", hen ethe term"photoni ". Most ommonly in PhCs, likein the present
work,
ǫ(~r)
is aperiodi repetitionofpo ketsofamaterialwithdiele tri onstantǫ
m
in aba kgroundmaterialwith diele tri onstantǫ
b
. One ouldimaginesome stru tureswithmore ompli atedfun tionsforthepermittivityǫ(~r)
orpermeabilityµ(~r)
.2.1 Light propagation in periodi media
Maxwell's equationsdes ribethe behaviorof ele tro-magneti wavesin any given
medium. For an ele tri ally polarizable, non-magneti , lo ally ma ros opi ally
isotropi andlosslessmaterialwherewenegle tfree hargesand urrentswehave,
inSIunits,in thelinearregime[44℄:
∇ · (ǫ
0
ǫ(~r) ~
E) = 0
(2.1)∇ × ~
H − ǫ
0
ǫ(~r)
∂ ~
E
∂t
= 0
(2.2)∇ · ~
H = 0
(2.3)∇ × ~
E + µ
0
∂ ~
H
∂t
= 0
(2.4)where
E(~r)
~
istheele tri eld,H(~r)
~
isthemagneti eld,ǫ
0
isthepermittivityin va uum,µ
0
thepermeabilityin va uumandǫ(~r)
isthelo al relativepermittivity.Byintrodu ingtimedependen eoftheeldsasanharmoni os illatingfun tion,
intoequations2.1to2.2weobtainaftersomealgebrathefollowingmasterequations
forthe
E
~
-eldandH
~
-eld,respe tively,ˆ
Ξ ~
E
ω
(~r) = (
ω
c
)
2
E
~
ω
(~r)
(2.5)ˆ
Θ ~
H
ω
(~r) = (
ω
c
)
2
H
~
ω
(~r)
(2.6)where
ω
is the y li frequen y,c
the light velo ity,Ξ =
ˆ
1
ǫ(~
r)
∇ × ∇×
andΘ =
ˆ
∇ ×
1
ǫ(~
r)
∇×
.2.2 Two-dimensional PhC latti e
Thefo usofthisthesisisontwo-dimensional(2D) airholephotoni rystals. We
will provide herethe on eptsneededto des ribe2DPhCs. We onsider here2D
PhCs made from a triangular latti e of airholes in adiele tri material, i.e., an
InP-basedheterostru ture. Thelatti e onstant
a
and thediameter oftheholesd
are dened on Fig. 2.1. Theair llfa torf
, usually expressed in %,is the ratio betweenthesurfa eo upiedbyairtothetotalsurfa eofthePhC.Itis al ulatedfrom aunit ellandisgivenby
f =
π
2
√
3
(
d
a
)
2
≈ 0.907d
2
/a
2
.The re ipro al latti e is generated by the basis ve tors
b
~
1
andb
~
2
onstru ted from thereal spa e basisve torsa
~
1
= (a, 0)
anda
~
2
= a(
1
2
,
√
3
2
)
and are given by~
b
1
=
2π
a
(1, −
1
√
3
)
andb
~
2
=
2π
a
(0,
2
√
3
)
. Figure 2.1 shows the real and re ipro al spa easso iatedwiththetriangularlatti e2DPhC,aswellasthemainsymmetrydire tions
Γ
MandΓ
K,andtherstBrillouinzone.Introdu tion of simple defe ts in a 2D PhC latti e in the form of removed
holes (or holes lled with a material dierent from the host matrix) allows the
formation ofwaveguides[9℄, resonators[12, 45℄. Re ent works haveshown that a
lo almodi ationofthelatti e(modi ationof
a
toformphotoni heterostru tures [39℄, shift in position of latti e points [46℄) or of the basis ([40, 47, 48℄) of thephotoni rystal anenhan etheopti alpropertiesgreatly.
2.3 Planar Photoni Crystals
By denition, two-dimensional rystals extend innitely in the dire tion normal
(verti aldire tion) to the plane of periodi ity. Thus for opti al devi es based on
PhCs, a onnementintheverti aldire tionisne essary. This onnementis
ob-tained by aplanar waveguide with the light onned in the ore layerof higher
refra tiveindex. Several solutionsfor this material sta kingare possibleand are
generally lassiedintohighorlowindex ontrastsystemsa ordingtothe
surround-Figure2.1: a)S hemati ofa2DPhCinrealspa e. The ir lesrepresentthebasis
(hole) repeated at ea h latti e point.
a
~
1
anda
~
2
are theprimitivelatti e ve tors. Thelatti e onstanta
and hole diameterd
are also shown; b) s hemati s of the re ipro al spa e orrespondingto thelatti eshownin a).b
~
1
andb
~
2
are thebasis ve torsof the re ipro allatti e. The rst Brillouin zone is shown, as well as thesymmetrypoints
Γ
,KandM.ingmaterial(s). Itistobenotedherethattheindex ontrastreferstotheverti al
stru tureandnotto theindex ontrastbetweenba kgroundandPhCholes/rods.
Highindex ontrast system
The high index ontrast domain is ommonly dened by the ondition
∆n ≥ 2
. Thematerials aboveandbelowthe orelayermaybemadeofdierent materials.Examplesinthis ategoryaremembranes[12, 49,50℄, semi ondu tor-on-insulator
stru tures where the ore layer is asymmetri ally bounded by air and a low
in-dexdiele tri layerwhi h ouldbeoxides [51, 52℄, polymers[53, 54℄ or lowindex
substrates(e.g., sapphire[55,56℄).
Low index ontrastsystem
Themost ommonwaytofabri atePhC devi esinthelowindex ontrastsystem
makes use of semi ondu tor heterostru tures to onne the light in the verti al
sket hofthisverti alstru tureindi atingtherespe tivevaluesofrefra tiveindex
(at 1.55
µ
m) as well as themodeproleζ(z)
ofthe thefundamental modeof the ele tri eldE(z)
~
in the waveguidefor TE polarization. The small dieren e inFigure 2.2: S hemati diagram of the verti al stru ture omposed of
InP/InGaAsP/InPas wellastheverti alproleoftheTEfundamental mode
refra tiveindi esdoesnotallowforastrongverti al onnement,i.e.,
ζ(z)
extends deepinto thesubstrate. Thisweak onnementofthemodeprolehassigni antimpli ations on the fabri ation of PhC devi es and on their opti al properties.
These issueswill begivenattentionin Chap.5.
2.4 Light propagation in planar photoni rystals
In asesforwhi hwe andenetwodistin tpolarizationstatesTE(E
z
=Hx
=Hy
=
0
) and TM(Hz
=Ex
=Ey
= 0
) the master equations for the z- omponent of the respe tiveeldsaregivenby:∂
∂x
1
ǫ(~r)
∂
∂x
H
z
+
∂
∂y
1
ǫ(~r)
∂
∂x
H
z
+
ω
2
c
2
H
z
= 0
(2.7)1
ǫ(~r)
∂
2
∂x
2
+
∂
2
∂y
2
E
z
+
ω
2
c
2
E
z
= 0
(2.8)We annowintrodu etheperiodi ityofthediele tri fun tion:
ǫ(~r) = ǫ
m
h
1 + C exp(−j ~
G · ~r)
i
(2.9) withG
~
are ipro allatti eve torsu hasG =
~
P
n
i=1
m
i
b
~
i
withn
beingthenumber of dimensionsofthe onsidered problem,m
i
anintegerandb
~
i
thebasisve torsoftheoremtodealwithlineardierentialequationswith periodi oe ientsin
one-dimension,andmu hlaterasimilar on eptwas developedbyFélixBlo h [59℄in
threedimensions(3D) tosolvetheproblemofthemotionofele tronsinaperiodi
potential. Therefore,solutionsoftheMaxwell'sequationsdes ribingthe
propaga-tionofele tro-magneti wavesin aperiodi diele tri mediumareoftenreferredto
as "Floquet-Blo h waves" (FB waves) or simply "Blo h waves". Mathemati ally
aFB wave, for examplefor TM polarization for whi h theele tri eld in the
~z
dire tionisE = E~z
~
, an beexpressedas,∞
X
n=−∞
V
n
(~k) · exp(−j ~
k
n
· ~r)
(2.10)with
~k
n
= ~k
0
+ ~
G
n
where~k
n
is thewaveve torofthen
-th omponent of theFB wave,~k
0
thewaveve torof therst harmoni ,G
~
n
are ipro al latti eve tor,andV
n
(~k)
theFourier oe ientfun tion.The dira tion properties of one- or two-dimensional gratings and the opti s
oftheFB wavessupportedin su h stru tureswereanalyzedin thelate80's [2, 3℄
using thepowerful framework of waveve tordiagrams. Su h diagrams havebeen
appliedto anisotropi rystalsand laterto photoni rystalstru tures. Theywill
be used in Chap.8 to explainthe physi sof the dispersionrelated ee ts seenin
2DPhC stru tures.
The eigenvaluesof the masters equations 2.7 and 2.8 givethe band stru ture
ω(~k)
of the PhC and is also alled the dispersion relation [60℄. It is ommon to onsider the band stru ture in the rst Brillouin zone either in the full zoneor along the main symmetry dire tions (
Γ
MK) only. Figure 2.3 displays an exampleshowingthebandstru tureforatriangular-latti e2DPhCindi atingthebandgap (sometimes alled stopgap) and the rst three bands are labeled. The
masterequations 2.7and 2.8 arewavelength s alable. Thepropertiesof thePhC
stru turesdonotdependontheabsolutewavelengthofthelightbutratheronthe
ratioofthewavelengthto thelatti eperiodi ity. Thus itismeaningfulto usethe
normalizedfrequen y
u = a/λ
, as in Fig.2.3 to des ribethe opti alproperties of PhCs.In the rst band lose to the
Γ
point the waveve tor of the light inside the stru tureismu h smallerthanthelatti e onstant: weareinthelongwavelengthlimit(i.e.,
λ
is mu h largerthan the PhC period). Thelight doesnot "feel" the periodi diele tri perturbationandthepropagationissimilartothatinanisotropimedium of refra tive index
n
ef f
, the ee tive refra tive index of the stru ture. Awayfromthislongwavelengthlimit,asλ
de reasesthewaveseestheperiodi ity: weentertheregimeofBraggree tion. Dependingonthestrengthofthediele tri"potential"(that anbe alled"photoni strength"),theenergyofthewavewillbe
moreor lessdistributed in between theFourier omponents of theex ited mode.
Fig.2.3 omparesthebandstru turesforTEandTMpolarizedlightforaPhC
stru ture of
f
= 30%,airholesinaba kgroundof refra tiveindexn = 3.24
. The bandstru tureforTEshowsthepresen eofabandgapwhereasthereisnobandgapforTM.Anotherdieren eisthat forTM theseparationbetweenthese ondand
thethird bandsislarger.
Figure2.3: Bandstru tureforTEmodesina2Dtriangularlatti ealongthemain
symmetrydire tionswithairllfa tor
f = 30%
;b)Bandstru tureforTMmodes. TheairlightlineisshownintheΓ
Mdire tion.Therelationshipbetweenthewavelengthoflightanditswaveve torinairorin
theba kgroundmaterialisgivenby:
u =
a.k
2πn
(2.11)where
n = n
mat
orn
air
istherefra tiveindexoftheba kgroundmaterialorofair (=1), respe tively. These linearrelations are alled thelightlines. OnFig.2.3b,thelightline orrespondingtoairisindi atedinthe
Γ
Mdire tiononly. Lightlines areoftenusefulto identifyguidedand leakymodes. Amoredetaileddis ussionisgivenin Se .2.6.
Fig.2.3alsosuggeststhatforagivenPhCthebandstru ture anbeinvestigated
byvaryingthewavelengthoflightand/orthelatti eperiod
a
. Inpra ti e,thelatter an beusefulwhentheavailablewavelengthrangeofthesour eislimited. Thisisalled lithographytuning, andisappliedin Chap.5.
Modelingoftheopti alpropertiesof2DPhCdevi eswiththeInP/GaInAsP/InP
verti alheterostru tureis usuallyperformedwith2Dmodelingtools. The
refra -tiveindexoftheba kgroundmaterialis giventhevalueoftherefra tiveindex for
theheterostru ture
n
ef f
andisobtainedfromtheee tiveindexapproximation. Deningthepolarizationofmodesinaplanarphotoni rystalisnotastraight-forwardissue. Firstofall,the onventionsusedinthe"waveguideapproa h"andin
the"2DPhC"aredierent. Fromthepointofviewoftheverti alwaveguideaeld
axisof theholes. Itis thereforeimportantto statewhi hdenitions areused. In
this thesis the 2D PhC denition is applied. There is an added ompli ation
areal 2D PhC isstri tly a 3Dobje t andthe Blo h oe ientsof the eld have
omponentsinthethreedire tionsandthestri tdenitionsforTEandTMmodes
given above do nothold anymore. An a epted way is to dene an approximate
polarizationstatein thePhC alled TE-likeandTM-like. Howeverthisdenition
holdsonlyifone andeneasymmetryplaneintheverti alstru ture(atthe
sym-metryplane,polarizationisstri tlyTEorTM).Inthelowindex ontrastsystem,
this is not valid and results in polarization mixing. We refer to the polarization
statesaspseudo-TEandpseudo-TM.Keepingthisinmind,intherestofthethesis
thepolarizationmodesinthePhCwill besimplyreferredtoas TEorTM.
3D al ulationtoolsarerequiredtosimulatePhCsbasedonhigh-index ontrast
stru tures. However, in the ase of low-index ontrast system, even though 2D
simulations give a reasonable agreement with experiments more a urate results
anbea hievedby3D al ulations. Dieren esinwavelengthshiftortransmission
levelsbetween2Dand3Dsimulationshavebeenreported[61℄.
2.5 Computational methods
Dierentmethodshavebeensu essfullyusedtosolvetheMaxwell'sequationsand
tomodeltheopti al hara teristi sofphotoni stru tures. Hereonlyalistofafew
(most ommon)methodsisgiven. Inasket hypi ture,themethods anbedivided
intwomain ategoriestreatingMaxwell'sequationseitherinrealspa easafun tion
oftime (where Finite-Dieren eTime-Domain (FDTD) isthe dominantmethod)
orin re ipro alspa e (frequen ydomain). Inthefrequen ydomain, methods are
distinguishedbythebasis usedfor theexpansionof theeld. Therstand most
straightforwardmethodisthePlaneWaveExpansionmethod(PWE)usingabasis
of plane waves. It is a parti ularly su essful method to obtain the PhC band
stru ture. To address simple defe ts in PhC stru tures it is possible to use the
super ell method. Howeverothermethodsare morepowerfultosimulate omplex
stru tures. For instan e, the MultipleMultipole Expansion method using abasis
ofspatialBessel/Hankelfun tions[62℄,orWannierfun tionsisespe iallyusefulin
the ase oflo alizedelds (i.e., avities) [63℄. S attering/transfer matrixmethod
isa widelyused omputational method for photoni s. This method and has also
been appliedto photoni rystals, sometimes together witha modalexpansion of
theeld. Modal methods an expand theeigenmodeson aBlo h mode basis(as
in the freely available software CamFR [64℄ or on a Fourier basis (e.g., Fourier
modalmethod[65℄). Inthisthesis,PWEandFDTDareused,andamoredetailed
dis ussionofthosetwomethodsisgivenbelow.
Plane Wave Expansion
stru -majorimplementationsare available: afull ve torial al ulationusedbythe MIT
program MPB [66℄, or a more omputationally ee tive method proposed by Ho
[67℄. Whenwe introdu e theperiodi ity ofthe latti ewe an expandthe eld in
termsofBlo hmodes
ϕ
~
k
(~r) = exp(j~k · ~r) ~
u
k
(~r)
withu
~
k
(~r)
ave torialfun tionwith the same periodi ity as the PhC latti e. Owing to this periodi ity, we an nowexpand
u
~
k
(~r)
into aFourierseries. Inorderto solvethemasterequations2.7and 2.8numeri allyoneneedstoexpandallthe oe ientsandtrun atetheseriestoanite number
N
ofplanewaves. Onepossibilityistoexpand1/ǫ(~r)
. However,for reasonsofimproved onvergen eHo'smethodrst al ulatestheFouriertransformof the diele tri map
ǫ(~r)
, trun ates it to the number of onsidered plane wavesN
, andthen inverts it. In thisthesis, the2DPWE resultswereobtainedusing a MatlabimplementationofHo'sPWEmethod [67℄.In thePWE method the simulatedPhC elds are innite in the2D plane of
periodi ityandtranslationinvariantalongtheverti aldire tion(i.e.,innitelylong
holes). Itispossibletosimulatestru tureswithsimpledefe tsinthe rystallatti e
(su has avities,waveguides)usingthesuper ellapproa h. Wedeneaunit ellfor
the rystal(mu hlargerthantheoriginalunit ellofthePhClatti e)in ludingthe
defe tthatwe all"super ell"andthenrepeatthisunit ellto overthefullspa e.
In order to avoid artifa ts arising from arti ial oupling between the repeated
defe ts the super ell should be su iently large. One understands immediately
that thismethodissuitabletoverysimplegeometriesonly.
Finite Dieren e Time Domain
In theFDTD method Maxwell'sequations aresolved in time domainusing nite
dieren e operatorsto approximate thedierentials. Itis apowerfulmethod,
al-beitoften des ribedas a"brute for e"method, and is ableto simulate anygiven
geometry anddoesnotrelyontheperiodi ityofthestru ture. Itisawidespread
omputational method to solve a large variety of ele tro-magneti problems. It
has been proposed in 1966 by Yee [68℄. A grid (Yee's grid) is used to dis retize
spa e. Atea hpointofthegridtheele tri omponentoftheeld attime t+1/2
is al ulatedfromthevalueofthemagneti eldattimet,thenthemagneti eld
at time
t
+1 is obtained from the ele tri eld value att
+1/2, and so on. The dis retizationintimedependsonthespatialgridsize. Forinstan e,in2Dthetimestepisexpressedas:
∆t = β
1
c
a
q
1
∆x
2
+
1
∆y
2
(2.12)where
∆x
,and∆y
arethegridstepsizes,a
isthelatti e onstant(usuallysetto1),c
thelightvelo ityandβ
a oe ient. Theeld anbeex itedbydierentsour es su haspointsour es,planewavesorwaveguidemodes. The omputationaldomainis surrounded by Perfe tly Mat hed Layers(PML) [69℄ that are absorbinglayers
towards zero over a boundary layer. FDTD al ulations are very demanding in
termsof omputationalpower,espe iallyforsimulationof3Dstru tures.
In this thesis, only 2D al ulationswere performed with an innitestru ture
inverti aldire tion(i.e.,innitelylong ylindri alholes)and anite geometryin
the2Dplane. TheFDTD method anprovidethespatialeld distributionatany
giventime stepfor mono hromati ex itation,or the spe tral energydistribution
(transmission, ree tion, ...) by al ulating the value of the Poynting ve tor at
agiven position in thestru ture ex ited by a pulse. It is possibleto indu e loss
in thestru ture by theintrodu tion ofan imaginarypartin the denition of the
refra tiveindexof oneor moreof theregions. For amoregeneralreferen eabout
the FDTD method see Ref.[70℄. In this work, we have used the freely available
FDTDprogramF2P[71℄.
2.6 Leaky modes
Figure 2.4: Comparison of the in-plane waveve tor with the out-of-plane sphere
denedbythewaveve torinair
ModessupportedbyaPhCslab ansuerfromlossesifsomeBlo hharmoni s
arelo atedinside thelight ones. For a2DPhC,modes anbeguided intheslab
orleakyintothe laddinglayers. Inthe aseofthelowindex ontrastsystem,leaky
modesareradiativemodesinairorin thesubstrate,whi hbydenitionintrodu e
energy losses. If the in-plane wave-ve tor
k
~
in
of a given Blo h harmoni in the PhCslabisshorterthan thenormofthe orrespondingwaveve torin air~k
0
,it is possibleforlightto oupleintoradiativemodesin air. Theverti al omponentofthe
p
th
Blo hmode omponentisexpressedas:
~
k
p
z
=
q
n
layer
~k
0
− ( ~
k
in
+ p ~
G)
(2.13)where
n
layer
isthe refra tiveindex of airor ofthe substrate. From this equation one an see that ifthere exists anintegerp
su h that~k
in
+ p ~
G < n
layer
~k
0
thewaveve tor
~k
p
z
will be real and this omponent of theBlo h modewill leak. The groupofve torsink
-spa e,normsofwhi hareequalton
layer
~k
0
denethelight one. In order to nd theleaky modes of the stru ture one should ompare theposition ofthe ex itedBlo h modesin
k
-spa ewithrespe tto thelight ones for airand forthesubstratematerial. OnFig.2.4, themodeof waveve tor~k
in−1
an ouple into radiativemodesin air, whereasthemodeof waveve tor~k
in−2
will be evanes entin air. Theinuen e of leaky modes onthe opti alpropertiesof PhCFabri ation steps for photoni
rystals
The fabri ation of semi ondu tor-based 2D PhCs involves several pro ess steps,
typi ally in lean room environment, and dierent pro ess equipments are used.
Ea h ofthepro essstepsshouldbepre isely ontrolledand alibratedin orderto
obtainstru turesofdesiredquality. Inthis haptertherelevantpro essstepsthat
were used for PhC devi e fabri ation are des ribed, followingtheir sequen e in a
typi alpro essow.
3.1 Epitaxy
TheInP/GaInAsP/InPverti alwaveguideisgrownbymetal-organi vapor phase
epitaxy(MOVPE)onanInPsubstrate. Thegrowthpre ursorsaretrimethylindium,
trimethylgallium,arsineAsH
3
andphosphinePH3
andthegrowthis ondu tedatatemperatureof 680
o
C. Thethi kness ofthe topInP ladding is200nm (unless
otherwisestated). The orelayerismadeofGa
x
In1−x
Asy
P1−y
withx = 0.24
andy = 0.52
latti emat hedtotheInPsubstrate. Therefra tiveindexofthismaterial atλ
=1.55µ
misn
=3.35anditsabsorption edgeatλ
edge
=1.22µ
m. Somesamples hada10-nmInGaAsPquantumwell(QW)for arrierlifetimemeasurements,withanemission wavelength of
λ
=1.14µ
m. Thethi knessof the orelayeris 420nm, unless otherwise spe ied. All the layersare undoped. For internal light sour emeasurements ( hap.5), twoQW with emission wavelengths around 1.47
µ
m and 1.55µ
mareembedded intotheInGaAsP orelayer.3.2 Mask deposition
The SiO
2
mask is deposited by Plasma Enhan ed Chemi al Vapor DepositionTable3.1: Pro essparametersforthePECVD pro essesforSiO
2
deposition Pro essC Pro essP Substrate temperature 230o
C 300o
C Plasmapower 15W 20W Pressure 700mT 800mT Depositionrate 1nm/s 1.2nm/s Refra tiveindex 1.47-1.49 1.51-1.52Gasesandows 5%SiH
4
in He: 2%SiH4
in N2
:40SCCM 740SCCM
N
2
O:200SCCM N2
O:425SCCMChemi alvapor deposition (CVD) is apro ess in whi h gaseous spe ies rea t
onasubstratetoformathinsolidlm[72℄. Inplasmaenhan edCVD mostofthe
energyrequiredto generate hemi allyrea tivespe iesisprovided by theplasma.
ThusPECVDallowsdepositionofthinlmsatrelativelylowpro esstemperatures.
Theplasma isgenerated at afrequen y of13.56 MHz. Inthis thesiswork we
used twotypesofdepositionpro essesas listedonTable3.1. The onformalityof
thedepositedlms anbetailoredtosomeextendbyadaptingenergyandangular
distributionoftheions. ThegasesweusedtoformSiO
2
thinlmsaresilane(SiH4
)and dinitrogen oxide ("laughing gas") N
2
O.Details aboutthe PECVD pro essesused inthisthesisarefoundinTable3.1.
3.3 Ele tron Beam Lithography
Thepatterndenitionofnanostru turesisa riti alstepfordevi egeneration.
Al-thoughalternativemethodshavebeenproposed(deep-UVlithography[73℄,
nanoim-printlithography[74℄,selfassembly[75℄),ele tronbeam(e-beam)lithographyisstill
by far the dominant method owingto its exibility. Inthe presentwork,two
e-beam systems were used: a Raith 150 e-beam system at the KTH Nanophysi s
departmentandaLEICAe-beamsystemattheLaboratoryforPhysi sand
Nano-stru tures(CNRS-LPN,Mar oussis,Fran e).
Ebeam resists
Thedoseistheamountofele tri al hargesre eivedbytheresistperunitarea. It
isusuallyexpressedin
µ
C/ m2
. The riti aldose orrespondstotheminimaldose
requiredbytheresistinorderfortheexposedregionstoberemoved ompletelyby
thedeveloper. Whenele tronspenetrateintoamaterial,theyaresubje tedto
for-wardandba kwards attering. Thesedependonthematerial(natureofsubstrate
fortheexposureoftheresistinareasawayfromthea tualbeamposition. This is
alledtheproximityee t(PE)denedaspatternspe i linewidthvariations[76℄.
The magnitude of this ee t depends on the a eleration voltage an operation
at100keVallowsadiminution ofthePEdue toalargerpenetrationdepth ofthe
ele tronsin the substratematerial. Inthis ase, thelargeangle s atteringo urs
furtherawayfrom thesubstrate/resistinterfa eand theprobabilitythat theyget
s attered/absorbedinthematerialbeforerea hing theresist ismu h higher. This
isnotthe aseforlowera elerationvoltages,at25keVforinstan e,forwhi h ase
thePE an beveryimportant. Workat verylowbeamenergies an beasolution
toredu es atteringbutitsuersfromamaindrawba kintermsoftheresistresist
thi kness.
Figure3.1: a)Chemi alstru ture ofPMMA;b) Chemi alstru tureofZEP.
Manye-beamresistsareavailable,dieringintermsofresolution,sensitivityor
resistan eto dry-et hing. Howeversome of them are predominantly used due to
theirhighsensitivityor theirresistan etodryet hing. Inthiswork,twotypesof
positiveresistswereused: PMMAandZEP 520-A.
Poly-methyl-metha rylate(PMMA)(Fig.3.1a)is onventionallyadoptedforits
highsensitivityallowingverynestru turestobepatterned. ZEP-520A,provided
by Nippon ZeonCo, hasthestru ture presentedin Fig.3.1b[77℄. It onsistsof a
virtual1:1 opolymerof
α−
hlorometha rylateandα
-methylstyrenewhi hexhibits apositiveresistbehavioruponele tronbeamexposure. Thehigh-sensitivityisat-tributedtothe
α
- hlorometha rylategroups,whereastheresistan etodry-et hing isdueto theα
-methylstyrenegroups. Theglasstransition temperatureof ZEP is 145o
C.We hose hereto hardbaketheresist at 180
o
C and p-xylenewas used as
thedeveloper.
The Raith 150 E-beam lithography system
Ane-beamlithographysystemis a omputer- ontrolledS anningEle tron
Mi ro-s ope(SEM)providedwithabeamblankerandapatterngeneratorunit. Insome
asesan interferometri stage forexposureof patternsoveralargesample areais
also available [76℄. A s hemati drawingof the Raith 150e-beamsystem is
pro-videdon Fig.3.2. TheSEM hamberismaintained under highva uum(range of
Figure3.2: S hemati sfortheRaith150system. Reprodu edwithpermissionfrom
The SEM olumn is aGemini olumn, spe ially builtto provide high
perfor-man eatlowa eleratingvoltages. The olumnisequippedwithaS hottky
eld-emissionele tronsour eofthehot- athodetype(atungstentip withazir onium
oxide ollar). Theva uum in theele trongun hamber("gun va uum")is in the
10
−9
mBarrange. Thepatterngeneratorisaunit ontrollingthebeamblankerand
thedee tionofthebeambysendingvoltagesignalstothes an oils. The
displa e-mentofthesamplewithrespe ttothe olumnisadjustedbyalaser-interferometer
ontrolledstage,whi hintheRaithsystemweusedhasaresolutionof5nm. The
movementofthestage allowsstit hing ofwritingeldswhen exposinglargeareas
(i.e.,largerthanwhatisa hievablebydee tion ofthee-beam). IntheRaith150
system, the writingeld length is limited to the range 60
µ
m to 1400µ
m. How-ever,atrade-oshould bemadedependingontherequirementsinthea ura yofpatterning. Alargewritingeldareaallowstoexposeextremelystit hingsensitive
patterns (su h as PhC areas) without moving the stage. However, large writing
eld areas should be avoided if distortion sensitive patterns are present far from
the entral area. In the ase of stit hing, smaller writingelds will also provide
betterresults.
Thewritingspeeddepends onseveral parameterssu h astheavailable e-beam
urrent (dose), the maximum frequen y of operation of the DAC ontrolling the
beam dee tion (10MHz, step displa ement) and the settling time whi h is the
timethee-beamneedsea h timeit movestoanewposition onthesample(5 ms
in the Raith systemwe use). The pattern designneeds to be divided into small
re tangularandtriangularunitsthatthesystemwillexposefollowingparallellines.
Thisisaproblemfor ir ularfeatures;thustheholesareapproximatedbypolygons.
Typi alholediameters forPhC operatingat near infrared(
λ
around1.5µ
m) are intherangeof100toafew100nm.TheparametersoftheexposuresperformedwiththeRaith150systematKTH
are ana eleration voltage of 25keV, apertures of 7.5
µ
m for PhC holes (e-beam urrent≈
10-20pA)and of20to60µ
mforlargeareas(ebeam urrent≈
3nA),a workingdistan e ofapproximately5mm,awritingeld sizeof100µ
mandastep sizeofafewnm.Beforeexposureoneshould orre tforfo us,astigmatismandbeamalignment,
as in any onventional SEM system. In the ase of exposure over large areas, a
writingeld alignment(WFA) pro edure has to be performed. It is a"learning"
pro edureaimedat alibratingtheorthogonalityands alingfa torofthedee tion
axis to the high pre ision sample stage. The stage is assumed orre t and the
dee tion systemis adjusted to it. One hooses asmall feature that is pla edat
the enterofthegivenwritingeld(WF) orrespondingtothe enterofthe olumn
forzerodee tion. Thenthesystemmovesthestageinordertopositionthefeature
atoneoftheWF orners. Thesystemthendee tstheele tronbeamtos anover
anareaattheWF orner(seeFig.3.3),thegeneratedimageisusedtoindi ateto
thesystemthea tualpositionofthefeaturewithinthes anarea. Thisisrepeated
at regularintervalsthat thefeature isstillatthe enterofthewritingeld. Ifthe
dee ted beamdoesnotndtheexa tpositionitwillresultsinstit hingerrors.
3.4 Mask opening - Rea tive Ion Et hing
Rea tiveionet hing(RIE)isawidespreadte hniqueusedinmi roele troni s.
Gen-erally a apa itively oupledradio-frequen y(RF) plasmais omposedofasour e
operating at 13.56MHz and two planar ele trodes in a va uum hamber. The
system is self-biased, whi h means that the ion energy annot be independently
ontrolled. It depends on the RFpower,the operatingpressure and gas
ompo-sition. The et hing of the SiO
2
mask is performed by Fluorine based RIE. Therea tivegasis usuallyCHF
3
. CF4
analso beused, howeverin this asetheet hsele tivityoftheSiO
2
overtheresistisrather poor. Onesolutionisto addH2
totheCF
4
plasma. Thepresen eofhydrogenin reasesthepolymerizationpro esses(C-Hbonds)thusin reasingthesele tivity. Theparametersofthepro essusedin
this thesis are apowerof 45 W, abias
V
bias
of -300V,and apro ess pressureof 15mT.Thegas ompositionwasH2
(10SCCM)/CF4
(28SCCM).3.5 Et hing of the semi ondu tor material
Thepatternsdenedbye-beamlithographyandtransferedtothehardmaskneed
to be now transfered to the semi ondu tor material by highly anisotropi et h
pro esses. Dry-et hing methods are ideally suited for this purpose. All the PhC
stru turespresentedinthisthesiswereet hedusingAr/Cl
2
CAIBE.Theimplemen-tationofthisparti ularet hingte hniqueaswellasthestudyofits hara teristi s
and impli ations on the properties of the fabri ated PhC represent a signi ant
partin thisthesiswork. Adetailed presentationof CAIBEandits hara teristi s
ispresentedin thenext hapter.
Mostofthealternativete hniquesforPhCfabri ationarealsodry-et hingbased
andarelistedonTable3.2. Anequipment-spe i omparisonofthemaindryet h
te hniquesusedinthe ontextofPhCet hingispresentedinthenext hapter. Wet
hemi alet hing is ex luded in most of the ases due the isotropy of theet hing
and theimpossibilitytoobtainhighaspe tratioforthePhCholes. Howeverhigh
aspe tratios anbeobtainedbyanele tro- hemistrymethodofporeet hing[97℄.
Table3.2 lists outmostofthe reportedInP et hing te hniques forPhC
fabri- ationand omparestheirreportedperforman es. Thevaluesfortheaspe tratios
giveninTable3.2refertothetotaldepthoftheholesdowntothetaperedbottom.
Depending on the hole shape the aspe t ratio of the straight portion of the hole
will besmallerthanthat givenin thetable. Typi allytheapproximatelystraight
portion orrespondsto about70%of thetotal depth. However, thisestimate has
to betaken autiously. RIE ismorelimitedin a hievableaspe tratiosthanother
pro esseswithmoredenserplasmassu hasICP.HighestreportedARareprovided
manu-Table3.2: Comparisonofet hing te hniques forInP-basedPhCfabri ation
Te hnique Chemistry Performan e Referen e
RIE CH
4
AR≈
2 [79℄CH
4
/Ar/H2
AR≈
2 [80, 81℄ECR-RIE Cl
2
/Ar AR=8 [82℄ylindro- oni al
Cl
2
AR≈
2 [83℄ICP-RIE SiCl
4
(/Ar) AR=14 [84℄, [85℄ylindro- oni al Cl
2
AR≈
10 [86℄ Cl2
/O2
AR=16 [87℄ 250o
C ylindro- oni al ylindri al Cl2
/N2
AR≈
8 [88℄Cl
2
/Ar/N2
(/He) AR=16 [89℄oni al Cl
2
/Xe AR≈
5 [90℄ ones swelled ylinders Hi/Xe AR=13 [91℄, [92℄ oni al CAIBE Ar/Cl2
AR≈
18-20 [81, 93,94℄ ylindro- oni al FIB lowAR [95, 96℄fa turinga tivedevi es. It is reasonable to expe t that high bombarding energy
will reatemore damage. However thenature of the reateddefe ts will also
de-pend onthe typeof dry-et h hemistry. Themostmaterial damagingfabri ation
te hniqueis Fo usedIon Beamet hing, whi h bombardsthe samplewithgallium
ionsa eleratedtoahighenergy(typi ally15to30keVforsputtering,oftheorder
of100keVforimplantationregimes).
3.6 Post-et hing pro ess steps
AfterthePhCiset hed,dependingonthedevi esandmeasurement onguration,
additionalpro essstepsareusuallyne essary. Ifoneneedstoa essthePhC
stru -turesele tri ally( aseoflateral ondu tion(Chap.7)),oneshoulddeneele tri al
onta ts. Weuse onventionalopti allithographytopatternareasforthe onta ts,
thenweevaporateametallayer(Ni/AuGe/Ni/Aufor
n
- onta ttoInP)andusea lift-ote hniqueto isolate the onta t pads. An annealing step in neutralatmo-sphere(N
2
) around450o
themetallayerandthesemi ondu torbyalloying. Ifthephotoni rystaldevi es
aretobemeasuredbyanin-plane ouplingte hnique,su hastheend-remethod
des ribedinSe .8.3,otherpro essstepsareusuallyne essary. Theopti al oupling
viathe leavedfa etofthesamplerequiresgoodopti alquality. Inordertoa hieve
this it ispreferableto thin down thesubstrate toathi knessofabout100
µ
m by a lapping te hnique. Then the samples are be leaved at a suitable length andChemi ally Assisted Ion Beam
Et hing of InP-based photoni
rystals
Severalte hniques havebeenused foret hing PhCsin InP-basedmaterialsas
de-s ribedinthepre eding hapter. Argon- hlorine hemi allyassistedionbeam
et h-ing(Ar/Cl
2
CAIBE) isone ofthe mostsu essful pro essesfor high aspe t ratioet hing. Theawareness and ontrol of fabri ation and understanding of the
ma-terialissuesforphotoni rystalsarene essaryforhighperforman ePhCdevi es.
The hara terization of theet hing is thereforene essary. The quality of et hing
ofphotoni rystalsisvery riti alasitdire tly inuen estheiropti alproperties
in termsof losses. Inaddition, et hing hara teristi ssu h as et h depth,
rough-nessand shape invariablydepend onthe feature size. This hapter presents the
results obtained in Paper A. We introdu e the CAIBE te hnique, present the
experimental resultson thefeature size dependen e of theet hing and developa
physi o- hemi al model forthe et hing me hanism. The phenomenon behind the
featuresizedependen e oftheet hingisexplainedanditsimpli ationsintermsof
theopti alpropertiesofPhCsisdis ussed.
4.1 Chemi ally Assisted Ion Beam Et hing
CAIBE is an et hing te hnique relying on the bombardment of a sample with
a hemi ally inert ion beam under a rea tive gas atmosphere in a high-va uum
hamber. Intheexperimentsreportedhere,theneutralgasisargonandtherea tive
spe iesCl
2
. Chlorineisinje tedviaagasringoverthesampleandargonionsareextra tedfromaremoteplasma. Itallowsindependentvariationsoftheionenergy
anddensity. CAIBEwasperformedwithaNordiko3000ionbeamet hingsystem,
equippedwithatwo-gridiongun(Fig.4.1). Inthissystem,theplasmaisgenerated