KTH engineering Sciences

Contact poling of RKTP with silicon needles

Hoda Kianirad

Master of Science Thesis

Laser Physics

Department of Applied Physics

School of Engineering Science

Royal Institute of Technology

Stockholm, Sweden 2013

TRITA/FYS 2013:06

ISSN 0280-316X

ISRN KTH/FYS/--13:06-SE

پدر و مادرم به

i

(a)

Abstract

Quasi-phase-matching (QPM) is a method to get efficient and tailored second-order

nonlinear interactions. Several techniques exist for fabrication of periodic domain

structures in ferroelectric crystals for QPM- based frequency conversion. By far, electric

field-poling using lithographically patterned electrodes on the z-face of the crystal is the

most common one. High-quality, periodically inverted ferroelectric domain structures in

flux-grown KTiOP (KTP) crystals were fabricated already in the late 90’s using this

technique. It has been shown that a slight Rb doping of the KTP crystal (RKTP) facilitates

the periodic poling. However, fabrication of two-dimensional (2D) domain structures in

RKTP has not yet been investigated. A disadvantage with the lithographic patterning is

that each sample needs to be patterned individually, which is tedious and time

consuming. Moreover, when it comes to small domain features, which are required by

the next generation of nonlinear optical devices, a more versatile poling technique has to

be developed due to the limitations of conventional photolithography.

In this work, we present a new technique for 2D domain inversion in a 1 mm thick

RKTP crystal and demonstrate the densest 2D lattice in a KTP isomorph. First, 2D

periodic arrays of silicon spikes with 20 µm periods and silicon pillars with 5 µm

periods were constructed using isotropic dry etching. Second, the silicon arrays were

used as a contact electrode in order to periodically pole the RKTP crystal. A 2D domain

pattern with 20 20 and 5 5 periods were obtained. A high normalized

conversion efficiency of 1.27 % was obtained for frequency doubling of a CW

Ti-Sapphire laser at 894 nm for the 5 µm period pattern. The measured temperature

bandwidth was 4.25 from which we estimated an effective crystal length of 5.4 mm,

which is very close to the physical structure length of 6 mm.

This novel poling technique has several advantages. First, the silicon electrode is

reusable and there is no need for patterning each sample individually. Second, the

crystalline structure of silicon provides high accuracy and reproducibility in the

electrode fabrication. Finally, Si array electrodes can be designed for any desirable

period or electrode geometry. Therefore, contact electrode poling with this technique

can be a more convenient, flexible, and suitable method for making small feature domain

patterns, and it also can reduce the fabrication costs.

ii

Acknowledgment

This work could have never been done alone and without technical and emotional

support of several people and I express my gratitude to all of them.

First, Andrius Zukauskas, I would like to thank you for your supervision. Thank you for

teaching me all the small theoretical details, how to work in the lab and always having

time to answer my questions regarding the experiments and other things. On top of that,

thank you for helping me to present my work and guiding me in writing the report.

I would like to gratefully acknowledge the supervision of Thomas Frisk in the fabrication

part of my project. Thank you for providing me with knowledge, teaching me your skills

and being abundantly helpful. You have given me the chance to work independently and

to believe in myself by happily guiding me through this path of life with all the small

hints about everything.

I am deeply grateful to Prof. Fredrik Laurell and Carlota Canalias, not only for letting me

work on the fascinating and interesting subject of nonlinear optics but also for

generously giving me an unlimited access to every instrument and facility that I needed,

trusting me in all different parts of my project and willing to discuss my problems at any

time. Most importantly, thank you for letting me be a part of Laser physics family, you

inspired me to find my passion for research.

I would like to express my thanks of gratitude to Niclas Roxhed for the ICP machine

recipe, Jens A. Tellefsen, Jr for proofreading this thesis and for the great guides about

writing, Charlotte Liljestrand, for introducing the laser physics group to me, Katia Gallo

for all the late night laughs and good portions of food and Arezoo Ghanadian for being a

wonderful office mate with all the laughters that made late night work much more fun.

I wish to thank all the members of Laser physics research group for happily sharing

their knowledge, giving me hope and help when my mind got stuck, for all the laughters

during lunch times and for making me feel so welcome in the group.

I am grateful to all my friends, most importantly to Mandana and Saeed for being the

surrogate family during these years of living in Sweden and for their continued love and

support thereafter.

Of course, I would like to thank my family for their support through these years of my

education. Thank you for believing in me, for your unconditional love that made the

physical distance unnoticeable. A special thanks to my parents for their full support in

many different ways.

iii

Contents

1 Introduction 1

1.1 Background .................................................................................................................................... 1

1.2 Objective of the work .................................................................................................................... 2

2 Nonlinear optics 3

2.1 Linear and nonlinear polarization ................................................................................................. 3

2.2 Second order nonlinear processes ................................................................................................ 4

2.3 Susceptibility coefficient ............................................................................................................... 4

2.4 The coupled wave equation .......................................................................................................... 5

2.5 Phase-matching ............................................................................................................................. 6

2.5.1 Birefringent phase-matching and quasi-phase matching ...................................................... 7

2.5.2 Quasi-phase matched second-harmonic generator ............................................................... 8

2.6 Fabrication methods of QPM structures ....................................................................................... 8

2.7 1D and 2D quasi-phase matched crystals ..................................................................................... 9

3 KTP and RKTP crystal properties 10

3.1 Crystal structure ..........................................................................................................................10

3.2 Crystal properties ........................................................................................................................11

3.3 Domain switching ........................................................................................................................11

4 Silicon electrode fabrication 13

4.1 Method ........................................................................................................................................13

4.2 Photolithographic mask design ...................................................................................................15

4.3 Fabrication process .....................................................................................................................16

4.3.1 RCA cleaning .........................................................................................................................17

4.3.2 Wet oxidation .......................................................................................................................17

4.3.3 Resist coating ........................................................................................................................17

4.3.4 UV exposure .........................................................................................................................17

4.3.5 Developing ............................................................................................................................18

4.3.6 Hard bake .............................................................................................................................18

4.3.7 Silicon oxide etching .............................................................................................................18

4.3.8 Silicon etch ...........................................................................................................................19

4.3.9 Masks strip ...........................................................................................................................19

4.4 Modifications ...............................................................................................................................19

4.5 Final fabricated Si masks .............................................................................................................20

5 Fabrication of periodically poled RKTP by silicon stamp 25

iv

5.1 Sample preparation .....................................................................................................................25

5.2 Poling setup .................................................................................................................................25

5.3 Fabrication of PPRKTP .................................................................................................................25

5.4 Poling monitoring ........................................................................................................................26

6 Characterization of domain structures 28

6.1 Selective etching ..........................................................................................................................28

6.1.1 Domain evaluation for the 20 µm period grating ................................................................28

6.1.2 Domain evaluation for 5 µm period grating .........................................................................31

6.2 Optical characterization ..............................................................................................................33

6.2.1 Conversion efficiency ...........................................................................................................33

6.2.2 Temperature acceptance bandwidth ...................................................................................35

6.2.3 Domain merging point estimation .......................................................................................36

7 Conclusions 38

7.1 Further developments .................................................................................................................39

8 References 40

 

1  

1Introduction

1.1Background

Nonlinearfrequencyconversionisaveryattractivemethodforachievingcoherentradi‐ation in spectral regions inaccessible by available lasers, leading to potential applica‐tions ranging from optical communication, spectroscopy and remote sensing, projec‐tors,materialprocessingtodiagnosticandmedicaltreatment.Manyapplicationsoffre‐quencyconversionarebasedonsecond‐ordernonlineareffectsutilizingthe com‐ponent of the susceptibility tensor, which is a characteristic of non‐centrosymmetricmaterials,suchaslithiumniobate(LN),potassiumtitanylphosphate(KTP),lithiumtan‐talate (LT), etc. These nonlinearmaterials have relatively high nonlinear coefficientsandareusefulforapplicationssuchassecondharmonicgeneration(SHG),opticalpar‐ametricoscillation(OPO),andopticalparametricamplification(OPA).

Duetodispersioninthematerial,incidentandgeneratedwaveswithdifferentwave‐lengthstravelatdifferentvelocitiesandbecomeoutofphase,thusreducingtheconver‐sionefficiency.Forefficientfrequencyconversion,itisnecessarytomaintainthephasematchbetweentheinteractingwavelengths,i.e.,totransfertheenergytothegeneratedwave, the relativephaseof the interactingwavesmustbe kept constant.Quasi‐phasematching(QPM)bringstheinteractingwavesintophasebyaddinganartificialmomen‐tum vector. The artificial momentum vector in QPM originates from a periodic QPMstructure.Byperiodicallyinvertingthepolarizationvector,thesignoftheχ(2)suscepti‐bility can be altered and the accumulated phase mismatch between the interactingwavesisreset.ChangingthepolarizationlocallyinferroelectricmaterialtoachieveQPMis called ferroelectricdomainengineeringorperiodicpoling.With this technique it ispossible to phase‐match any conversion in the transparency range of the nonlinearcrystal,usuallyfrommid‐infraredtonearultra‐violet.

Therearemanydifferent techniques forachievingpolarizationswitching in ferroe‐lectricmaterials,includingchemicaltreatment,modulationofthesignofthenonlineari‐tyduringthe ferroelectriccrystalgrowth,electronbeamlithographyandelectric fieldpoling.Thelatteristhemostcommontechnique,whichisimplementedbyapplyinganelectricfieldperiodicallyovereverycoherent‐lengthoftheinteractionintheferroelec‐tricmaterial. Conventional electric‐fieldpoling relies on the creationof periodic elec‐trodesviacontactphotolithography,limitingthedomainsizeto1µm.Thelithographyprocessforpatterningmicro‐structuredelectrodesoneachcrystalisacomplicatedandtime consuming procedure. Thewidth of the domain‐inverted regionswith this tech‐niqueisalwayslargerthanthatoftheinitiallydepositedelectroderegion.Thiseffectiscalled domain broadening and it influences the accuracy of the duty cycle. Domainbroadeningisduetothetangentialcomponentoftheelectricfieldwhichresultsinpo‐larizationreversalunderthephotoresistisolatinglayer[1].Inordertoimprovethein‐

 

2  

verteddomaingratingqualityandcompensateforthedomainbroadening,ithasbeenproposedtopatternthemetalelectrodeswithlessdutycycle[2].Hence,inthecaseofalargewavevectormismatch theQPMstructureneedstobe in thescaleofsub‐micron.Thiscorrespondstosub‐micronelectrodedimensionsforsmalldomainpatterngratingwhichishardtoachievebythephotolithographyprocess.Latesteffortstoreducethedomainsizetosub‐micronscalewithlessdomainbroad‐

eninginclude1)periodicpolingofKTPcrystalwith720nmperiodforbackwardsecondharmonicgeneration[3],2)polingofRb‐dopedKTP(RKTP)crystalwith690nmperiod[4],and3)fabricationofperiodicallypoledKTP(PPKTP)crystalswith800nmperiodformirrorlessopticalparametricoscillation[5],[6].Inordertosimplifythepolingpro‐cessandtospeedupthefabricationprocedure,etchedSistamperelectrodewith3µmperiodhasbeeninvestigatedasanalternativeelectrodeforformationof1DdomainsbycontactpolinginLiNbO [7].However,theseelectrodeswerefabricatedbywetetchingandthesamplethicknesswaslimitedto~200µm.

1.2Objectiveofthework

Theaimof this thesis is to investigateanew technique for ferroelectricdomainengi‐neering on themicron scalewhich can reduce the domain broadening aswell as thecomplexityandlithographyproblemsoftheconventionaltechnique.Inthisapproachanarray of silicon needles is used as a periodic electrode for applying voltage over thecrystal in a periodic fashion. To explore the feasibility of this technique, two‐dimensionalgratingswithdifferentperiodicityarefabricatedassiliconneedlesonap‐typesiliconwafer.Thissiliconarrayofneedles isusedtocreate two‐dimensionaldo‐mainstructures inRKTPcrystal.Reproducingthegratingstructureofthesiliconelec‐trodewithhighaccuracyonacrystalcanleadtonewachievementsinperiodicpolingtechniqueandnewapplicationsofquasi‐phasematcheddevices.Finally,theferroelec‐tricdomainstructures,poledwithsiliconelectrodes,wereanalyzedandtheopticalper‐formanceofperiodicallypoledcrystalswasevaluated.

 

3  

2NonlinearopticsIn classicaloptics the light‐matter interaction inducesoscillatingdipoleswith the fre‐quencyoftheincominglight.Nonlinearoptics isthefieldofopticsthatstudiesthein‐teractionof lightwithmatter in theregimewhere theresponseof thematerial to theelectromagnetic radiation is nonlinear and depends on the electric field amplitude ofthe radiation. If thepowerof the incoming light ishighenough, thepropertiesof thematerialbecomedependentontheintensityoftheillumination.Theinducedoscillatingdipoles createphotonswith frequencies thatarenotpresent in the incoming light. Inthefirstnonlinear‐opticalexperimentofthelasererawhichwasperformedbyFrankenetalin1961,arubylaserradiationwithawavelengthof694.2nmwasusedtogeneratethesecondharmonicinaquartzcrystalatawavelengthof347.1nm[8].Thisworkwasfollowedbythediscoveryofarichdiversityofnonlinearopticaleffects[9].Therefore,inordertodescribefrequencyconversion,itisnecessarytounderstandtheprinciplesofnonlinearoptics.InthisChapter,thephysicsbehindthenonlinearfrequencyconver‐sionwillbediscussed.

2.1Linearandnonlinearpolarization

When an electromagneticwave passes through a dielectricmedium, the electric fieldinducesapolarizationinthematerial.Forlowintensityelectromagneticwaves,thema‐terialresponsecanbeapproximatedinthefollowingway:

P t ϵ χ E t ,(2.1)

whereϵ isthepermittivityofthevacuumand istheelectriccomponentoftheelec‐tromagneticwave.ThisequationshowsalineardependenceofthepolarizationPontheelectricfieldwithaproportionalityfactorχ (firstordersusceptibility),whichappliestothelinearopticalphenomena.Whenitcomestomoreintenseelectromagneticwaves,Eq.(2.1)canbeextendedto:

P t ϵ χ E t χ E t χ E t ⋯ ,(2.2)

or

P t P P P ⋯.(2.3)

Theparametersχ andχ inEq.(2.2)arethesecondandthirdordersusceptibilities,respectively.Regardingthelinearandthenonlinearcomponents,Eq.(2.3)canbewrit‐tenas:

P t P P .(2.4)

Ifthenonlinearpartoftheinducedpolarizationisstrongenough,theoscillatingdipolesinsidethematerialemitphotonswithotherfrequencies[10].

 

4  

2.2Secondordernonlinearprocesses

The term ϵ χ E t in Eq.(2.2) describes the second order nonlinear interactions,whichareillustratedinFig.(2.1).

 

 

 

 

 

 

                                                  Figure 2.1 Second order nonlinear interaction processes. 

Figure(2.1.a)showsthesecond‐harmonicgenerationprocess,wheretwoincidentpho‐tons,eachwithfrequencyω,arecombinedtogiveonenewphotonwithfrequency2ω.Theinducedpolarizationinthenonlinearmediumisthenasfollows:

P 2ϵ χ EE∗ ϵ χ E e c. c. .(2.5)

ThefirsttermofEq.(2.5)representsthezerofrequencycontribution,knownasopticalrectification. The second term represents the generated second‐harmonic signalwithfrequency2ω.Second‐harmonicgeneration is inpracticeuseful toaccess thespectralrangeofshorterwavelengths.

Figure (2.1.b) shows the case of sum‐frequency generation (SFG) and difference‐frequencygeneration(DFG).InSFG,twoincidentphotonswithfrequencies and propagatingthroughthenonlinearmediumarecombinedtocreataphotonwithhigherenergy, . In the DFG case, a lower energy photon is createdwith a fre‐quencyof (observethathere ).

Figure(2.1.c)showsanotherpossiblesecondordernonlinearinteractioncalledfre‐quency down‐conversion or optical parametric generation (OPG). Here, one photonwithfrequencyω propagatingthroughthenonlinearmediumissplitintotwophotonswithfrequenciesω andω repectively.Ifthenonlinearmediumisplacedinanopticalresonator,theconfigurationiscalledopticalparametricoscillator(OPO)[11].

2.3Susceptibilitycoefficient

Thenonlinearsusceptibilityisintroducedasatensorofrankm+1forχ ,but,usually,anonlinearcoefficient,d,isusedinsteadofχtensorandisdefinedinthefollowingway:

 

 

 

2 SHG

SFG, DFG

OPG, OPO

2 2

a

b

c

 

5  

d χ .(2.6)

If the condition of permutation symmetry applies,d d , the d‐tensor can be ex‐

pressedasa3 6‐elementmatrix,andthepolarizationofthegeneratedphotonsinCar‐tesiancoordinatesisgivenbythecomponents:

P

P

P

2ϵ Kd d d d d dd d d d d dd d d d d d

E EE EE E

E E E EE E E EE E E E

 ,             (2.7)

whereω is the frequencyof thegeneratedphotondue to theannihilationof thetwophotons with frequencies ω and ω . K is the degeneracy factor, given by K=½ ,whenω ω (SHGcase)andK=1whenω ω [12].

Consideringa second‐ordernonlinear interaction, the susceptibility coefficient χ onlyappearsincrystalswithoutacenterofinversion.Inmanymaterials,therefore,χ vanishesduetotheircentrosymetriccrystalstructure.

2.4Thecoupledwaveequation

Thenewfrequencycomponentsoftheelectromagneticfieldcanbeexpressedbyintro‐ducingawaveequation. If thenonlinearpart isaddedtothestandardwaveequation,onecanextractthecoupledwaveequationbetweentheinteractingwavesbythefollow‐ingexpression:

E μ σ μ ϵ μ ,(2.8)

whereμ isthevacuumpermeabilityandσistheconductivityofthecrystal.

Assumingawavepropagatinginthex‐direction.Inthesecondharmonicgenerationprocess,ittakestheformofplanewaves,E x, t ½ E x,ω e c. c,andwith

theapproximationofa slowlyvaryingelectric field, i.e. ≪ k , the threecoupled

equationsateachfrequency,canbederivedasfollowing:

α E Kd E E∗e ∆  , 

α E Kd E E∗e ∆ ,(2.9)

α E Kd E E∗e ∆  . 

 

 

Here,d

ofthei

tioncobeingesity,the

2.5Ph

CompleConsiddamentisnotfure (2.matchi

tion of

fromth

calledc

differen

Afterobacktolatebacthecoh

 

Fig

m

Severalinteracthecon

d isthee

ncidentfie

efficientanequalto∆ketotalphas

hase‐matc

etephase‐meringtheStalandthefulfilled,i.e.2) illustrangconditio

f the crysta

hecrystale

coherencel

nceof isa

necoherenothefundackandfortherencelen

gure 2.2 Intens

atching and c) 

ltechniquectingwavesnceptofqua

ffectiveno

eldandcan

nd∆kisthk k

semismatc

ching

matchingisSHprocess,eSHwavese.∆k 0,aates the intons.Forac

al lengthw

edgetothe

lengthand

accumulate

ncelength,amentalfielthbetweenngth[14].

sity of SHG  in d

Quasi phase m

eshavebees.Twodiffeasi‐phasem

nlinearten

bederived

etotalwavk .For

chshouldb

sreachedwduetothediffer,nandthenontensity ofcertainmis

with a perio

epoint tha

isequalto

edbetween

theintensld(nophasntheharmo

different phase 

atching [14].  

endeveloperentmethmatching,w

6 

nsor,which

dfromEq.(

vevectormefficientinbezero[13

whenthephemateriald

n ,sotnlinearfrethe seconsmatchof∆

od of L∆

at theSHG

oL .O

nthetwoin

sityofthessematchinonicandth

matching cond

pedtominihodswillbwhichhasb

hdependso

(2.7).Thet

mismatchonteractiona].

hasemismdispersion,themomenquencycond‐harmoni∆k,theSH

∆ along th

intensitya

verthisch

nteractingw

secondharng).Inthishefundame

ditions. a) Perfe

imizethepeshortlyrbeenemplo

, L

onthepola

ermα

oftheprocandthehig

atchisequtherefractntumconsenversionisc signal foHGintensity

he crystal a

attains itsm

aracteristic

waves,i.e.,

monicfieldcase,theinentalwaves

ect phase matc

phasemismeviewedwoyedinthis

arizationd

isthe

essandisghestoutpu

ualtozero,tiveindiceservationcosnotefficieor differenyvariesas

axis. The d

maximumv

cdistance,

L∆.

dstartstontensitywisperiodica

ching, b) No ph

matchbetwwithmainfsthesis.

irection

absorp‐

definedutinten‐

∆k 0.soffun‐onditionent.Fig‐t phaseafunc‐

distance

value is

aphase

convertilloscil‐allywith

hase 

weenthefocuson

 

7  

2.5.1Birefringentphase‐matchingandquasi‐phasematching

Thereareseveraltechniquesforcompensatingthephasemismatchinthecrystal.Onetechnique takesadvantageof thenaturalbirefringenceof themediumitself. Thebire‐fringenceisdefinedasthedifferencebetweentherefractiveindicesoftheordinaryandthe extraordinary wave defined on the principal axis of the crystal. In birefringentphase‐matching, by choosing a specific “cut” or crystal orientation, the interactingwavesexperiencethesamerefractiveindices.Consequently,theypropagateatthesamephasevelocityand∆kisthuszero.Growth,preparationandalignmentofthesecrystalsis straightforward. However, the applicability of simple birefringent phase‐matchingcan be limited by a number of factors. First, since themethod relies on thematerialproperties(thedispersionandthebirefringence),findingacrystalorientationthatsat‐isfiesthephase‐matchingconditionischallenging.Second,thenonlinearitydependsonthepropagationdirectioninthecrystal.

Anotherphasematchingtechnique,whichisbasedonthecompensationofphaseve‐locity differences between the interacting waves, is quasi‐phase matching (QPM). InQPM, thesignof thenonlinearsusceptibility ismodulated in thespatial coordinate topreventaccumulationofphasemismatch.Suchaspatialmodulationcanbeobtainedinferroelectriccrystalsbyperiodicallyalteringthecrystalorientationandinthiswaytheeffective nonlinearity changes between d and d . The interacting waves stillpropagatewithdifferentphasevelocities,but,whentheaccumulatedphasemismatchreachesπ, the sign of the drivingnonlinear susceptibility is also reversed so that thephasedifference isreset tozero.Thiscreatesastep‐wisegrowthintheoutputpoweralongthecrystallengthascanbeseeninFigure2.2.c.Ideally,themodulationisdoneaf‐tereachcoherence length,L , and is referredtoas the first‐orderQPM.OnecanalsousehigherorderQPMwherethematerialismodulatedwithaperiodofseveralcoher‐encelengths.Besidesahigherefficiency,QPMstructurescanbemadewithanyperiodandstructureaswell asprovidephasematching foranynonlinearprocess inside thetransparencywindowofthematerial[15].

InaQPMstructure,theperiodicallymodulatednonlinearcoefficientisusedtocom‐pensateforthephasemismatchinthecrystal.Thisnonlinearcoefficient, ,canbewrit‐teninthespatialcoordinatebyaFourierexpansion:

d x d ∑ G exp ik x ,(2.10)

whered isthenonlinearcoefficient,k isthem ordergratingvectorwhichsatisfies

thephase‐matchingcondition,andG istheFouriercoefficientofthem harmonicde‐finedas[16]:

G sin mπD .(2.11)

Here,Disthedutycycle,whichisdeterminedbytheratioofthereverseddomainlengthand the structure period. From Eq.(2.11) it follows that the most efficient phase‐

 

8  

matchingwillbe for the firstorderQPM(m=±1)andwithaduty‐cycleof50%(D=1/2). Whereas for second‐order QPM, efficient phase‐matching is gained for a duty‐cycleof25%(D=1/4).

2.5.2Quasi‐phasematchedsecond‐harmonicgenerator

ThewavevectormismatchforQPMSHGcanbewrittenas:

Δk k 2k k ,(2.12)

wherek ,thesocalledm harmonicgratingvector,isdefinedas:

k .(2.13)

When theSHGand the fundamentalwavesarequasi‐phasematched, i.e.,Δk 0, the

gratingstructureperiodcanbewrittenas:

Λ 1,2,3, ….(2.14)

Thelengthofthegratingstructurecanthenbeadjustedformaximizingtype‐Iandtype‐IIQPMSHGprocesses.Type‐Iandtype‐IISHGarerelatedtod andd ,respectively.

2.6FabricationmethodsofQPMstructures

Inprinciple, aQPMdevice canbedesigned as long as one can change the signof thenonlinear coefficient in the material. There are several methods, including crystalgrowth[17],ionexchange[18],andelectronbeamwriting[19],tofabricateQPMdevic‐es.However, themost common technique formakingQPM structures is electric‐fieldpoling.High‐quality,periodicallyinvertedferroelectricdomainstructuresinfluxgrownKTiOPO (KTP) crystalswere fabricated in the late90’sbyLaurell andKarlssonusingthistechnique[20]. Aschemeforsamplepreparationandsubsequentelectric‐fieldpol‐ingisshowninfigure2.3.

 

 

 

Figure 2.3 Contact electrode poling process. 

 

 

 

V

Cutting and polishing, Solvent cleaning 

Resist spinning and Baking 

Resist exposure and Devel‐opment  Metal deposition (usually Al)   

Electric‐field poling

 

9  

Theferroelectriccrystaliscutintosmallpiecesandthec‐faceispolishedtoprovideoptimum light beampropagation. After cleaning the samples, the photoresist is spin‐coatedonthepolarfacesofeachcrystal.Theperiodicstructureiscreatedbyastandardphotolithographicprocess.Finally,metalelectrodesareformedbymet‐aldepositionontopofthephotoresist.Thelaststepisapplyingavoltageacrossthecrystaltocreatethedomainsofreversedpolarizationunderthemetalelectrodes. 

TheQPMdeviceshavea veryhighefficiency,when thedomain structures areuni‐formthroughoutthewholethicknessofthecrystalaswellasinthedirectionoftheop‐ticalbeampropagation. Fabricationofperiodically‐poled crystals canbedoneonly incertain crystalmaterials. Thesematerials are ferroelectric nonlinear crystals such asKTP, LiNbO andLiTaO .

2.71Dand2Dquasi‐phasematchedcrystals

For1DQPM,theincidentplanewaveisusuallypropagatinginthex‐directionandthephasemismatchcanbecompensatedforinastructurewiththeperiodΛ,whichisequal

toamultipleofthefundamentalspatialfrequencyofthestructure, .Ina2Dcrystal,

the electromagnetic radiation is propagating in the x‐y plane. In this situation, morethan one k vector appearwhich are not in the same direction as the initial planewave(oppositeto1DQPM).Figure2.4showsthepossiblek inthecaseoftheSHGprocess for2DQPM.Foragivenpump frequencyω, severalgratingvectorsk cancompensateforthephasemismatch.BychangingtheangleoftheincidentbeamonecanobservedifferentcasesofSHcorrespondingtothesedifferentk ’s.Eachpropagationdirectioninsucha2DQPMstructurecanresultinseveralSHpeakswithvaryinginten‐sity[21]. 

 

                                    

                                 Figure 2.4 Possible k‐vectors for compensating phase mismatch in a 2D QPM crystal. 

x

y

k

k

k

k

k

k

k

k

k

k

k

k

 

 

3KTPImplemlarizatifulfilledandsecoftheisities,cquencysizeoftanylpdamageshownwhileRischose

3.1Cry

Thecryb=6.3graphicneouspsistsofconsistchannetodomarelocarahedraK ions

                  

                  

PandRmentingofQionofa ferdinordertcond,ferrointeractingcrystalswityconversiothecrystalhosphate(e,excellentthatasligRKTPhassenasferro

ystalstru

ystalstruct99 andccdirectionpolarizatiofTiO octahtofshortaelalongthemaininversiatedinthea.Applingsandresul

 

 

 

 

 

 

                          

                          

RKTPcryQPMisusurroelectrictobesuitaelectricity.gwaves.Inthstablechon.Dependiisalsoan(KTP)isaptmechanichtRbdopiimilarpropelectricno

ucture

tureofKTP=10.584sa,b,and,n.ThecryshedrachainandlongTiec‐axis. Thionalongtesechannelavoltageatsinionicc

                  

                     Fig

ystalprouallyachievcrystal.TwblefornonFurthermodifferentehemicalaningontheaimportantpromisingccalstabilityingoftheKpertiesasKnlinearopt

Pisorthorh.Theprincofthecrstalstructunswhicha−Obonds,ischiralstrthec‐axisalsandarewasintheeleconductivit

ure 3.1  The cry

10 

opertiesvedbyperiwomainrenlinearoptore,thecryenvironmenndmechaniapplicationissue.Afercandidateby,andhighKTPcrystaKTPwithlticalmater

hombicandncipalaxis,rystal,wheureisshowrelinkedbwhichbuiructureisaandwillbeweaklybonectricfieldty[22],[23

ystal structure o

siodicallyinequirementtics:First,aystalhastontalconditicalpropernandtheprrroelectricbecauseofnonlineari

al(RKTP)faowioniccoialinthisw

ditslatticex,y,and,zrethec‐ax

wninFig.(3.byPO tetrld‐upachianimportaediscussedndedtobotpolinglea3].

of  [24

nvertingthets forsucharelativelyobetranspionsandvartiesareberocessingscrystalsucitshighreity.  Ithasracilitatesthonductivitywork.

econstantscorresponxisisthedi.1). Thecryahedras.ThiralstructuantparameinSectionththeoctadstothem

4]. 

espontaneacrystalm

yhighnonlparentinthvaryinglighestchoicessteps,theavchaspotasesistancetorecentlyalheperiodicy[4].Hence

are:a=12ndtothecrirectionofystalnetwoheTiO ocurerepreseeterwhenin3.3.TheKahedraandmovemento

eouspo‐mustbeinearityherangehtinten‐forfre‐vailablesiumti‐oopticalsobeencpolinge,RKTP

2.819 ,rystallo‐sponta‐orkcon‐tahedraentingatcomesK ionsthetet‐ofthese

 

11  

Thenonlinearopticalmaterialwhich ischosen for thisproject is KTPclosestrela‐tive,Rb‐dopedKTP,RKTP.InRKTPcrystal,the ionsaresubstitutedby ions.TheRKTPcrystalhassimilarpropertiesasKTPwithlessionicconductivity.

3.2Crystalproperties

Ferroelectricity

Ferroelectric crystals are materials with spontaneous polarization, which can beswitchedbetween twoopposing statesbyapplying anelectric field. The spontaneouspolarizationisdefinedbythecompositionofnegativeandpositiveions.Inequilibrium,thecenterofpositiveandnegativeionsdoesnotcoincide.Therefore,eachpairofnega‐tiveandpositiveionscanbeconsideredasanelectricdipole.Thesumoftheseindividu‐aldipoleswiththesamedirectionconstitutesthespontaneouspolarization.Anyregionofaferroelectriccrystalwithuniformspontaneouspolarizationiscalledadomain.Theinterfaceborderbetweeneachdomaininsidethecrystaliscalleddomainwall.ForKTPandRKTP, thedomains areorientedalong the c‐ direction forming the180 ° domainwallsbetweenthem.Thedomainwallsareparalleltothe100crystalplane[25],[26].

Ionicconductivity

Ionicconductivityplaysan importantrole fortheprocessofdomainswitching inKTPcrystals.ThenatureoftheKTPmaterialallowsthediffusionofatomsinthestructure.Undercertainthermalconditions,thepotassiumions(K )tendtohopoverthevacan‐ciesandtherebyintroducinganionicconductivityinthec‐axisdirection.Theconductiv‐ityalongthepolaraxiscanbeuptofourordersofmagnitudelargerthanthedirectionperpendiculartothepolaraxis[25].Highionicconductivitycauseslargecurrenttoflowthroughthecrystalwhenahighvoltageisapplied;thiscanresultinelectricaldamagesofthecrystal[27].InordertodecreasetheionicconductivityoftheKTPcrystal,onecanintroduceadopantlikeRb.SinceRb hasalargerradiusthanK ,itismoredifficultfortheRb topassthroughtheconductivitychannelsinRKTP.TheRb alsoblocksthelat‐ticeforthemovementofthemajorityoftheK ions,whichotherwisewouldhopfromsite to site.This isnotpossible, since theneighboringspacesarealreadyoccupiedbyRb .ThehoppingrateofK isthenreducedwithrespecttothatofRb astherearenopossibilities for theK ion topass theRb ion [28]. The ionic conductivity ofRKTP istypicallyabout2ordersofmagnitudelowerthanthatofflux‐grownKTP[29].

3.3Domainswitching

Nonlinear crystals such as KTP, which are used for QPM frequency conversion, aregrowntohaveonesingledomainandarecutalonga,bandtheccrystallographicaxes.Inordertoalterthepolarization,theappliedelectricfieldmustexceedacertainthresh‐old,calledcoercivefield.Thevalueofthecoercivefielddependsonparameterssuchasthe frequency, thewaveformof theappliedvoltage, thetemperatureandtheshapeofthecontactelectrode.

 

12  

Ferroelectricdomain inversionbycontactpoling, isbasedonthreesteps.First,nu‐cleationofthedomainwhichstarts fromtheedgeofeachelectrodewherethefield isstrongest.Second,thedomainwillrapidlypropagatesforwardalongthepolardirectionofthecrystalandatthesametime,thedomainwillslowlygrowssideways[30].Nucle‐ationistheformationofsmallantiparalleldomainsundercontactelectrodesinsidethecrystal,which grow and subsequentlymerge together. These domains can growbothalongtheferroelectricaxisorbysidewaysmotion.  It isestimatedthat inKTP,thedo‐main‐wall growth velocity in the polar direction is, at least, two orders ofmagnitudelargerthaninthex‐yplane.Thevelocityalongthey‐axisisaround30timeslargerthanthatalongthex‐axisduetothecrystalstructureofKTP[31].Domainmovementinthesidewaysdirectionsisnotdesiredinelectricfield‐poling,sinceitcausesdeviationfromtheidealdutycycle,domainmergings,andwillruinthegratingquality.

 

 

4Sili

Siliconbecontobservtallinemancequired

4.1Me

Theree[32], wmasksflexible(heightthenee

Oneetchingingfolldlesfabthesisbstoppedfortheture,etfrustumshouldthusteerperithecry

Anothavean

iconelec

isagrouptrolledoveedincryststructureto be fabrshape,itis

ethod

existsavarwhere a deis given. Te and shouttowidthredles.

ofthefirstg.Inthismelowingthebricatedbybecauseofdatthe(1(100),onetchantconcms. Furthebealigneddiousandodsandstiystalplanes 

Figure 4

ther technnydesirab

a 

ctrodef

IVmateriaerawideraalline,polyprovides hricated. Sinsthefirstch

rietyofmetetailed reviThe techniquld be ableratio).Itsh

tmethodsethod,thecrystalplayagroupinthefollowi111)plane.eshouldhacentration,rmore, todalongthetimeconsuillhavehigs,limitingth

   

4.1 Various type

ique is theleheight .

fabricat

alwithuniangebyinty‐crystallinhigh repronce Si canbhoicemate

thodsformiew of diffquewhiche tomakehouldalsob

forfabricaresultingnanesofthenUSA.ThisingreasonsHowever,aveagood,andtimeachieve a[110]crysuming.Besighaspectrhepossibil

         

es of silicon nee

eBoschproFigure(4.1

13 

tion

iquepropetroductione,aswellaoducibilitybemanipuerialforele

makingsilicferent fabrwould bestable andbepossible

ationofsilineedleshavsiliconsubstechniques:firstofasincethee

dcontroloninordertopyramida

staldirectioidethesepratiostructitytovary

edles obtained b

ocess.The f1.b)shows

b

rties.Foreofdopantsasinamorpand allowlated andctrodefabr

conneedlesrication tecbeneficiald sharp neeetovaryth

iconarraysvepyramidbstrate.Figeisnotsuiall,theetchetchratefontheetchoprevent tl shape, thon[33].Thproblems,muresisconthepitch.

    

by different tech

fabricatedsseveralco

example,itss.Whilethiphoussilicos devices wfabricatedrication.

s.See,forechniques fofor this predleswithespacing(

sofneedlesalshapedugure(4.1.a)itableforthingprocessor(411)plconditionstheneedleshe side ofismaskalimakingpyrnstrainedb

             

hniques [33] , [3

electrodesonstructed

sconductivispropertyon,thesingwith highin almost

example,reor siliconroject needhigh aspe(orpitch)b

sisbaseduetothew)showssuhepurposeshereneedlaneisfastssuchastes tobecomthe squarignmentisramidswithbythegeom

34] , [35]. 

sarepillarpillarsof

c 

vitycanycanbeglecrys‐perfor‐any re‐

eferenceneedlesds to beect ratiobetween

onKOHwetetch‐uchnee‐eofthisdstobeterthanempera‐mebluntre maskcritical,hsmall‐metryof

         

swhich150µm

 

 

heightenoughenoughThe

termedduringdeeplyshownanoptitainevdles,aofbreaRega

heightcess.Thshownmaskisthemaspikes

RoxhapproadeeprethecurneedlesdleswiThe

tric‐fieltweentthatcaused toobtain

and withhtobestabhandthefa

thirdtechddeep reactheetchininthevertinFig.(4.1mizedratioventallernverytallelakingwhen

arding thewhilethethishasbeein Fig.(4.2sunder‐etcasks shouldof200µm

hedetal.hachisbasedeactiveionrvatureofts.The fabrithatiprad

methodusldpoling,istheneedleninclude8o fabricateneedleswi

a radius oblewhileinabricationp

niqueformctive ion etng. In this tticaldirecti.c),areshaoofSF ,Oeedles[35lectrodemitcomesin

needles shtipradiusseninvestiga2.a). By usichedandpdbe twiceheight,the

Figure 4.2 a) 

havemodifdonalternnetchingththeneedlericationprodiusbelow

sedinthissthemodifsarewant86chipswiseveralwithaheight

a 

of 40 µm mncontactwprocessist

makingSietching (DRtechnique,ionandsligarp(1µmtgases,an].Althoughightleadtoncontactw

hown in Fstaysintheatedinrefeing a perferobeswiththedesiredistanceb

     

silicon spikes an

fied the silatingbetwhroughthetipandtheocess is fol100nm,se

worktofafiedisotropted,thefabith4 8m

wafers simutof1‐2µm

14 

made withwiththeKTtosomeext

electrodesRIE))wherplasmaetcghtlylateratipradius)ndpressurehthistechnoadditionawiththecry

ig.(4.1.c), oemicroscaerence[36ectly isotroh20µmtipedheight obetweenthe

nd b) long, sha

liconmicroweenisotroBoschproeanisotropllowedbyeeFig.(4.2.b

abricateSipicSF plasbricationprmm dimenultaneouslym.

h this techPcrystals,tent,compl

isavariane theprocchants reaally,atthesandtall(1e,theaspecniqueresualproblemystals.

one possibalerange,is]andtheropic plasmapradiusareof the spikeespikesne

rp needles [36]

oprobe fabpicSF plaocess.Theipicetchingoxidesharpb).

micro‐neesmaetchinrocessstarnsion.Furthy. The litho

nique. Thehowever,tlicated[34]

ntoftheBoessparamectwith thesametime.50µmheigtratiocanltsinthedsduetoth

le solutionsoptimizinresultingma etching (efabricatedes.Therefoeedstobe4

 , [37]. 

bricationpasmaetchinisotropicetdeterminepeningwhi

edlearray‐eg.Sincediftswith4‐inhermore,bographyma

b

e pillars artheyareno].

oschproceeters are ce silicon su.Thefinalnght).Bythbeadjustedesiredshahehighpro

n for reducngtheetchimodifiedspi(SF ), thed.Thediamore, for fab400µm.

 

process [37ngandanistchingdetestheheighichresults

electrodeffferentperinchsiliconbatchproceask is desi

re thickotsharp

ss(alsochangedubstrateneedles,euseofdtoob‐arpnee‐bability

cing theingpro‐ikesarecircularmeterofricating

7].Theirsotropicermineshtoftheinnee‐

forelec‐iodsbe‐nwafersessingisgned to

 

15  

  

4.2Photolithographicmaskdesign

Thedesignofthephotolithographicmaskforthesiliconarrayofneelesfabricationcon‐tainssevendifferentperiodsinsquareandinhexagonalarrangement.Thedistancebe‐tweentheneedlesinthedifferentarraysincludes:3.2µm,5µm,6.4µm,8µm,9µm,15µmand20µm.Forhavingsharpneedlesof1‐2µmheight,acircularmaskofroughly4µmdiameter shouldbedesigned foreachneedleas it is shown inFig.(4.3).However,this 4 µmdiameter value cannot apply for the 3.2 µmperiod due to the distance re‐striction.Forthe3.2µmperiod,thecircularmaskdiameteristakentobe2.2µm.Thustheheightoftheneedlesforthisperiodwillbesmallerthanfortheotherperiods. 

 

Figure 4.3 Schematic of mask, opening area and ratio of exposed and non‐exposed area for a 5 µm period grating. 

Theperioddiversityfordifferentelectrodearrayresultsinspecificdensityofcircu‐larmasks on each array. Furthermore, considering the same diameter of the circularmaskfortheperiodofmorethan5µm,theopeningarea(etcharea)willbedifferentforeacharray.Theopeningareaincreasesbyincreasingtheperiod.Thesetworeasons,be‐sidesthefactthattheetchratesarenotthesameforallpartsofthewafer(duetogasand plasma distributions in the chamber), lead to the predicted non‐identical needleshapesonthearrayswithdifferentperiods.Onewaytocontroltheseparametersistocalculate the ‘exposedarea’ (etching area) and the ‘non‐exposed area’ and arrangingdifferentelectrodearrayswithregardtothehighandthelowetchratepartsofthewa‐fer. Figure (4.3) shows a schematic of themask, the opening area and the estimatedetchingshapefora5µmperiod.Theparameter representstheratiobetweentheex‐posedandthenon‐exposedareaforthisperiod.

Aftercalculatingthevaluefor forallthespecificperiods,arrayswithhigher val‐ue were located close to the perimeter of the wafer, where the etch rate is higher,whereasarrayswith lower valuewere located in the centerpart where theetch isslower. Table (4.1) summarizes the different parameters of each array regarding theopeningareas,maskdiametersandthecorresponding value.

 

   

 

5µm A

    

5µm

4µm1µm

 

2µm

   

 

16  

Period (µm)  Type  Mask diameter (µm) : Non‐exposed area : Exposed area  ⁄

3.2  Hexagonal  2.2  3.73 5.02 0.74 

Square  2.2  3.73 6.38 0.58 

5  Hexagonal  4  12.56 9.09 1.38 

Square  4  12.56 12.44 1.01 

6.4  Hexagonal  4  12.56 22.86 0.55 

Square  4  12.56 28.44 0.44 

8  Hexagonal  4  12.56 42.64 0.29 

Square  4  12.56 51.44 0.24 

9  Hexagonal  4  12.56 59.14 0.21 

Square  4  12.56 70.25 0.18 

15  Hexagonal  4  12.56 182.2 0.07 

Square  4  12.56 212.44 0.06 

20  Hexagonal  4  12.56 333.85 0.04 

Square  4  12.56 387.44 0.03 

Table 4.1 Mask diameter, exposed and non‐exposed areas for different gratings on the wafer. 

Althoughthearrangementof thearrayswasdone to get the same result in all cases, alargedifferenceinthe valueleadtoantici‐pateddifferentneedleshapes.Comment:Theless exposed area open to the etch gasesthereisonthewafer,themoreuniformetchresultswillyield(duetolessconsumptionofetch‐gasmoleculesandthusamorehomoge‐nous etch plasma with less excausted vol‐umes). In reality, very low :s areunpracti‐cal, as you get very few etched features onyourwafer. 

4.3Fabricationprocess Figure 4.4  Schematic of the fabrication process. 

Theprocessstartedwithfiveone‐sided,500µmthick,p‐dopedwafers.Thedopantele‐ment was boron, the wafer orientation was (100) and the resistivity was 0.005 ‐0.02Ωcm.Thesecharacteristicsmadethewafersuitable forthecontactelectrodepol‐ing.AschematicofthefabricationprocessisillustratedinFig.(4.4).

           

Silicon wafer

 

 

 

 

 

a 

 

b 

 

c 

 

d 

 

e 

f 

 

g 

 

h 

 

i 

  

j 

Oxidation 

Resist deposition

UV exposure

Developing 

Hard baking 

Oxide etching 

Si etching 

Resist strip 

Oxide removal 

 

17  

4.3.1RCAcleaning

Thefirstphaseoftheprocessiswafercleaning(Fig.(4.4.a))toeliminateanycontamina‐tion,andalsotopreparethewaferforabetteradhesionofthephotoresist.Thestand‐ard cleaningprocedurecontainsfoursteps:

1. Removing organic contaminants by placing the wafers into a solution of H SO +H O at80‐130°Cfor5minutes,a.k.a.Piranhadipor7‐Upcleaning.

2.Washingindeionized(DI)waterN2bubblerfor5minutes.

3.Removingthethinnativeoxidelayer(~10 )byputtingthewafersintoasolutionofHF(50%)+H Oat25°Cfor100seconds.

4.WashingagaininDIwaterandbubblerfor3moreminutes,followedbydryinginhotN2

4.3.2Wetoxidation

Aftercleaningthewafers,theyweretransferredtothe‘wet’(O2andH2)oxidefurnace(Fig.(4.4.b)).Theoxidegrowthrateat1050 isapproximately5.3nm/min.Therefore,a1µmthickoxidelayer,requiresathreehoursrun.Theoxidelayerwilllateractahardmaskforthesiliconneedlefabrication,beingresistanttotheSF6etchplasma.

4.3.3Resistcoating

Theresistcoating is illustratedinFig.(4.4.c). It includesfourstepsthatareperformedsemi‐automatically:

Topreventpeelingoftheresist,thewaferswerecoatedwithanadhesionlayercon‐sistingofhexamethyldisilazane(HMDS)at130°C.Thesurfaceoxidelayeronthewaferforms long‐range hydrogen bonds with water absorbed from the air. When resist isspunontosuchasurface,itadherestothewatervaporratherthantothewafersurface,and this results in poor adhesion. Before applying theHMDS, the temperature of thechamberwasincreasedto130°Cinordertodehydratethesurface.ThenHMDSgaswasspunontothedehydratedsurface,providingamoreefficientadhesionofthephotore‐sist.Thewafersthenhadtobecooleddownto22 in20secondstopreventareduc‐tionof theresistviscosity.Thephotoresist (SPR7001.2) wasspincoatedwith5000rpmonthewafersresultinginthefinalresistlayerbeing1.2µmthick.Toevaporatethesolventandprepareitforexposure,thewafersweresoftbakedonahotplateat90°Cfor60seconds.Exposingawaferwithoutcoolingitdownresultsinathermalexpansioneffectofthephotoresist.

4.3.4UVexposure

TheUVexposurestepisoneofthechallengingpartsofthefabricationprocessduetothefactthat itrequireshighresolutiontoachievemicron‐scalestructures.Criticalpa‐

 

18  

rametersarethewavelengthoftheUVlightandtheexposuretime.Thewavelengthofthemaskalignerwassetto405nmwithan intensityof18.4MWcm .TheexposuretimecanbecalculatedbyknowingtheUVlamp’sintensityandthethicknessofthepho‐toresist,however,itneededmodificationforgettinggoodlithographicresults.

Reproduction of the photolithographic mask structure on to the photoresist wasdoneby ‘proximityexposure’whichretainsagapbetweenthemaskandthewafer topreventstictionordamage(Fig.(4.4.d)).SincethewavelengthoftheUVlightiscompa‐rablewith themask features, thegapmaycausesomediffractioneffectsandmayde‐gradetheresolutionofthepattern.Therefore,thegapalsoneededtobemodified.TheUVexposurestepwasstartedwith fourdifferentparameters tooptimize theprocess.Theexposuretimeandthelampintensitywerekeptatthesamelevelbuttheexposuremodeandthegapbetweenwaferandmaskwerechanged.ThedetailedparametersarelistedinTable(4.2).

Mode Exposuretime Gap Lampintensity

Wafer2 Softcontact 5 s 50 µm 18.4MW/

Wafer3 Softcontact 5 s 40 µm 18.4MW/

Wafer4 Vacuumcontact 5 s 50 µm 18.4MW/

Wafer5 Softcontact 5 s 30 µm 18.4MW/

Table 4.2 Exposure parameters for 4 different wafers in batch process. Wafer 1 is not listed since it is used as dummy. 

4.3.5Developing

Developingtheexposedphotoresistbeganwitha60secondssoft‐bakestepat110 invacuumfollowedbya15secondscoolingdownto22 .Thesoftbakerearrangestheexposedandnon‐exposedmoleculesofthephotoresist,averagesthestandingwavein‐tensity and prevents over‐exposure and/or under‐exposure. Moreover, it smoothensthephotoresistsidewallsandincreasestheresolution[38].ThebakingwasfollowedbydevelopingthephotoresistinaCD26developer.TheCD26developerisacombinationof<95%waterand3%tetramethylammoniumhydroxide.Thisdeveloperreactswiththeexposedphotoresistandthehydroxidecarboxylgroupsofthephotoresistaredis‐solvedinit.ThefinalprofileisdepictedinFig.(4.4.e).

4.3.6Hardbake

Before continuing theprocesswith dry etching, thewafers needed to be hardbaked.Hardbakingdecreasestheetchrateofthephotoresistandpreventsthetotalconsump‐tionoftheresistduringtheetchprocess.Furthermore,duringthehardbake,allthesol‐ventinthephotoresistisevaporated,whichimprovesadhesionofthephotoresist.

Thehard‐baketimeneededtobeadjustedduetothetypeofphotoresistemployedand its thickness. An under‐baked photoresist has low adhesion, is not polymerizedproperlyanddisplayshighetchrate.Ontheotherhand,over‐bakingofthesamplere‐

 

19  

sultsinreflowingofthephotoresistintheopeningsandwilltherebydegradethereso‐lution.

Inthisexperiment,oursampleswerehardbakedfor45minutesat110 .Fig.(4.4.f)illustratesthesampleafterthehardbakingstage.

4.3.7Siliconoxideetching

In this step, the hardmask (SiO layer) was patterned through the photoresists softmaskbyetching.Theprocesswasperformedinadry‐etchingmachinewithacombina‐tionof gases:N ,Ar,O , CF andCHF .Theestimatedoxideetch rate in thismachinewas 60 nm/min. Considering the the oxide thickness should be 950 nm, the sampleswereprocessedfor16minutes.Fig.(4.4.g)showstheschemeofthesampleafterremov‐ingtheoxidelayer.

4.3.8Siliconetch

Thesiliconwasetched isotropically throughtheopeningsof theoxide(hardmask)toformtheneedlesbyusingSF in thereactive ionetchingmachine.Toset therunningtimeoftheprocess,theetchrateneededtobeestimated.Theratedependsonseveraladjustableparameters: pressure, initial pump‐out time, speed flowof theplasmaandplasma power type can be varied. A detailed description of these parameters can befoundin[37].However,basedonthisreference,thetimeestimationforneedlesof1‐2µmheightis65seconds.ThefinalneedlearraysaresupposedtolooklikethoseshownintheschemeofFig.(4.4.h).

4.3.9Masksstrip

Afterformingtheneedles,thenextstepwasremovingthemasksandexposingthesili‐conneedles.Theremovalofthephotoresistwasdonebyapplyinganoxygenplasmafor15minuteswith500sccmat1000W(Fig.(4.4.i)).Theoxidemaskwasremovedfromthewaferbydippingitinhydrofluoricacid(50%).Hydrofluoricacidisagoodoptioninthiscaseduetoitsselectiveetchingofthesiliconoxide.2minuteswasfoundtobesufficientetchtime.RinsingwasinDIwaterfor30seconds.Themostreliablewaytoas‐certainthepointwhentheoxideiscompletelyremovedfromthewaferistestingitwithDIwater.Sincesiliconoxidehasahydrophilicsurface, ifafter thisstep, ifa filmofDIwater is stillobservedon thewholewafer, theoxidehasnotbeenproperly removed.This processwas continued until the surface showed a hydrophobic behavior,whichmeans that the oxide layer was removed [38]. Figure (4.4.j) shows the final needlesshapeonthesiliconwaferafterremovingthemasks.

4.4Modifications

High‐densitypatterns(5µmperiodicity)andlow‐densitypatterns(20µmperiodicity)inthesamephotolithographicmaskintroduceddifficultiesinthephotolithographyandthesiliconetchingsteps.Differentetchratesduetothedifferentopeningsmayresultin

 

20  

varyingheightoftheneedles.However,itcanbemodifiedbyarrangingdifferentarraysonthesamemaskasitwasexplainedinSection4.2.Consideringthelithographicprob‐lems,thediffractioneffectduetothesmallgapbetweenthemaskandthewafer,photo‐resistspreadingafterthehardbakeandanon‐optimalexposuretimewillleadtoapoorlithographic quality inmany of our patterns.Moreover, a small period between eachsinglecircularmasks,resulted inahigherriskofover‐exposuredue tocontaminationbetween exposed andnon‐exposedphotoresist anddecreasedquality of the lithogra‐phy.Usingvacuumcontactexposuremodeandthusdecreasingthegapsizeandtheex‐posuretimeresultedinbetterresolutionfordifferentfeaturesizes.

Regardingthesilicondioxideandthephotoresistmask,whichareusedforthepat‐terningandsiliconetching,theirlayerthicknessisachallengingissueinthefabricationprocess.Forhavingwell‐controlledetchingofsilicon,arelativelythicksiliconoxidelay‐erisneeded.This,consequently,requireslongeroxideetchingtime.Ontheotherhand,alongersiliconoxideetchingdirectlyresultsinahigherphotoresistconsumption.Thus,a thick layerofphotoresist isneededwhichpossiblyresults inapoorpatternquality.However,athinnerphotoresistlayeralsorequiresathinoxidelayer.Thisleadstonotwell‐controlledsiliconetchingand,again, lowresolutioninpatternquality.Therefore,thethicknessesofthesetwolayersneededtobeoptimized.Theoptimizationwhichwasusedinthisthesiswasathinneroxidelayerandaslightlythickerphotoresistlayerwithregardstotheirinitialvalues.Hardbakingtimewasdecreasedinordertopreventre‐flowofthephotoresistintheopenings.TheinitialandmodifiedvaluesofthedifferentparametersarelistedinTable(4.3).

Oxide

thickness

Photoresist

thickness

Exposure

time

Proximitymode Proximity

gap

Hardbake

time

Initial 950nm 1.2µm 5 s Softcontact 30‐50µm 45min

Modified 470nm 1.4µm 4 s Vacuumcontact 10µm 30min

Table 4.3 Initial and final parameters for needles fabrication. 

4.5FinalfabricatedSielectrodes

Thefabricationprocessofthesiliconmicroneedleshasbeenoptimizedineveryfabrica‐tionstepinordertoconstructneedleswiththedesiredperiodanddimensions.Howev‐er, due to the reasonsmentioned in Section 4.4, the resulting needles have differentshapes.Figures(4.5)‐(4.8)illustrate5µmperiodpatternsindifferentstepsoffabrica‐tion.

 

 

         

                  

            

           

Fortcularmreason2.2µmthe resThe saetching

      

                  Figu

peri

Fi

pe

Fi

af

              

the3.2µmmaskscollaforthecol

mforthispast).The secme etch dug,plateau‐s

                  ure 4.9 Optical m

iod pattern afte

igure 4.5 Optic

eriod pattern af

igure 4.7 SEM  i

fter isotropic Si 

mperiodpaapsedon thllapsingforatternwhiccond reasouration forshapedelec

microscope ima

er lithography. 

cal microscope 

fter lithography

image of 5 µm

etch.         

tterns,whehesurfacerthisperiochissmallon ishavingr the 3.2 µctrodes,and

age of 3.2 µm 

image of 5 µm

y. 

 period pattern

20 µm

10 µm 

16 µm

21 

              

                  

entheSietdue tooveodisthefalerthanforg a variablµm, 5 µm adveryshar

                

m 

n 

tchprocesser‐etchingactthatthertheotherle circularand the 20rpspikes,r

Figure  4.10  O

µm period pat

Figure 4.6 Opt

period pattern 

Figure  4.8  SEM

mask removal,

swasbeing(Fig.(4.9) ‐circularm

rperiods(4maskdens µm casesespectively

Optical microsco

ttern after oxide

tical microscope

after oxide etch

M  image  of  5 

formation of p

gstopped,‐ (4.12)).Tmaskdiame4µmdiamsityonone resulted iy.

cope  image  of 

e etching. 

e  image of 5 µ

ching. 

µm  period  af

plateaus. 

10 µm 

15 µm 

16 µm 

 

 

thecir‐The firsteterwaseterforewafer.in over‐

 

3.2 

µm 

fter 

 

 

      

Figuthesuretching

Unlikeover‐ettheundthesep(4.13)‐etching

       

 

       

 

Figu

afte

Fig

per

Fig

afte

ure (4.12)srfacewith3gprocess,t

thepatterntched,respdercutisexpatterns, th‐(4.16)shgoftheSia

ure 4.11 SEM im

er isotropic Si et

ure 4.13 Optica

riod pattern aft

ure 4.15 SEM i

er isotropic Si e

shows the3.2µmperiheirtraceo

nswith5µpectively,thxactlyequahe spikes aowthepatandthemas

mage of 3.2 µm

tch, masks drop

al microscope  i

ter lithography.

mage of 20 µm

etch.

patternsaiods.Sinceonthesurfa

µmperiodshepatternsaltothecirare very shtternimagesksremova

 period pattern

p. 

image of 20 µm

 

m period patter

16 µm

20 µm

40 µm

22 

               

            

after removthecirculafacedamag

sand3.2µmswith20µrcularmaskharpwith aesafterthealstepsfor

              

              

n  F

a

m 

rn 

         

ving thecoarmaskscoedthewho

mperiods,µmperiodskdiametera tip diameelithographthe20µm

Figure 4.12 SEM

after mask rem

Figure 4.14 Op

period pattern 

Figure  4.16  SE

tern after mask

ollapsedcirollapsedduolepattern.

whichareareisotrop(4µm).Aseter less thhy,siliconperiodpat

M image of 3.2 

oval, over etche

ptical microscop

after oxide etc

EM  image  of  2

k removal, need

rcularmaskuringtheis.

under‐etchpicallyetchaconsequhan1 µm.oxideetchttern,respe

µm period pat

ed pattern. 

pe image of 20

ching. 

20  µm  period  p

dles are formed

10 µm 

20 µm 

20 µm 

 

ks fromsotropic

hedandhedanduence,inFigureshing,ICPectively. 

 

 

ttern 

0 µm 

pat‐

d. 

 

 

DespcouldbportantvariatioapproxFig.(4.1µmand

     

 

     

 

Figure

riod pa

of 50 n

Figure

od pat

approx

piteoveretbeusedfort.Forthe2onsoverwximately5017)–(4.18)datopdiam

e 4.17 SEM imag

attern with heig

nm. 

e 4.19 SEM imag

ttern with heigh

ximately 2 µm 

chingaswrintended20µmperiowholesurfa0nm,whic).Regardinmeter<2µ

ge of a needle i

ght of 1.73 and

ge of a needle i

ht of 2.8 and to

2

ellasnot‐euse;periododpattern,ceofthesh issatisfangthe5µmµm,asshow

in 20 µm pe‐

d tip diameter 

in 5 µm peri‐

op diameter of 

1 µm 

2 µm 

23 

enoughetchdicpoling.,thespikesiliconelectactory fortmperiodpawninFig.(4

                      

                      

hing,thefaTheirheigshaveahetrode.Thethepurposattern,the4.19)–(4.20

Figure 4.18  S

Figure 4.20  S

abricatedspghtandtopightof1.73tipdiametseof thisthplateausha0).

EM image of 20

EM image of 5 

pikesandppdiameter3µm,withteroftheshesis,asshaveaheigh

0 µm period pa

µm period patt

10 µm 

4 µm 

plateausareim‐aslightpikesishowninhtof2.8

 

 

ttern.  

tern.

 

24  

Figure 5.1 Schematic of poling technique with silicon electrodes. 

5FabricationofperiodicallypoledRKTPbysiliconstampConsideringthefabricationofQPMcrystalswithelectric‐fieldpoling,themostcommondefectsinthefabricationaremissing/mergeddomainsanddomainbroadening.Fabri‐catingsub‐micronstructuresbecomesmorechallengingsincethefringing‐fieldeffectismoresevere. Inthiscase, theusualtechniqueforpatterningofthemetalelectrodeonthesamplesurfaceisphotolithographyprocess(seeSection2.6).Thephotolithographyprocessinvolvesseveralstepswhichmustbeperformedonthesurfaceofeachindivid‐ualsample.Thisprocessistimeconsumingandcanleadtoslightlydifferentpatternsoneachsample.Furthermore,byhavingthesamethicknessofphotoresist,insub‐microndomainfabrication,theaspectratioisincreasedandthequalityofphotolithographyisdecreased.Therefore,thelithographyprocessneedstobemodifiedinordertoachievehighqualityinthefabricateddomainswithlessdefects.

In thisworkwe investigate anew techniqueof electric fieldpolingusing aboron‐dopedsiliconarrayofneedlesasanelectrode.Unliketheprevioustechniques,herethewholearrayofelectrodes ismanufacturedindependently fromthecrystalonasiliconwafer.Siliconisaflexibleandhighlydevelopedmaterial,therefore,transferringlithog‐raphysteps fromthecrystal surface toaSielectrodecanbehelpful inorder toover‐come the photolithographic problems of the conventional technique. As illustrated inFig.(5.1),thefabricatedsiliconarrayofneedlesismountedonthecrystalandthepolar‐izationisalteredbyapplyingavoltageacrossthecrystalandthewholeelectrodestruc‐ture. 

 

 

 

 

 

 

 

 

 

Inorder to study theelectric‐fieldpolingusing theSielectrode,differentelectrode

gratingswithdifferentperiodsrangingfrom3.2μmto20μmwerefabricated.Thetopdimensionofeachelectrodeonthesurfacevarybetween50nmand1.9µm.Withthehelpofthesiliconarrayofelectrodes,theperiodicpolingcanbedoneusingfewerstepsandonlyonesiliconelectrodestructurecanbeusedforpolingofseveralsamples.

V 

RKTP

 

      

RKTP crystal

Poled

Silicon electrode structure 

with nano‐sized needles 

 

 

Inthistyofthly‐pole

5.1Sam

ThemaRKTPwcrystalldirectiosamecbekept

5.2Po

Thepothehigshownotheratalmansaturatthecry

5.3Fa

Severalbestreduratiorameteforthecontrolidone,onds.S[4].Thethe nuc

chapter,pehecreatedddcrystalsi

mpleprep

aterialusedwaferwaslographic aonineachwconductivitytinthesam

olingsetup

olingsetupgh voltage.inFig.(5.2andtokeepnually. ThetedKClsoluystalandth

brication

lparametesults.Thepon, thepulsers. Inthefappliedvolled(duetotheelectriShortelectresiliconelecleation sta

eriodicpolidomainstruisevaluated

paration

d in thiswcut into saaxes, respewafer,eachy [20].Themeorderas

p

includedaThe samp),andwerpthemincoe crystal anutionwhichesiliconel

nofPPRKT

ers canbepressurebse shapeaframeoftholtage.Theothemanuicalpulsesric‐fieldpuectrodestrarted from

ingwiththucturesissd.

work isRb‐amplesofdectively. Sihcolumninerefore,dusintheiniti

aholderanle holder ceusedtofontact.Thend the silichgivesamectrode.

Figure 

TP

adjusteddetweenthend themahiswork,apressurebualaligningwerechosulsesdecreructurewasm this face.

25 

histechniqustudiedand

‐dopedpotdimensionince the conthex‐axisuring theclialwaferb

ndanelectrconsisted ofixtheelecesiliconelecon electromoreunifor

       

5.2  Poling cell

during theecrystalanagnitudeofallparametbetweenthgandmounsentobetrasethedosconnecteThe polar

ueisdiscusdtheoptica

tassium titas10 5onductivityswasexpecleaningprooxtomake

riccircuitwof two liqutrodestrucectrodewasodewere cmconnecti

. 

polingprondtheelectf theapplieterswereshecrystalannting).Sincriangularwmainspreadtothecrization inv

ssed.Furthalperforma

anylphosp1mmalo

y mainly vactedtohavocedure, thefurtheran

whichwasuid electrodctureandtscenteredonnected tionoverth

cess inordtrode,theeedelectricsetatconstndmaskcoethepolinwithaduraadingbeyopolarfaceversionwa

hermore,thanceofper

phate (RKTong thea,aries alongvemoreorhesamplesnalysissimp

usedforade cellswhthecrystalontopofthto the circuhewholesu

 

der toachielectric‐fielfieldarestantvaluesouldnotbengprocessiationof5mondtheeleeofthecrysas executed

hequali‐riodical‐

TP).Thebandcg the y‐lessthes shouldpler.

applyinghich aretoeachhecrys‐uit by aurfaceof

ieve theldpulseuchpa‐sexcepteideallyisarap‐millisec‐ectrodesstalandd for 24

 

26  

RKTPsamples.Twodifferentsiliconmasks,withaperiodicityofΛ=5µmandΛ=20µm,wereusedfortheperiodicpoling.

5.4Polingmonitoring

In low‐conductivity ferroelectrics such as LiNbO , the domain inversion can be con‐trolledbymonitoringthecurrentflowinginthepolingcircuit[2].Forhigh‐conductivityferroelectrics, such as KTP, the applied field leads to a strong ionic current passingthroughthesample.Therefore,monitoringthepolingprocessbythecurrentflowisin‐accurate.Inordertocontrolthepoling,andtobeabletoreproducetheresults,thepol‐ingprocesswasinsteadmonitoredbytheelectro‐opticeffectandin‐situSHG.

A schematic of the experimental setup is shown in Fig.(5.3). TheHeNe laser beamwaslinearlypolarized45°tothez‐andy‐axisofthecrystalbythepolarizer.Thebeamwentparalleltothex‐axisofthecrystalandtheintensityofthebeamwasmeasuredbyananalyzerrotatedby90°withrespecttothefirstpolarizer.Whenthedomainspropa‐gatedownthroughthecrystal,theoutputintensityfromtheHeNelaserwillbemodu‐latedinthecrystal[39].Thiseffectcanbeobservedwithaphotodetectorandanoscillo‐scope.

 

 

 

 

Figure 5.3 Poling monitoring setup. 

Theelectro‐opticmonitoringmethodgives informationas towhether thepolariza‐tion isreversedornot.Hence,onlyaqualitative indicationof thepolingresultcanbeobtainedbythistechnique.  

Inorder toobtainmoredetailed informationon thehomogeneityof the fabricateddomains,thein‐situSHGtechniquewasapplied[29].ATi:sapphirelaserwaslooselyfo‐cused on the crystal aperture and tuned to the phase‐matchingwavelength. The SHGsignalwasmeasuredbyapowermeter.ThelaserbeamwasscannedinthehorizontalandtheverticaldirectionstocheckthequalityofthedomainsbyobservingtheSHGindifferentpositionsacrossthecrystalaperture.

  

 

 

  Silicon Mask

+45° ‐45°

High voltage

Detector

 Oscilloscope

  

 Ti‐Sapphire

HeNe   

Polarizer

Lens Filter

 

 

Power meter

Wave plate 2  

 

27  

AsaCWTi:sapphire laser(SpectraPhysicsModel3900S,pumpedbya10WMilenniaX )hasemissionlinesbetween700nmand950nm,theQPMgratingperiodof5µmcanbeevaluatedbyusingthe1storderSHG,i.e.,λ 894nm.FortheperiodsofΛ=20µmthe5thorderSHGhastobeusedwitha fundamentalwavelengthofλ 842nm(ac‐cordingtoEquation2.14).

 

 

6Cha

6.1Sel

ThedoThesamat80faceofthedom

6.1.1Do

ItwasoplesarstructuthecryexplainNeverthesinsoBoth

contactheightthe polcrystaluniform

a:    Dom

a) 

aracteri

lectiveet

omainstrucmplewaspforatimetheRKTPcmainstruct

omaineval

observedthenothomoureonthecystalfeaturnedbyanisheless,a2DomeoftheRhofthepolt(Fig.(6.1))ofthespikling cell isandthesilmcontactin

main dimension

izationo

ching

ctureafterplacedinaetypicallycrystalwhiture[40].

luationfor

hatthedomogeneous.c polarfaesa1DstrsotropicdoDdomainsRKTPsamplarfacesco).Thisnonkesleadingnot optimliconelectrnsomeare

n 3.7   3.7 µm. 

ofdoma

polingwasolutionorangingbeilethec su

the20µmp

mainshapeSomepartsce.Howeveructure.Thmaingrowstructurehples.ontain1Da‐uniformitytoaninho

mized for elrodearepreasofthesa

20 µm 

28 

ainstruc

as revealedofKNO anetween5‐urfaceisle

periodgrat

eanddimesofthepoer,corresphereasonfowthintheKhasbeensu

and2Ddomyiscausedomogeneoulectric‐fieldressedtogeampleand

       b:  

b

ctures

dbya selecndKOH(2:15min.Thftessential

ting

nsioninthledsamplepondingdoorthiskindKTPcrystaluccessfully

mainstructdononehauselectroded polingwetherinthenon‐unifor

  Domain dim

b) 

ctiveetchin1moleratiheetchantllyuntouch

ecaseofthesexhibitdmainsontdofdomal,asdiscusfabricated

turesduetondbyslighecontact.Owith this neepolingcermcontact

ension 1.5   3.

ngof thesio)in20mattacksthehed,thusre

heΛ=20μdomainswithec polarainmergingsedinSectonbothpo

othenon‐uhtvariationOntheotheew techniqellwhichreinotherpa

.7 µm. 

20 µm

amples.mlwaterec sur‐evealing

μmsam‐itha2Drfaceofgcanbetion3.3.olarfac‐

uniformnsintheerhand,que. Theesultsinarts.

m 

 

 

c:    Dom

e:    Dom

µm in betw

Figure 6.1

kV/mm w

Fortheandspiinginmspikesingthe1.6µmFig.(6.2

 

Fig

fro

c) 

e) 

main dimension

main dimension

ween. 

1 Domain struct

ith the triangul

e20µmgriketipsaremanybrokeforthispeecrystalanmandtheti2.b).

 

 

 

                

 

gure 6.2 SEM p

om 50 nm to 50

a) 

n 3.9   16 µm. 

n 5 µm at the pl

tures of a poled

lar pulse of 3 m

ratingelecteverysharpenspikesariodwas1ndtheSielipdiameter

picture of a nee

00 nm during co

1 

lace of needles,

d RKTP sample 

ms duration.  

trode,thedp.Thereforand,thus,a1.73µmwitectrodetogriscorresp

edle in 20 µm p

ontacting silicon

µm

20 µm

20 µm

29 

d:    D

, 1.6 

f:    D

with 20 µm gr

densityofsre,thepresanetinhomthatipdiagetherintpondingly

period a) before

n chip and crys

d

f

Domain width 3

Domain width 2 

rating period of

spikesoverssureoneamogeneousameterof5thepolingcreducedto

e and b) after p

tal.   

b)

d) 

f) 

3 µm. 

µm. 

f Si stamp. Appl

rtheelectrchsinglespcontact.Th50nm.Howcell,theheo500nm,s

poling. The tip d

lied electric fiel

rodesurfacpikeishighheinitialhwever,afteeightisredseeFig.(6.2

diameter increa

1 µm

20 µm

20 

ld is 5.8 

eislowhresult‐heightofrpress‐ducedto2.a)and

ases 

m 

µm

 

 

Besidesbrokencaused

Fig

Sc

Extremvariatiotherebyspike tsharpspartof

Fig

ar

sc

FigurenificatiHence,electric

sthehighpnspikes.Figbyasharp

gure  6.3  SEM  p

cratch and brok

melysharpsoninthedoyatraceintip insideaspikemaythedomain

gure 6.4 SEM p

rea dimension: 

cratch, width: 20

(6.4.a)illuon.Therinitisprobacfieldcaus

a

a) 

pressureongure(6.3.apspikedue

pictures  of  scra

ken electrode tip

spikesontomains.Asnsomeofta scratcholeadtothenareawhe

pictures of dom

1.5 µm, spike s

00 nm.  

stratesaringshapesaablethatthedbyvery

a) 

neachspika)illustratetotheman

atches.  (a)  Scra

p (lying inside) o

theSielectsharpspikethedomainn the crysteringshaperethepola

mains formed by

cratch dimensio

ing‐shapedarealsoevieformatiosharpspik

1 µm

2  µm

30 

ke,themanesascratchnualalignm

ratch  of  one  sp

on the RKTP cry

trodesurfaecanpenetnareas.Figtal surfaceedomain.arizationis

y a sharp spike

on:  450 nm. (b

ddomainwidentinthenoftherinkes.

b)

nualalignmhonthecment.

pike  on  the  cry

ystal surface.

cecanbeatrateintotgure(6.3.b). The shapRing‐shapenotinvert

e. (a) Outer dim

b) Not inverted a

withthescredomainswngshapesr

)

b)

mentisanotfaceofth

stal  surface af

anotherreahecrystals)showsapeof electreddomainsed. 

mension: 3 µm, 

area inside a do

ratchinsidewithoutscrreliesonth

1 µ

1.5 µm

therreasonhecrystalw

fter alignment.

asonforthsurfaceandpossiblebrric field forsfeaturea

inner not inver

omain without 

eitinhighratch(Fig.(heinhomog

µm

nforthewhichis

  (b) 

eshapedleavesrokenSir such acentral

rted 

any 

ermag‐(6.4.b)).geneous

 

31  

Figure (6.1)shows thedomainstructuresofoneparticularRKTPsample fordifferentpartsof thecrystal.Thereare2Ddomainstructuresaswellaselongated2Ddomainsand1Ddomainson thec polar face(Fig.(6.1.a), (6.1.c), (6.1.e)).Theircorrespondingdomainsonthec faceareshownontherighthandside(Fig.(6.1.b),(6.1.d),(6.1.f)).Inmostofthedomainsonthec face,ring‐shapedandspikescratchescanbeobserved.

Sincethespiketipdiameterisincreasedafterthesamplemounting,thefabricateddo‐main dimensions are larger thanwhatwas anticipated. Due to the quasi‐one dimen‐sionalstructureoftheRKTP,2Ddomainstendtomergetogetherintheb‐directionandforma1Ddomainstructure.Therefore,insamplescontaining2Ddomainstructures,thefabricateddomainsarelargeralongtheb‐axis(~3.5‐4.2µm)andnarroweralongthea‐axis(~2.4‐3.8µm).Theircorrespondingdomainsonthec polarfacefollowsthesamecriteria.Insampleswith1Ddomainsonthec polarface,thewidthofthedomainsinthea‐directionislargeratthelocationofthespikecontact(~4.1μ )andnarrowerinbetween (~1.6 µm). The information regarding thedomains dimensions in bothpolarfacesarelistedinTable(6.1).Thereasonforthisisthehighvalueofthefringingfieldaroundthespiketipandalsothehardcontactoftheelectrodeandthecrystalsurface.Thecorresponding1Ddomainsonthec polarfacearehomogenousandhaveawidthof1.9µm.

Domaintype 2D Elongated2D 1D

Dimensioninc 3.7µm 3.7µm 3.8µm 16.5µm 1.6&4.1µm

Dimensioninc 1.5µm 3.7µm 2.9µm 1.9µm

Table 6.1 Domain shape and dimension in Λ = 20 µm grating 

6.1.2Domainevaluationfor5µmperiodgratingSixRKTPsampleswerepoledwitha5µmgratingelectrode.Gratingpatternsof5µmperiods are successfully reproduced on the polar faces of the crystal by applying anelectricfieldof5.4kV mm⁄ withatriangularpulseof5msduration.

Fabricateddomainstructureswitha5µmperiodsiliconelectrodeismoreuniformacrossthecrystalsurface.Uniformdomainstructurescomefromthefactthatthecon‐tactareaofthecrystalsurfaceandtheelectrodeisnowmorehomogenousthanintheΛ=20μmcase.Severalparametersleadtotheuniformcontact.First,alltheplateausonthesiliconarrayofelectrodeshavethesameheight.Second, topareasoftheplateausare perfectly flat because they have the initial surface of the siliconwafer. Third, thedensityof theplateaus ishighoverthewholesurfaceandthisresults ina total lowerpressureoneachindividualneedleelectrode.Therefore,betterdomaingratingqualityistobeexpectedfortheRKTPsamplespoledwithΛ=5μmSielectrodes.However,theproblemof themanualalignmentand thepressingof thesiliconarrayandcrystal to‐getherstillaffectsthepolingprocess.

 

 

Figure 6.5

length. (a)

Figure 6.6

length. (a)

Due todirectiothesamentsamin thesturesa

Figure 6.7

length. (a)

a) 

a) 

a) 

5 Domain struct

)    Domains d

6 Domain struct

)    Domains d

the quasi‐on.Howevemedimensimplespoledse two samrereprodu

7 Domain struct

)    Domains d

tures of poled R

dimension 2.5 

tures of poled R

dimension 2.5 

‐1D crystaer,inthea‐ions.Fig.(6dwithanemples haveucedsucces

tures of poled R

dimension 2   3

RKTP sample wi

 4.1 µm (b)

RKTP sample wi

 3.8 µm (b)

l structure‐direction,6.5)andFigelectricfieldsimilar shssfully.

RKTP sample wi

3 µm (b)   Do

15 µm

15 µm

15 µm

32 

ith 5.4 kV/mm t

  Domains widt

ith 5.4 kV/mm t

  Domains wid

e of KTP, thdomainsog.(6.6)shodofthesamhapes and t

ith 5 kV/mm tria

omains are distr

b

b

m

m

triangular pulse

th 2.5 µm. 

triangular pulse

dth 2.6 µm. 

he domainonboththewthedommemagnitthe dimens

angular pulse o

ributed random

b)

b)

b)

e of 3 ms durati

e of 3 ms durati

s tend to epolarfaces

mainstructuude.Thefasions and t

of 3 ms duration

mly width: 1 µm

ion and 5 µm g

ion and 5 µm g

elongate inshavefairluresoftwoabricateddthe domain

n and 5 µm gra

m. 

1

rating 

rating 

n the b‐lymuchodiffer‐domainsn struc‐

 ating 

15 µm

15 µm

15 µm

 

33  

On the other hand, inhomogeneous pressure, lower applied electric field, and lack ofgoodalignmentresultinmanymissingdomainsasisshowninFig.(6.7).

6.2Opticalcharacterization

Thesiliconspikesareverysharpforthe20µmperiodelectrodes.Thisresultsinanon‐optimal duty cycle of the fabricated domain structures. 5th order SHG process in thedomainstructureswithnon‐optimaldutycycleresultsinaverylowSHsignal.However,inthecaseofthe5µmgrating,sincethedutycycleinthiscaseisclosetotheidealcaseof50%,the1storderSHGprocesswithameasurablesignalpower,isobserved.Inordertoevaluatethequalityof the fabricatedgratings in theRKTPcrystals, theopticalper‐formanceoftwoofthesampleswiththeΛ=5µmgratingwereinvestigated.Thegener‐ated SH signal which was created in this experiment is close to c polar face, corre‐spondsto[1,0]reciprocallatticevector(RLV).ThisRLVisequivalentto1Dgrating.

6.2.1Conversionefficiency

Oneway to check thepolingquality is to observe the SH conversion efficiencyof thepoledcrystal.ForaGaussianbeam,theconversionefficiency isgivenbythefollowingexpression[41]:

η Lh B, ξ . (6.1)

According to Eq.(6.1), the SH conversion efficiency increases linearlywith the funda‐mentalpower,P ,andthecrystallengthL.TheconversionefficiencymeasurementweredonefortheSHGofaTi:Sapphirelaserbeamat894nmfortwoPPRKTPsampleswith5µmgratingperiods[22],[29].Figure(6.8)showstheexperimentalsetup.

Figure 6.8 SHG setup for measuering conversion efficiency. 

ThemeasuredefficienciesforthesetwosamplesaregiveninFig.(6.9)andrepresenttheefficiencyofpolingin1Ddomainstructureclosetoc polarface.Theexperimentaldataoftheconversionefficienciesfitsverywellwiththelineartheoreticalfunction.

 

 

 

 

Power meter

Ti‐Sapphire     

Polarizer Lens Filter

  2  

Half‐wave plate

sample

  

 

34  

0 100 200 300 400-0,05

0,00

0,05

0,10

0,15

0,20

0,25

0,30 efficiency (sample #1) efficiency (sample #2)

effic

ien

cy(%

)

PF(mW)

0 50 100 150 200 250 300 350 400 450

0

200

400

600

800

1000

1200 SH power (sample #1) SH power (sample #2)

PS

H(µ

W)

PF(mW)

 

 

 

 

 

 

 

 

Figure 6.9 The SHG conversion efficiency as a function of fundamental power for 2 samples with 5 µm grat‐

ing period. Circles and squares are measured data and red lines are linear fis, respectively.  

Inordertocomparetheconversionefficiencyforthetwodifferentsamples, it iscom‐mon to use their normalized conversion efficiencies. The normalized conversion effi‐ciencyreferstoaninputof fundamentalpowerof1Wattandacrystal lengthof1cm(withunitof%/Wcm).Itisdefinedasfollows:

η h B, ξ .(6.2)

Themaximumnormalizedconversionefficiencyobtainedforthebestspotinsample#1was1.27%/Wcmand1.1%/Wcmforsample#2.ThegeneratedSHpowerswere1050µWand912µWforaninputpowerof373mWand375mWinsample#1and#2re‐spectively.BasedonEq.(6.2),theSHpowerhasaquadraticdependenceonthefunda‐mentalpowerwithaproportionalityfactorof .Figure(6.10)belowshowstheSHsignalsdependenceonthefundamentalpower.Theexperimentaldataagreeswellwiththetheoreticalfit.

 

 

Figure 6.10 The SH power versus  the  fundamental power  in 2 different samples with 5 µm grating period. 

Solid squares and circles are experimental data and red solid line is the theorical curve. 

 

35  

6.2.2Temperatureacceptancebandwidth

Thephase‐matchingphenomenon is sensitive tomanyparameters, suchas thewave‐length, temperature, and the polarization of the interacting waves. The acceptancebandwidth,alsocalledthephase‐matchingtolerance,isdefinedastherelativechangeintheparameterthatcausesadropoftheoutputpowertoonehalfofthemaximumvalue.In thiswork, the temperature tuning characteristics of the SHG nonlinear interactionwasinvestigated.Forafixedwavelength,thetemperaturebandwidthcanbeexpressedas[16]:

∆ . Δ Δ ,(6.3)

whereαisthethermalexpansioncoefficientinthex‐directionandΔnisthedifferenceintheindexofrefraction.Listhelengthofthestructurecontainingtheuniformperiodicferroelectricdomaingrating.Temperature bandwidthmeasurementwas performed for frequency doubling of a

Ti:Sapphirelaserat894nm.ThesetupisshowninFig.(6.11)belowandthemeasure‐mentwasdoneonly for sample#2.The fundamentalbeamhad 369mWpower andwaspolarizedinthez‐directionbya ‐halfwaveplateandalinearpolarizer.Thebeamwasfocusedbyalensintothecrystal,closetothec polarface.Acopperelementwasused together with a temperature controller (ILX Lightwave LDT‐5525) in order toachievethetemperaturetuning. The temperature in the copper element increased from 22

to 39 with 0.2 steps and the SH power was monitored at every step.

Figure 6.11 Temprature tuning setup for SHG. 

ThetemperaturetuningcurveobtainedforthePPRKTPcrystalwithaperiodof5µmisshowninFig.(6.12).Thesolidlineisatheoreticalfunction,whichfitstheexperimentaldatawell [42].Thisconfirmsthatahigh‐qualityQPMstructure inthiscrystalwasob‐tained.

  

Power meter

Ti‐Sapphire     

Polarizer Lens Filter

  2  

Half‐wave plate

 

Temperature controller

 

36  

20 22 24 26 28 30 32 34 36 38 40-100

0

100

200

300

400

500

600

700

800

900

1000

PSH sample #2

PS

H (µ

W)

T (C)

CT 25.4

 

 

 

 

 

 

Figure 6.12 Temprature tuning curve for SHG, black squares show observed data and red line is the   fit‐

ting funcion. 

Ifthepolingqualityislowandtherearemissingdomainsinthestructure,theeffectiveinteraction lengthof aperiodically‐poled crystalwill be correspondingly reduced.Ac‐cordingtoEq.(6.3),bymeasuringthebandwidth,theeffectivecrystallengthcanbees‐timatedasfollows:

L .

∆Δn αΔn . (6.4)

Fromthemeasuredtemperaturebandwidthof4.25 ,theeffectivecrystallengthises‐timatedtobe5.4mm.Thisvalueisveryclosetothephysicalgratinglengthof6mm.Inthisconfiguration,theQPMSHGprocessutilizesthed33coefficientoftheRKTPcrys‐tals, therefore, the effectivenonlinear coefficient is related to thed33 in the followingway[16]:

d d .

The experimentally obtained nonlinear coefficient of our PPRKTP sample can be ob‐tainedfromrearrangingEq.(6.2).Thus,theeffectivenonlinearcoefficient iswrittenasfollows:

d,

. (6.5)

Inordertoevaluatethequalityofthepoledcrystal,d canbecomparedwiththevalue,calculatedusingd ,astakenfromreference[43].Inourexperiment,theRKTPcrystalexhibitsaneffectivenonlinearcoefficientof9.39pm/VforamaximumoutputSHpowerof896µW.Thisvalueisclosetothecalculatedvalueofd 10.7pm/VforthefirstorderSHG.

 

37  

6.2.3Domainmergingpointestimation

Thefabricateddomainswiththe2Darrayofsiliconelectrodeinthe5µmperiodsaretwo‐dimensional on the c faceof the crystal (where the electrode is in contact) andone‐dimensionalon thec polar face.Therefore,2Ddomainstructuresmergedalongtheb‐axisduringthepropagationandbecame1Donthenon‐contactfaceofthecrystal.Inordertofindtheexactpositionofthedomainmergingalongthepolardirectioninthecrystal,anSHGexperimentwasconducted inoneof thePPRKTPcrystals.Theexperi‐mentalsetupisillustratedinFig.(6.8)andwasperformedusingtheCWTi:Sapphirela‐serlasingatawavelengthof894nm.BasedontheconceptwhichisexplainedinSection2.7, the generated SH signal with this structure includes side peakswhich representmorethanonepossiblek‐vectorina2DQPMstructureclosetothec face.WhentheTi:Sapphire laser beamwas scanned over the crystal thickness, the side peaks disap‐pearedassoonas the2Dstructurebecame1D.The2Ddomainstructuremergedandbecamea1Ddittoatadistanceof0.21mmfromthec face.

 

38  

7ConclusionsIn thiswork,anewmethod for fabricationofperiodically‐poledcrystals ispresented.Thismethod is based on using an array of silicon spikes as an electrode for contactwhenperforming electric‐fieldpolingofRKTP.Theprocess for fabricatingthesiliconspikes is designed and subsequently optimized for different electrode shapes andformes.Thefinalmethodisflexible,repeatable,andeasytomodifyforproducingnewtypes of electrodeswith different sizes and geometries. The periodicity of the spikesranges from3.2 µmup to 20 µm. The process of producing spike array electrodes isoverviewedindetail. The constructed Si spike arrays are used as a 2D contact electrode for poling the

RKTPwithtwodifferentperiodicitiesof20µmand5µm,respectively.Eachindividualsiliconelectrodeonthefinalsiliconarrayofelectrodeshasaconicalshapeinthe20µmperiodcaseandaplateaushapeinthe5µmperiodcase.Theconical‐shapedelectrodeshavevery sharp spike tips resulting in apoorly controlledpressure in thepoling cellemployed. In the 20 µm period, the fabricated domains show pattern broadening.Nonetheless, the broadening can be reduced bymaskmodification and by improvingthe poling cell. The plateau‐shaped electrodes produce domain structures which aresimilartothefabricateddomainsusingconventionaltechnique.Thedomainbroadeningfor the 5 µm period case has the same scale as that of the broadening with metal‐depositedelectrodes.However, it canbe reducedby improving thepolingcell andbycontrolling the pressure between theRKTP crystal and the silicon electrode. The do‐mainbroadeningisalsoaffectedbytheelectrodegeometryandcanbeadjustedbymod‐ifying the geometryof theplateau fordifferentperiods.Moreover, the fabricateddo‐mainsforthisperiodarereproducedbyapplyingthesameparametersfortwodifferentsamples.Tothebestofourknowledge,thisisthefirsttimethistechniquehasbeenusedforfabricatingperiodicallypoledRKTP(PPRKTP). Theduty cycleof the fabricateddomains for the5µmperiods is close to the ideal

caseof50%.Therefore,theopticalperformanceofthePPRKTPcrystalswiththisperi‐odwas further evaluated.Two samplespoledwith5µmperiodareused to generatesecond‐harmonic (SH) signals and to investigate the conversion efficiency of the SHprocess for fabricatingQPMdevices. Thenormalized conversionefficiencies for thesetwo samples are higher than 1%/Wcm and, thus, this indicates good optical perfor‐mance. The evaluated effective grating length from the measured temperature ac‐ceptancebandwidth is5.4mm,which is close to thephysicalgrating lengthof6mm.Thus, the fabricateddomainsareuniformlystructureddemonstratingthehigh‐qualitypolingobtainedwiththistechnique.

 

39  

7.1Furtherdevelopments

Thereisavarietyofinterestingfutureexperimentsthatcanbedonebasedonthiswork.Theycanbedivided into twomaincategories:1)modificationof thesiliconelectrodestructureand2)moredetailedinvestigationofthepolingprocess.

Itwasobservedthatbyusingsharpspikes for thepoling, resulted innon‐uniformcontactandnon‐homogenousring‐shapeddomainstructures.Therefore,inthecaseofsilicon electrode modification, first, it is beneficial to polish the spike electrode andthereby reduce the sharpness. Second, the spikes and the plateaus can be fabricatedwithalargerheightbyintroducingtheBoschprocess.Anarrayofsiliconspikesorplat‐eauswithlargeheightcanbepartiallycoveredwithalayerofphotoresistresultinginabetterisolationduringthepolingprocess.Third,inordertocontrolthedomainbroad‐ening,thegeometryofthesiliconspikeandplateauelectrodescanbemodified.Reduc‐ing theelectrodedimension in theb‐directioncouldresult in lessdomainbroadeningleadingtoacomplete2Ddomainstructureonboththepolarfacesofthecrystal.Forth,theelectrodematerialcanbechangedtoanalyzetheinfluenceofdifferentmaterialsonpoling.Thep‐type siliconelectrode canbe replacedbyann‐type siliconormetalizedpolymerspikes.Thefabricationofpolymerspikesisflexibleandwelldeveloped.There‐fore, a metalized polymer array of spikes can serve as a good alternative electrodechoicetothesiliconspikes.Furthermore,modifiationoftheelectrodestructurecanbedoneinordertoinvestigatethe1DperiodicpolingofKTP.Inordertoaccessthistech‐niqueforfabricatingdomainsinthenano‐scale,siliconelectrodespecificationsneedtobemodified.

Regardingthepolingprocessitself,thefollowingaspetscanbeimprovedupon.Thepolingcellshouldbeupgradedsoastoprovideabetterpressureandalignmentcontrol.Thepolingparameterssuchasthepulseshape,pulsedurationandthenumberofpulsescanbeoptimized.Thus,periodically‐poledcrystalswithhigherqualitycanbeobtained.

 

40  

8References

[1] G.Rosenman,D.Eger,andM.Katz,“Domainbroadeninginquasi‐phase‐matchednonlinearopticaldevices,”AppliedPhysicsLetters,vol.73,no.7,p.865,1998.

[2] K. Kiantaka, M. Fujimura, T. Suhara, and H. Nishihara, “High‐Efficiency LiNbO3Waveguide Second‐Harmonic Generation devices with Ferroelectric‐Domain‐InvertedGratingsFabricatebyApplyingVoltage,”LightWaveTechnology,vol.14,no.3,1996.

[3] C. Canalias, V. Pasiskevicius, M. Fokine, and F. Laurell, “Backward quasi‐phase‐matchedsecond‐harmonicgeneration in submicrometerperiodicallypoled flux‐grownKTiOPO4,”AppliedPhysicsLetters,vol.86,no.18,p.181105,2005.

[4] A.Zukauskas,G.Strömqvist,V.Pasiskevicius,F.Laurell,M.Fokine,andC.Canalias,“Fabricationof submicrometerquasi‐phase‐matcheddevices inKTPandRKTP,”OpticalMaterialsExpress,vol.1,no.7,p.1319,Oct.2011.

[5] C.CanaliasandV.Pasiskevicius,“Mirrorlessopticalparametricoscillator,”NaturePhotonics,vol.1,no.8,pp.459–462,Aug.2007.

[6] C.Canalias,V.Pasiskevicius,R.Clemens,andF.Laurell, “Submicronperiodicallypoledflux‐grownKTiOPO4,”AppliedPhysicsLetters,vol.82,no.24,p.4233,2003.

[7] H.Kintaka,K,Fujimura,M,Suhara,TandNishihara,“Fabricationofferroelectric‐domain‐invertedgrating inLiNbO3byapplyingvoltageusingetched‐Sistamperelectrode,”vol.34,no.9,pp.2–3,1998.

[8] P. A. Franken, A. E. Hill, C.W. Peters, and G.Weinreich, “Generation of opticalharmonics,”PhysicalReviewLetters,vol.7,no.1,pp.118–120,1961.

[9] F.Trägerand...,SpringerHandbookofLasersandOptics.2007,pp.157–248.

[10] S.BahaaE.A.,FundametalofPhotonics,no.born1920.1981,pp.773–817.

[11] R. C. Miller, “Optical second harmonic generation in piezoelectric crystals,”AppliedPhysicsLetters,vol.5,p.17,1964.

[12] R.W.Boyd,NonlinearOptics,Third.AP,2007,pp.9–17.

[13] L.E.Myers,R.C.Eckardt,M.M.Fejer,R.L.Byer,W.R.Bosenberg,andJ.W.Pierce,“Quasi‐phase‐matched optical parametric oscillators in bulk periodically poledLiNbO_3,”JournaloftheOpticalSocietyofAmericaB,vol.12,no.11,p.2102,Nov.1995.

[14] F. Laurell, “Periodically poled materials for miniature light sources,” OpticalMaterials,vol.11,no.January,pp.235–244,1999.

 

41  

[15] H. Suchowski, B. D. Burner, A. Arie, and Y. Silberberg, “Broadband NonlinearFrequencyConversion,”OpticandPhotonicNews,pp.36–41,2010.

[16] M.M.Fejer,G.a.Magel,D.H.Jundt,andR.L.Byer,“Quasi‐phase‐matchedsecondharmonic generation: tuning and tolerances,” IEEE Journal of QuantumElectronics,vol.28,no.11,pp.2631–2654,1992.

[17] Y. Lu, Y. Lu, X. Cheng, G. Luo, C. Xue, and N. Ming, “Formation mechanism forferroelectricdomainstructuresinaLiNbO3opticalsuperlattice,”AppliedPhysicsLetters,vol.68,no.19,p.2642,1996.

[18] J.D.Bierlein,D.B.Laubacher,J.B.Brown,andC.J.vanderPoel,“BalancedphasematchinginsegmentedKTiOPO4waveguides,”AppliedPhysicsLetters,vol.56,no.18,p.1725,1990.

[19] M.C.Gupta,W.P.Risk,A.C.G.Nutt,andS.D.Lau,“DomaininversioninKTiOPO4using electron beam scanning,” Applied Physics Letters, vol. 63, no. 9, p. 1167,1993.

[20] H.KarlssonandF.Laurell,“ElectricfieldpolingoffluxgrownKTiOPO4,”AppliedPhysicsLetters,vol.71,no.24,p.3474,1997.

[21] V.Berger, “NonlinearPhotonicCrystals,”PhysicalReviewLetters, vol.81,no.19,pp.4136–4139,Nov.1998.

[22] C.Canalias,“DomainEngineeringinKTP,”RoyalInstituteofTechnology,2005.

[23] F. C. Zumsteg, J. D. Bierlein, and T. E. Gier, “KxRb1−xTiOPO4: A new nonlinearopticalmaterial,”AppliedPhysics,vol.47,p.4980,1976.

[24] C.Canalias,DomainEngineeringin.2005.

[25] S. Wang, “Fabrication and characterization of periodically‐poled KTP and Rb‐dopedKTPforapplicationsinthevisibleandUV,”RoyalInstituteofTechnology,2005.

[26] J. D. Bierlein Company and F. Ahmed, “Observation and poling of ferroelectricdomainsinKTiOPO4,”AppliedPhysicsLetters,vol.51,no.17,p.1322,1987.

[27] Q.Jiang,P.A.Thomas,K.B.Hutton,andR.C.C.Ward,“Rb‐dopedpotassiumtitanylphosphateforperiodicferroelectricdomaininversion,”JournalofAppliedPhysics,vol.92,no.5,p.2717,2002.

[28] a. Fragemann, V. Pasiskevicius, J. Nordborg, J. Hellstrom, H. Karlsson, and F.Laurell, “Frequency converters from visible to mid‐infrared with periodicallypoledRbTiOPO[sub4],”AppliedPhysicsLetters,vol.83,no.15,p.3090,2003.

 

42  

[29] S. Wang, V. Pasiskevicius, and F. Laurell, “High‐efficiency frequency converterswithperiodically‐poledRb‐dopedKTiOPO4,”OpticalMaterials,vol.30,no.4,pp.594–599,Dec.2007.

[30] C. Canalias, S.Wang,V. Pasiskevicius, andF. Laurell, “Nucleation andgrowthofperiodic domains during electric field poling in flux‐grown KTiOPO[sub 4]observed by atomic forcemicroscopy,”AppliedPhysicsLetters, vol. 88, no. 3, p.032905,2006.

[31] C. Canalias, J.Hirohashi, V. Pasiskevicius, andF. Laurell, “Polarization‐switchingcharacteristics of flux‐grown KTiOPO4 and RbTiOPO4 at room temperature,”JournalofAppliedPhysics,vol.97,no.12,p.124105,2005.

[32] N. Roxhed, “A fully integrated microneedle‐based transdermal drug deliverysystem,”RoyalInstituteofTechnology,2007.

[33] W.Trimmer,P.Ling,C.Chin,P.Orton,B.Brunett,M.Reed,R.Gaugler,S.Hashmi,and G. Hashmi, “Injection of DNA into plant and animal Tissues withmicromechanicalpiercingstructures,”IEEE,vol.2,pp.111–115,1995.

[34] D. .McAllister, F. Cros, S. P.Davis, L.M.Matta,M.R. Prausnitz, andM.G.Allen,“Three‐dimensionalhollowmicroneedle andmicrotubearrays,” inThe10th Int.Conf.onSolid‐stateSensorsandActuators,1999,pp.1098–101.

[35] S. Henry, D. V. McAllister, M. G. Allen, and M. R. Prausnitz, “Microfabricatedmicroneedles: A novel approach to transdermal drug delivery,” Journal ofpharmaceuticalsciences,vol.88,no.9,p.948,Sep.1999.

[36] B.StoeberandD.Liepmann,“Arraysofhollowout‐of‐planemicroneedlesfordrugdelivery,” JournalofMicroelectromechanicalSystems,vol.14,no.3,pp.472–479,Jun.2005.

[37] N. Roxhed, T. C. Gasser, P. Griss, G. A. Holzapfel, and G. Stemme, “Penetration‐Enhanced Ultrasharp Microneedles and Prediction on Skin Interaction forEfficientTransdermalDrugDelivery,”JournalofMicroelectromechanicalSystems,vol.16,no.6,pp.1429–1440,Dec.2007.

[38] J.D.Plummer,M.D.Deal,andP.B.Griffin,SiliconVLSITechnology.PrenticeHall,2000.

[39] H.Karlsson,F.Laurell, andL.K.Cheng, “PeriodicpolingofRbTiOPO4 forquasi‐phasematchedblue light generation,”AppliedPhysicsLetters, vol. 74, no. 11, p.1519,1999.

[40] F. Laurell, M. G. Roelofs, W. Bindloss, H. Hsiung, A. Suna, and J. D. Bierlein,“Detection of ferroelectric domain reversal in KTiOP4 waveguides,” Journal ofAppliedPhysics,vol.71,no.10,p.4664,1992.

 

43  

[41] G.D.Boyd,“ParametricInteractionofFocusedGaussianLightBeams,”JournalofAppliedPhysics,vol.39,no.8,p.3597,1968.

[42] W.Wiechmann,S.Kubota,T.Fukui,andH.Masuda,“Refractive‐indextemperaturederivatives of potassium titanyl phosphate.,” Optics letters, vol. 18, no. 15, pp.1208–10,Aug.1993.

[43] H.VanherzeeleandJ.D.Bierlein,“Magnitudeofthenonlinear‐opticalcoefficientsofKTiOPO4.,”Opticsletters,vol.17,no.14,pp.982–4,Jul.1992.