CFC3 2005 Conference on Lasers & Electro-Optics (CLEO)
Thermally
poled
germanosilicate
films
with
high
second-order
nonlinearity
A.Ozcan, M. J. F.Digonnet, G. S. Kino
Edward L. GinztonLaboratory; Stanford University,Stanford, California94305 Phone:(650) 723-0251Fax:(650) 725-2533 Email: aozcan(distanfordedu
F.Ay, and A. Aydinli
DepartmentofPhysics, Bilkent University, 06800, Ankara,Turkey
Phone: 90(312)290-1972Email:ay@,fen.bilkent.edu.tr
Abstract: Accurate measurements of the second-order
nonlinearity profile
ofthermally poled
low-lossgernanosilicate films grown on fused-silica substrates are
reported,
of interest aspotential electro-optic
devices. Afteroptimization,
wedemonstratearecordhigh
nonlinear coefficientd33
1.6pm/V,
atwo-foldimprovement
over highestreported d33 value in fused silica that we attributetothe presence ofgermanium.©2005 OpticalSociety ofAmerica
OCIS codes:(310.6860)Thinfilms, optical properties; (190.4400)Nonlinearoptics, materials 1.Introduction
Thermalpoling is one ofthemostreproducible andwidely used techniquestoinduceastablesecond-orderoptical nonlinearity in glass. Some of the important
potential
applications
ofthernally poled glass
include integratedelectro-optic
phase andamplitude
modulators orparametric
oscillators.[1]
Poledglass exhibits
significantadvantages compared to existingnonlinear materials, including low loss, broad transmission bands, high optical
damage threshold, and compatibility with fiber technology. The largest reported peak nonlinear coefficient for thermally poled silicaglass is d33 z 0.8pm/V.[2] As a resultof this relatively weaknonlinearity, allpoled-glass
devicesreported to daterequirefairlyhigh voltages and/or longlengths.Forthistechnologytobecome
practical,
thestrength of theinducednonlinearity hastobeimproved.
Inthisletter,weshow that thenonlinear coefficient can be doubled by doping the glass with germanium. After
optimizingthematerialcompositionandpoling
conditions,
wereportarecordpeakd33 coefficientof -1.6pm/Vinthermally poled germanosilicate films. These films are of interestprimarily for two important reasons: (1) their propagation loss can be verylow,[3] whichmakes themexcellent waveguidematerialswith a refractive index close tothatof silica, and (2)the addition of Ge tothe silicamatrix increases the third-order
optical susceptibility
J
3) ofthe glass, which should result in an increase in the induced nonlinear
coefficient.
These properties make poled germanosilicatefilmsapromisingcandidate forfuturelow-lossintegrated electro-opticdevices.Table1.Characteristics of different germanosilicate filmspoled in air at -5 kV and -280 'C.
Germaneflow Molefractionof X(G:SiO2 Refractiveindex Peakd33
SampleI rate GeO,7(%) X(3) at1064nn Thickness Poling time m/)
I Osccm 0 -1 1.469 4
gm
10mm 0.54 2 33 scem -20 1.54 1.497 4pm 5min 0.80 3 33sccm -20 1.54 1.497 4pm 10min 1.59 4 33sccm -20 1.54 1.497 4jm
15min 1.00 5 33sccm -20 1.54 1.497 2pm 10min 1.02 6 50sccm -30 2.22 1.514 4pm 10min 0.78 7 90sccm -56 4.77 1.553 4pm 10min 0.812.Germanosilicatethin filmgrowthprocess
Germanosilicatefilms were
deposited
on fused-quartzsubstrates(Infrasil,
25x25x0.15mm) by
plasma-enhancedchemicalvapor
deposition (PECVD) using
aparallel-plate
reactor(Plasmalab
8510C).
The filmsweregrownat350°C
andapressureof 1 TorratanRFpowerof 10Wat 13.56MHzapplied
to theplates.
Theprecursor gaseswereCFC3 2005ConferenceonLasers &Electro-Optics(CLEO)
silane (2%Sil{4/N2), germane (2%GeH4/He), andnitrous oxide (N20).The flow rates of silane and nitrous oxide werekept constantat180and 225 sccm,respectively, while that of germane waschanged from run to run between0 and 90 sccm. The growth rate of the films was-40nnm/min.
Annealing is almost always required to reduce thepropagation loss ofthe optical waveguides that utilize CVD-grownsilicon-based layers asthe core. Recently, we have reported the lowestpropagation loss values for as-grown germanosilicate films without the need for thermal annealing.[3] Inorder tooptimnize the poling process for this material, we grew seven individual germanosilicate filmswithdifferentcharacteristics, as listed in Table 1. Based on ourprevious
work,[3]
the propagation losses of the as-grown waveguides were estimated to be lessthan 0.15 dB/cmat1550 mn.i #4 I51M.2',; S4I Kk o t".2,
0. ri"E1. rt#$i
n.4 l. ~~~~ ~~~~~0.4~ A
a 2 s8 Islo is 14iS is O 2 4 8 8 i 200 t4 iS i o0 2 4 S 8 t0 12 14 18 1S
Depl:lVicn1.hs31n1cbaiu s(Pm Oap[hfnm lte Eim aukm (W DepM hirn Ehs anok suftea 8xn)
Fig. 1.Therecovered nonlinearity profiles of(a) samples #2, #3, and #4; (b) ) samples #1, #3, #6and#7; (c)samples #3and#5. 3.Thermal poling results
The grown germanosilicate-Infrasil structures were thermally poled using polished n-type silicon electrodes (the
positive electrode facingagainst thegermanosilicatelayer)in airat -5kVand 280 'C. Thepolingtimewasvaried between 5 and 15 min (see Table 1) to investigate its effect on the peak nonlinear coefficient and maximize this coefficient. The nonlinearity depth profile of each poled sample was then measuredby the Maker fringe-Fienup
technique.[2] Afundamental laser beam at 1064nm was focused onto the sample and the second-harmonic (SH) power generated within thepoled region was recorded as a function of the beam incidence
angle.[4]
In orderto avoidtotal internal reflection at the output face of the Infrasil substrate, and henceachieve high incidence angles (e.g.>85°) a pair of half-cylinders made of Infrasil wasclamped on both sides ofthe poled samples.[5] Theresulting Maker fringe (MF) curve was then correctedfirstforangle-dependent multiplereflectionsarising in the film from the refractive index mismatch between the film and the substrate, and second for Fresnel reflections at both the fundamental and the SH wavelengthsoccurring at the boundary betweenthe inputhalf-cylinder and the film andat the film-substrate boundary. Finally, the corrected MF curves were processed using an iterative Fouriertransforntechnique[2]
touniquely retrieve the induced nonlinearity depth profile d33(z),where z is the depth into the sample measured from the anode surface. For want of space,themeasured MF curves of the samples are not shown here. Therecoverednonlinearity profiles forthe seven poled samples of Table 1 are shown in Fig. 1. The results in Fig. 1 aregrouped according to the polingconditionsandfilmproperties. Figure1(a)
showstheprofiles of samples #2, #3, and #4, all ofwhich have a4-pm
thickgermanosilicate films grown at a 33-sccmgermane flow rate. Toidentifythe optimum polingtime,thesenominally identical sampleswerepoled under identicalconditions except for thepoling time,which was 5 min for sample #2, 10 min for sample #3, and 15 minfor sample#4. As showninFig.1(a),
the profiles recovered for these three samples exhibit similar features, namely asharppeakcentered about 0.5Am
below the anode, followed by a weak pedestal that is approximately constant to a depth of -9-12 jimand that graduallydecreases to zero at a depth of 13-16
gim.
This resultreveals that the optimum poling time for these germanosilicate-Infrasil structures at an applied E-field of -32.5 MV/m is -10 min, corresponding to sample #3. The peak d33 coefficient obtained under these poling conditions is as high as -1.6pm/V. To ourknowledge, this is the highest second-order optical nonlinear coefficient measured without any ambiguities in athermallypoled germanosilicateglass. The peak d33 coefficients measured for poling times of 5 and 15 min are -0.8
pm/V
and -1.0 pm/V,respectively. As physically expected, as the poling time is increased from 5 min to 15 min the depth of thepedestal gradually increases from -9
gm
to -12,Lm
(see Fig. l(a)). Thetotal depth of the induced nonlinear region (-13-16jim)
inFig.1(a)
issignificantlynarrowerthan for bulk Infrasil samplesthermallypoled undersimilarconditions,
for which the depth is typically -40pm.[2]
Furthermore, unlike in thermally poled Infrasil, the d33(z) profile of theCFC3 2005Conference on Lasers& Electro-Optics (CLEO)
poled
germanosilicate-Infrasil
structuresdoes notchange
sign. We believe that the reasonfor these differences isthat the
germanosilicate
film limits the diffusion ofpositive
ions such asH30
from the anode surface into thesample,
which results in theformationof a narrowerdepletion region
within the film and hence a shalloweroverall nonlinearregion.
A similarblocking
behavior ingermanosilicate
films,
which also resulted in narrowernonlinearwidths,
has beenreported
by
others.[6]
Afteroptimizingthe
poling
time,
weinvestigated
theeffectof thegermaneflowrate ontheinducednonlinearity
profile.
Forthispurpose,wepoled
samples #1,
#3,
#6,
and#7,which all havea4-rim thickgermanosilicate
film butwere grown at different germane flow rates,
namely 0, 33,
50 and 90 sccm,respectively,
so that their Ge concentrationsweredifferent(see
Table1).
Allfoursamples
werepoled
underidenticalconditions,
i.e.,
in airat-5 kV and 280 °C, for 10 min. The recoverednonlinearity profiles
for thesesamples
are shown inFig.
1(b).
Theprofiles
exhibitthesame overallshape
astheprevious
samples (see Fig. 1(a)).
Thepeak
d33
coefficients ofsamples #1, #6 and #7 are 0.54, 0.78 and 0.81 pm/V,respectively.
Thisinvestigation
shows that thehighest
peak d33coefficient
(1.6
pm/V)
isachieved foragermaneflow of 33 sccm(sample
#3). However,
ahigher
germaneflowrateproduces
ahigher
Geconcentrationand thusahigher
,j3)
(seefourth column of Table1,
wherethe listed23)valueswere calculated fromthemeasured
dispersion
curves ofthesamples),
sobasedonthisargumentalone we expect thatthehighest peak d33
shouldoccur atthehighest
flowrate.The factthatthepeak
d33
coeffilcient is maximum inthe 33-sccm
sample
suggests that the built-in fielddrops
athigher
Ge flow rates. Wemostly
relate these observations to an increase in the film electricalconductivity
as the Ge concentrationincreased,
which has beenpreviously
confirmed.[7]
On the otherhand,
thebuilt-in field ofthe 33-sccmsample
ishigher
than thatofthe0-sccm
sample (pure
SiOi),
although
thelatter hasalowerelectricalconductivity.
Thispoints
outthatthere shouldbean
optimum
electricalconductivity
range foragiven setofpoling
conditions. Thishypothesis
is supported by the observationthatundersimilarpoling
conditions, Suprasil,
which contains much lessimpurity
thanInfrasil and thus hasalowerconductivity,
develops
abuilt-in fieldnearly
oneorder ofmagnitudelower thanInfrasil.[1]Finally,
weinvestigatedtheeffect of the film thickness onthe inducednonlinearity
profile. Forthispurpose, wepoled sample #5,
grownat a33-sccm germaneflowrate toafinalthickness of2gm,
for 10min.Figure
l(c) showsthe
nonlinearity profile
recovered forthissample.
Forcomparison,
thenonlinearity
profile
ofsample #3, whichwas grownatthesame flowrate andpoled
underidentical conditions but is thicker(4pm),
is alsoshown inFig. 1(c). The twosamples
have very similarprofiles,
which wasexpected
sincethey
havethe samecomposition
andwerepoled
under thesameconditions.However,the totaldepth
ofthenonlinearity
islarger
forsample#5(A17
gm)
than forsample
#3('13
pm).
Aprobable
explanation
for thisdifferenceis that thethinner2-Am
germanosilicate film
insample
#5 acts asaweakerbarrierfor ion diffusionthanthe4-pm
film insample
#3. Furthermore,webelievethat thischarge spreading
is attheorigin
of the weakerpeak
d33
coefficient in sample#5 (1.02pm/V
vs. 1.6pm/V
insample #3).
4.
Summary
We have
reported
adetailedstudy
of thenonlinearity
profileofthermalpoledgermanosilicate films grown onfused-silica substrates
by
PECVD. Thisstudy
sheds a betterunderstanding
on the physics of thermal poling ingermanosilicate
films. Inferredprofiles
all exhibitasharp
peak-0.5gm
beneaththe anode surface,followed by a weakerpedestal
ofroughly
constantamplitude
downtoadepth
of13-16pm. Comparedtothermally
poled undopedsilica,
they
are shallower and do notexhibit
asign
reversal, which indicates that the germanosilicate filmsignificantly
slows down theinjection
ofpositiveions from air into theglass.Afteroptimizing the Ge concentration,thefilm
thickness,
andthepoling
time,
weobtainedarecordpeak d33
coefficient of 1.6pmn/V
inthe sample grown ata33-sccm germane flowrate(-20
mole%
GeO2) and poled for 10min. Combinedwith the lowpropagation
loss of thesefilms,
thisenhancednonlinearity
makespoledgermanosilicate filmsapromising candidate for planarelectro-optic
devices.[1]Y.Quiquempois,P.Niay,M.Douay,andB.Poumellec,"Advances inpolingandpermanently induced phenomena in silica-basedglasses,'
CurrentOpinioninSolid State & Materials Science7,89-95(2003)
[2]A.Ozean,M.J. F.Digonnet,andG. S.Kino,"Iterativeprocessingofsecond-orderopticalnonlinearitydepth profiles," Opt. Express 12, 3367-3376(2004),http://www.opticsexpress.org/abstract.cfrn?URI=OPEX-12-15-3367
[3]F.Ay,A.Aydinli,andS.Agan"Low-lossas-growngermanosilicatelayers foroptical waveguides," Appl. Phys. Lett. 83, 4743-4745(2003)
[4]P. D.Maker,R.W.Terhune,M.Nisenhoff,andC. M.Savage,"Effectsofdispersionandfocusing on production of opticalharmonics,"Phys. Rev. Lett.8,21-22(1962)
[5]A.Ozcan, M. J. F.Digonnet,andG.S. Kino,"Cylinder-assistedMaker-fringe technique," Electron. Lett. 39, 1834-1836 (2003)
[6]D.Faccio,A.Busacca,D.W. J.Harwood,G.Bonfrate,V.Pruneri,and P.G.Kazansky, "Effect of core-cladding interface on thermal poling ofgermanosilicate optical waveguides,"Opt.Comm.196,187-190(2001)
[7]R.T.Crosswell,A.Reisman,D.L.Simpson,D.Temple,andC. K.Williams,"Planarizationprocessesand applications:III.As-deposited and annealed filmproperties,"J.Electrochem. Soc.147,1513-1524(2000)