+ All Categories
Home > Documents > Submicron SOI waveguides - UGentphotonics.intec.ugent.be/download/ocs63.pdf · 2005. 3. 11. ·...

Submicron SOI waveguides - UGentphotonics.intec.ugent.be/download/ocs63.pdf · 2005. 3. 11. ·...

Date post: 27-Jan-2021
Category:
Upload: others
View: 1 times
Download: 0 times
Share this document with a friend
91
http://photonics.intec.UGent.be Submicron SOI waveguides Dries Van Thourhout Trento ‘05
Transcript
  • http://photonics.intec.UGent.be

    Submicron SOI waveguides

    Dries Van ThourhoutTrento ‘05

  • http://photonics.intec.UGent.be 2© intec 2004

    AcknowledgementsThe European Union IST-PICCO and IST-PICMOS projectThe European Space AgencyThe Belgian IAP-PHOTON networkThe Flemish Institute for the industrial advancement of Scientific and Technological Research (IWT)

    The Photonic Research Group at Ghent University – IMECPieter Dumon, Wim Bogaerts, Dries Van Thourhout, Dirk Taillaert, Bert Luyssaert, Peter Bienstman, Joris Van Campenhout, Gunther Roelkens, Ilse ChristiaensThe Silicon Process division at IMECVincent Wiaux, Stephan Beckx, Johan Wouters, DizianaVangoidsenhoven, Rudi De Ruyter, Johan Mees

  • http://photonics.intec.UGent.be 3© intec 2004

    OutlineSubmicron SOI-wires

    Introduction: why do we need themBasic properties: design, loss, wavelength, polarizationFabricationDevices: couplers, crossings, filters …

    III-V on SiliconIntroductionCoupling of lightFabricationPICMOS (Photonic Interconnect on CMOS)

  • http://photonics.intec.UGent.be 4© intec 2004

    Scale differenceElectronics

    Active opto-electronics

    Passive photonics

    1cm1mm100µm10µm1µm100nm

    AWG in Silica on SiliconBend radius

    linewidth in current PIC

    VCSELstripe laserLED

    detector

    gatewidth

    transistor

    taperspot-sizeconvertor

    2R regenerator

    fibre core

    flip-flop

    Wavelength-scale photonics

    interconnects

    Wavelength-scale photonics

  • http://photonics.intec.UGent.be 5© intec 2004

    PICs: today and futureToday (InP, Silica-on-Silicon...):

    Size of components on a chip (both functional components and interconnect components):

    103 - 106 µm2

    Number of components on a chip:1 - 103

    Future (10-20 years from now):Size of components on a chip (both functional components and interconnect components):

    1 - 103 µm2

    Number of components on a chip:103 - 106

  • http://photonics.intec.UGent.be 6© intec 2004

    Silica-on-silicon

    NTT (e.g. Miya e.a., IEEE STQE ’00 pp.38)

  • http://photonics.intec.UGent.be 7© intec 2004

    Reduce PIC-size / increase density WE NEED:Ultra-compact waveguiding with

    Sharp bends (Bend radius < 10µm)

    Compact splitters and combiners

    Short mode-conversion distances

    Compact wavelength selective functionsHighly dispersive element

    Small, high-Q resonators

    Compact non-linear functionsIncrease power density by using tight confinement

  • http://photonics.intec.UGent.be 8© intec 2004

    High refractive index contrast (>2:1)High refractive index contrast allows for:

    Very tight bendsCompact resonators with low lossWide angle mirrorsVery compact mode size

    strong field strength strong non-linear effectssmall volume to be pumped in active devices faster and/or lower power

    Photonic band gap effects

    air semiconductor

    dielectric

    High refractive index contrast is key for ultra-compact photonic circuits

  • http://photonics.intec.UGent.be 9© intec 2004

    Index ContrastConventional PICs Nanophotonics

    In-plane (effective) index contrast

    Out

    -of-p

    lane

    inde

    x co

    ntra

    st

    Low

    Low

    High

    High

  • http://photonics.intec.UGent.be 10© intec 2004

    Materials for nanophotonic waveguides

    In-plane indexcontrast

    Out-of-planeindex contrast

    Si/SiO2 (SOI) 3.5 to 1 3.5 to 1.5

    Si/air(membrane)

    3.5 to 1 3.5 to 1

    GaAs/AlOx 3.5 to 1 3.5 to 1.5

    InP/SiO2 3.3 to 1 3.3 to 1.5

    SiON/SiO2 2 to 1.5/1 2 to 1.5

    GaAs/AlGaAs 3.5 to 1 3.5 to 3.2

    InGaAsP/InP 3.3 to 1 3.3 to 3.17

  • http://photonics.intec.UGent.be 11© intec 2004

    SOI nanophotonics

    Start: SOI-Wafer• Thin Silicon layer• Thick SiO2 buffer

    BOX thickness

    Waveguide Definition

    Width + Height

  • http://photonics.intec.UGent.be 12© intec 2004

    SOI-wires

    EBeamyes1110.0 600260Oct. 03Columbia5.0 500200

    DUVyes115.0 300300Apr. '05LETI / LPM+oxidation0.8 20050G-lineyes132.0 500200Dec. '01MITEBeamno1105.0 400320Dec. '02YokohamaEBeamno36.0 300300Dec. '02NTTEBeamno35.0 470270Aug. '03CornellEBeamno23.6 445220Apr. '04IBMDUVno12.4 500220Apr. '04IMEC

    Fab.top clad

    BOX [um]

    loss [dB/cm]

    w [nm]

    h [nm]

    DateGroup

    (Table partly from Vlasov, McNab, Opt. Expr. ’04, pp1630)

  • http://photonics.intec.UGent.be 13© intec 2004

    OutlineSubmicron SOI-wires

    Introduction: why do we need themBasic properties: theory, design, loss, wavelength, polarizationFabricationDevices: couplers, crossings, filters …

    III-V on SiliconIntroductionCoupling of lightFabricationPICMOS (Photonic Interconnect on CMOS)

  • http://photonics.intec.UGent.be 14© intec 2004

    Back to basics: cavities and waveguides

    How does light propagate in waveguides and cavities?

    n1

    n2

    n2

    n1

    n2

    n2

    n1

    n2

    n2

    n1

    n2

    n2

    Propagation in waveguide

    Propagation through cavity

    Emission within waveguide/cavity

    Propagation through waveguide discontinuity

    What is the role of n2 /n1 ?

  • http://photonics.intec.UGent.be 15© intec 2004

    Back to basics: cavities and waveguides

    kx

    kz

    TotalInternalReflection

    Radiation modes

    n2k0 n1k0

    d n1

    n2

    n2

    k

    kx

    kz

    • dispersion:

    • continuity:

    For slab waveguide: n2 /n1 0.99 0.9 0.5Fraction of (2D) k-space confined by TIR = 14% 44% 87%

    For channel waveguide:Fraction of (3D) k-space confined by TIR = 2% 19% 75%

    “Light line”

    220 yx kkc

    nnkk +=== ω

    2,1, xx kk =

    2

    1

    21 ⎟⎠

    ⎞⎜⎝

    ⎛−

    nn

    2

    1

    21 ⎟⎠

    ⎞⎜⎝

    ⎛−

    nn

  • http://photonics.intec.UGent.be 16© intec 2004

    Back to basics: cavities and waveguides

    kx

    kz

    π/d

    Resonance if ↓

    d large → many resonancesd small → few resonances

    Two types of resonances:

    • guided modes confined by TIR

    • resonantly enhanced radiation modes

    d n1

    n2

    n2

    k

    kx

    kz

    Only one guided mode if

    dmkz

    π=

    ( )2121

    12

    nn

    nd−

    ≤λ

  • http://photonics.intec.UGent.be 17© intec 2004

    Back to basics: cavities and waveguides

    Waveguide/cavity types:• conventional waveguide: n2 /n1 ≈ 0.9 - 1

    single mode even for d substantially larger than λ/n1lots of radiation modes

    only small fraction of k-space well controlled by TIR

    hence bends, couplers need to be based onslow adiabatic transitions → longinterference with long coupling length between guided modes → long

  • http://photonics.intec.UGent.be 18© intec 2004

    Back to basics: cavities and waveguides

    Waveguide/cavity types:• high contrast waveguide: n2 /n1

  • http://photonics.intec.UGent.be 19© intec 2004

    Back to basics: periodic stacks

    n1

    n1

    n1

    n2

    n2

    Λ

    kx

    kz

    n2k0 n1k0

    Strong reflection and little transmission if:(for normal incidence)

    (Bragg condition)

    2π/Λd1

    d2

    Λ=

    Λ+

    Λπmdkdk 2211

    Λ+

    Λ=

    =Λ→

    22

    11:

    2dndnnwith

    nm

    av

    av

    λ

  • http://photonics.intec.UGent.be 20© intec 2004

    Back to basics: periodic stacks

    Cases:• low contrast stack: n2 /n1 ≈ 1

    many periods needed for strong reflection

    strong reflection only for narrow angular and spectral range

    • high contrast stack: n2 /n1

  • http://photonics.intec.UGent.be 21© intec 2004

    Spectral accuracy and geometrical accuracy

    High index contrast components:- interference based filters,

    with d the waveguide width (≈λ)

    - cavity resonance wavelengthwith d the cavity length (a few λ)

    - photonic crystalwith d the hole diameter (≈λ)

    dd∂

    ≈∂λλ

    if tolerable wavelength error : 1 nm ⇓

    tolerable length scale error : (of the order of) 1 nm

    dd∂

    ≈∂λλ

    dd∂

    ≈∂λλ

  • http://photonics.intec.UGent.be 22© intec 2004

    Basic PropertiesEffective Index

    1

    1.25

    1.5

    1.75

    2

    2.25

    2.5

    2.75

    300 400 500 600 700 800

    Waveguide Width [nm]

    Effe

    ctiv

    e In

    dex

    TE0

    TE1

    TM0TM0

    TE1

    Single-Mode Width

    Cladding (1.44)

    h=220nm – λ=1550nm – 2D calc

  • http://photonics.intec.UGent.be 23© intec 2004

    1

    1.25

    1.5

    1.75

    2

    2.25

    2.5

    2.75

    300 400 500 600 700 800

    Basic PropertiesEffective Index

    Waveguide Width [nm]

    Effe

    ctiv

    e In

    dex TE0

    TE1

    TM0TM0

    TE1

    Single-Mode Width

    Cladding (1.44)

    h=220nm – λ=1550nm – 2D calc

  • http://photonics.intec.UGent.be 24© intec 2004

    Basic PropertiesGroup Index

    h=220nm – λ=1550nm – 2D calc

    1

    1.5

    2

    2.5

    3

    3.5

    4

    4.5

    5

    1500 1525 1550 1575 1600Wavelength [nm]

    n eff

    -ngr

    oup

    w=400-600

    w=400-600

    1.52

    2.53

    3.54

    4.55

    400 450 500 550 600Waveguide Width [nm]

    νν+=

    λλ−=

    ddnn

    ddnnng Determines filter properties

    ngroup

    neff

  • http://photonics.intec.UGent.be 25© intec 2004

    Substate Leakage

    BOX Buffer Thickness [um]

    Subs

    trat

    e Le

    akag

    e Lo

    ss [d

    B/c

    m]

    w=300nmw=500nm

    1dB/cm

    h=220nm – λ=1550nm

  • http://photonics.intec.UGent.be 26© intec 2004

    Losses

    EBeamyes1110.0 600260Oct. 03Columbia5.0 500200

    DUVyes115.0 300300Apr. '05LETI/ LPM

    +oxidation0.8 20050

    G-lineyes132.0 500200Dec. '01MIT11EBeamno1105.0 400320Dec. '02Yokohama

    EBeamno36.0 300300Dec. '02NTTEBeamno35.0 470270Aug. '03Cornell

    2.5EBeamno23.6 445220Apr. '04IBM< 5DUVno12.4 500220Apr. '04IMEC

    σroughness [nm]

    Fab.top clad

    BOX [um]

    loss [dB/cm]

    w [nm]

    h [nm]

    DateGroup

  • http://photonics.intec.UGent.be 27© intec 2004

    0

    5

    10

    15

    20

    25

    30

    35

    40

    300 350 400 450 500 550

    Wire width (nm)

    Loss

    es (d

    B/c

    m)

    w400nm440nm450nm500nm

    33.89.47.42.4

    Propagation losses± 1.7 dB/cm± 1.8 dB/cm± 0.9 dB/cm± 1.6 dB/cm

    Loss (IMEC)

    22

    2s

    2s n

    dxEE

    ∆σ

    ∝α∫ Refractive index contrast

    Field at interfaceSurface Roughness

    Width [nm]

    Loss

    [dB

    /cm

    ]

    IBM

  • http://photonics.intec.UGent.be 28© intec 2004

    Loss (IBM)2

    2

    2s

    2s n

    dxEE

    ∆σ

    ∝α∫

    Wavelength [nm]

    Loss

    [dB

    /cm

    ]

    3.5dB/cm

    (Vlasov, McNab, Optics Express, ’04)

    w=450nmh=220nm

    TETM

  • http://photonics.intec.UGent.be 29© intec 2004

    Loss - otherOxidationDifference TE/TM

    TM higher substrate leakage

    TM higher scattering at vertical roughness

    TE higher field intensityRoughness Correlation lengthGrillot e.a., PTL ’04, pp. 1661

    (MIT)(Grillot)

  • http://photonics.intec.UGent.be 30© intec 2004

    Polarisation

    EBeamyes1110.0 600260Oct. 03Columbia5.0 500200

    DUVyes115.0 300300Apr. '05LETI/LPM

    +oxidation0.8 20050

    G-lineyes132.0 500200Dec. '01MIT11EBeamno1105.0 400320Dec. '02Yokohama

    EBeamno36.0 300300Dec. '02NTTEBeamno35.0 470270Aug. '03Cornell

    2.5EBeamno23.6 445220Apr. '04IBM< 5DUVno12.4 500220Apr. '04IMEC

    σroughness [nm]

    Fab.top clad

    BOX [um]

    loss [dB/cm]

    w [nm]

    h [nm]

    DateGroup

  • http://photonics.intec.UGent.be 31© intec 2004

    PolarisationIssues:

    Rectangular cross-section: very different neffSquare cross-section:

    neff,TE = neff,TMBut: polarisation conversion + higher losses

    Polarisation conversion in bends studied by SakaiFDTD: Conversion < 25dB (R>1um)Experiment: -13dB to -10dBReason: side wall angle (85o)

    Polarisation insensitivity: hopeless ??Use polarisation diversity

    (Sakai, Fukazawa, Baba, JLT ’04, pp. 520)

  • http://photonics.intec.UGent.be 32© intec 2004

    Temperature dependence

    K/pm140dTdK1079.1

    dTdn c14Si =λ⇒×= −−

    Classical Filters: dTdn

    dd c =

    λλ

    (λc : central wavelength)

  • http://photonics.intec.UGent.be 33© intec 2004

    Temperature DependenceUse Temperature Dependence for TO-switch

    Espinola e.a. (PTL ’03, pp. 1366)

    Lh=650um

    Switching time = 3.5us

    Switching power = 50mW

  • http://photonics.intec.UGent.be 34© intec 2004

    Waveguide Density

    Photonic Crystal Guides have smaller mode diameter but require several rows of holes !!!

    Further scaling: increase height, increase index (?)

    Surface Plasmon waveguides ?

    Min

    . Cen

    ter-

    to-c

    ente

    r [µ

    m]

    Waveguide Width [µm]

    Single mode

    Crosstalk < 20dB/cm

  • http://photonics.intec.UGent.be 35© intec 2004

    Bends

    4040

    0.150.05

    2.05.0

    500220LETI/LPM

    2 bends1.3resonant 400340Columbia

    poly-Si0.3resonant 12 bends0.51.0 500200MIT

    31.0 400320Yokohama0.173.0

    24 bends0.462.0 300300NTT05.0 ?2.0 ?1.0 500220IMEC05.0

    0.0132.0 20 bends0.0861.0 445220IBMNote

    Loss [dB/90]

    Radius [um]

    w [nm]

    h [nm]Group

    (Table partly from Vlasov, McNab, Opt. Expr. ’04, pp1630)

  • http://photonics.intec.UGent.be 36© intec 2004

    Bends

    Wide λ-range

    No need for resonant bends ?

    (Vlasov, Mc Nab)

  • http://photonics.intec.UGent.be 37© intec 2004

    Submicron SOI-wiresIntroduction: why do we need themBasic properties: design, loss, wavelength, polarizationFabricationDevices: couplers, crossings, filters …

    III-V on SiliconIntroductionCoupling of lightFabricationPICMOS (Photonic Interconnect on CMOS)

    Outline

  • http://photonics.intec.UGent.be 38© intec 2004

    Fabrication: Review

    RIE (CF4:Ar)AluminiumEBeamyes1110.0 Columbiamixed5.0

    HBr etchingSiO2DUVyes115.0 LETIoxidation0.8

    RIE (SF6)SiO2G-lineyes32.0 MITICP (CF4 + Xe)metalEBeamno1105.0 Yokohama

    SF6/CF4 etch, ECR-etchEBeamno36.0 NTTICPEBeamno35.0 Cornell

    CF4/CHF3/Ar (Oxide) + HBr (Silicon)SiO2EBeamno23.6 IBM

    Cl2/He/Hbr/O2ResistDUVno12.4 IMEC

    Etch MethodMaskFab.top clad

    BOX [um]

    loss [dB/c

    m]

    Group

  • http://photonics.intec.UGent.be 39© intec 2004

    FabricationEBeam

    Best suited for research, Features < 50nmSlow, not compatible with mass-fabrication (?)

    Standard Litho (G-line, I-Line)OK for 500nm linesProblem for smaller features (gaps in direction coupler, PhC, taper tips

    DUV (248nm, 193nm)Resolution OK (but characterisation needed !)Compatible with Mass-FabricationExpensive mask

  • http://photonics.intec.UGent.be 40© intec 2004

    Original fabrication process

    Si-substrate

    SiO2Si

    Photoresist Photoresist

    AR-coating

    wafer Photoresist(UV3)

    Bare Soft bake AR coating Illumination(248nm deep UV)

    bakePost Development Silicon etch Oxide etch Resist strip

  • http://photonics.intec.UGent.be 41© intec 2004

    Line width with exposure dose

    200

    300

    400

    500

    600

    700

    800

    900

    10 15 20 25 30 35 40

    200

    300

    400

    500

    600

    700

    DesignedLine Width

    Line width (nm)

    Exposure dose (mJ)

  • http://photonics.intec.UGent.be 42© intec 2004

    Hole size with exposure dose

    150

    200

    250

    300

    350

    400

    450

    500

    550

    10 15 20 25 30 35 40

    400/240400/320450/270450/360500/300500/400550/220550/330550/440600/240600/360600/480

    DesignedPitch/diameter

    Hole size (nm)

    Exposure dose (mJ)

    Sufficient ProcesswindowMarginally

    sufficient

  • http://photonics.intec.UGent.be 43© intec 2004

    -5%-10%-15% +5% +10% +15%

    -0.4

    -0.2

    Bestfocus

    +0.2

    +0.4

    Exposure Energy [∆E/E0 ]

    Bestenergy

    Focu

    s [µ

    m]

    λ = 248nmNA = 0.63resist = UV3

    λ = 248nmNA = 0.7resist = UV3

    λ = 248nmNA = 0.7resist = TIS

    Process WindowDesign Pitch/Size: 500nm /300nmTarget size: 300nm

    5% deviation ellipse

    λ = 193nmNA = 0.63resist = TIS

  • http://photonics.intec.UGent.be 44© intec 2004

    -5%-10%-15% +5% +10% +15%

    -0.2

    -0.1

    Bestfocus

    +0.1

    +0.2

    Exposure Energy [∆E/E0 ]

    Bestenergy

    Focu

    s [µ

    m]

    λ = 248nmNA = 0.63resist = UV3

    λ = 248nmNA = 0.7resist = UV3

    λ = 248nmNA = 0.7resist = TIS

    Process WindowDesign Pitch/Size: 400nm /240nmTarget size: 200nm

    5% deviation ellipse

    λ = 193nmNA = 0.63resist = TIS

  • http://photonics.intec.UGent.be 45© intec 2004

    Optical proximity effectsexample:

    triangular lattice

    pitch = 530nm

    Diameter = 420nm

    r/a = 0.4

    1um

    resist

    Bulk hole = 420nm

    Border hole = 380nm

    Corner hole = 350nm

  • http://photonics.intec.UGent.be 46© intec 2004

    Optical proximity effectsW1 waveguide

    pitch = 500nmhole Ø in bulk lattice = 300nm

    0.400.42 0.44 0.46 0.48 0.50k

    0.26

    0.27

    0.28

    0.29

    0.30

    0.31

    0.32a/λ

    Light cone

    Oddlattice

    modes

    Even lattice modes

    0.42 0.44 0.46 0.48 0.50k

    0.40

    MSB MSB

    Øborder=310nm Øborder=320nm λ(nm)

    1925

    1850

    1785

    1725

    1665

    1610

    1560

    0.42 0.44 0.46 0.48k

    0.40

    MSB

    Øborder=300nm

    0.50

    Light cone

    Oddlattice

    modes

    Even lattice modes

    Light cone

    Oddlattice

    modes

    Even lattice modes

  • http://photonics.intec.UGent.be 47© intec 2004

    Optical Proximity CorrectionProblem: Optical Proximity EffectsHoles at lattice boundary are different than holes in bulk due to interference effects

    Correction on mask required (also on PICCO_03)

    Original Mask layout Resist on wafer Mask layout with OPC

  • http://photonics.intec.UGent.be 48© intec 2004

    Optical proximity correctionsDetermine empirically from PICCO_01

    Example: 500nm pitch, 300→320nm holes

    Border hole

    bias (nm)

    Corner hole bias (nm)

    Hol

    e si

    ze d

    evia

    tion

    (nm

    )

    Bulk

    Border hole

    Corner hole

  • http://photonics.intec.UGent.be 49© intec 2004

    Deep Etch RoughnessExample: Ring resonator

    Straight wire=400nmRing wire = 500nm

  • http://photonics.intec.UGent.be 50© intec 2004

    OxidationThermal oxidation of top Silicon layer.

    Roughness reduction

    Lithography Si+SiO2 Etch 20-60nm thermal oxide

    Oxide removalwith HF dip

    optiona

    l

    10nmoxide

    30nmoxide 50nm

    oxide

    Roughness onair-oxide interface

  • http://photonics.intec.UGent.be 51© intec 2004

    Shallow etchingShallow etching → less roughnessBut: Scattering at bottom of hole

    Roughness reduction

    Lithography Si Etch 10nm oxide Oxide deposition

    Re-fill hole with oxide to reduce asymmetry

    200nm hole

    optiona

    l

  • http://photonics.intec.UGent.be 52© intec 2004

    Silicon-only etch

    Deep etching Si-only etching

    Less roughness

  • http://photonics.intec.UGent.be 53© intec 2004

    Etch bias with Silicon-only etchThick resist layer: 800nm UV3

    needed for deep etching200-300nm hole Ø: high aspect ratiocauses litho-etch bias

    800n

    m

    300nmhole Ø

    Litho Etch Result

    shadowof thickresist

    230-250nmhole Ø

  • http://photonics.intec.UGent.be 54© intec 2004

    Etch bias with Silicon-only etchOptimal Solution:

    new litho + etch development: no time.Short-term Solution:

    Resist-hardening/Resist Trimming plasma treatment

    800n

    m

    300nmhole Ø

    Litho Etch ResultRH

    300nmhole Ø

    smallershadow

    Still slightly sloped sidewalls

  • http://photonics.intec.UGent.be 55© intec 2004

    Updated Fabrication Process

    Si-substrate

    SiO2Si

    Photoresist Photoresist

    AR-coating

    wafer Photoresist(UV3)

    Bare Soft bake AR coating Illumination(248nm deep UV)

    bakePost Development Resist Hardening

    Silicon etch Resist strip

  • http://photonics.intec.UGent.be 56© intec 2004

    2-step processingTwo types of structures

    Waveguides: requires deep etch (al least through Silicon)

    Fibre couplers: require 50nm etch

    Two-step processingFibre couplers first: 50nm etch gives little topography

    Wafer-scale alignment: Alignment markers on the wafer and reticle periphery, not between the structures

  • http://photonics.intec.UGent.be 57© intec 2004

    2-step processing

    deep trench

    shallow fibre coupler

  • http://photonics.intec.UGent.be 58© intec 2004

    CMOS-compatible ?Well, it is SiliconIt is processed in a CMOS line

    ButCMOS = layered

    We: lines, holes, gaps, tips all in same layerCMOS = vias but rather low density

    Phot Crystals = superdense latticesLine-edge roughness: no issue in CMOS (till now)

    Roughness kills everything

  • http://photonics.intec.UGent.be 59© intec 2004

    Submicron SOI-wiresIntroduction: why do we need themBasic properties: design, loss, wavelength, polarizationFabricationDevices

    III-V on SiliconIntroductionCoupling of lightFabricationPICMOS (Photonic Interconnect on CMOS)

    Outline

    • Couplers• Crossings• Ring Resonators• AWG• Cascaded MachZehnder• Fibre-chip couplers

  • http://photonics.intec.UGent.be 60© intec 2004

    Couplers - SplittersDirectional Couplers

    Used in ring resonators, cascaded MZI …“Easy” to choose splitting ratioSensitive to fabrication issues (optical proximity, deviations in widths)

    Multi-mode interference couplers (MMI)Fabrication tolerant

    Standard Y-juncionSymmetric, narrow gap

    Advanced CouplersSakai, Fukazawa, Baba, IEICE Trans ’02, 1033

  • http://photonics.intec.UGent.be 61© intec 2004

    CouplersYokohama Nat. Univ

    Simulation:

  • http://photonics.intec.UGent.be 62© intec 2004

    CrossingsCrosstalk free crossings in optics ?

    Standard crossingLarge diffractionLarge crosstalk (-9dB)Large loss (1.4dB)

    Enhanced versionsBetter performanceLarger

    NOT acceptable for large density circuitsUse multiple waveguide layers ??

  • http://photonics.intec.UGent.be 63© intec 2004

    Fibre coupling

    InP ridge wg

    SM-fibre core

    SOI PhC wg

    µmMode mismatch between waveguide and fibre

  • http://photonics.intec.UGent.be 64© intec 2004

    Coupling to fiberImportant:

    Large bandwidth

    Low loss

    FabricationLimited extra processingTolerant to fabrication deviations

    Coupling tolerance If coupling to SMF: same for all types of taperCoupling to high-NA fiber: lower

  • http://photonics.intec.UGent.be 65© intec 2004

    Fiber-chip couplingRegular taper

    Difficult to fabricate

    Multi-mode

    Facet coating required

  • http://photonics.intec.UGent.be 66© intec 2004

    Coupling to fiberInverse taper

    Broad wavelength range

    Single mode

    Easy to fabricate (if you can do the tips)

    Low facet reflections

    0.4µm

    80nm

    0.2µm

    500 µm

    polished facet

  • http://photonics.intec.UGent.be 67© intec 2004

    Coupling to fiber

    3x32x002x2

    Cladding Size

    0.8Polymer60.0200.0 300300NTT< 4dBSiO2100.040.0 470270Cornell< 1dBPolymer75.0150.0 445220IBM

    tbdPolymer500220IMEC

    LossCladding Material

    tip width [nm]

    L [um]

    w [nm]

    h [nm]

    Group

    0.4µm

    80nm

    0.2µm

    500 µm

    polished facet

  • http://photonics.intec.UGent.be 68© intec 2004

    Coupling to fibreTip fabrication

    EBeam

    Modified DUV (resist trimming)

    220nm

  • http://photonics.intec.UGent.be 69© intec 2004

    spot size convertor

    Single modefiber core

    Coupling to fiberThe vertical fiber coupler

    use a grating to couple light from/to a fiber perpendicular to the PIC

    use a spot-sizeconvertor in plane

    wafer scale, no need to cleave/polish the devices

    good alignment tolerances

    relatively broadband

    works for TE only

    (artist’s impression)

  • http://photonics.intec.UGent.be 70© intec 2004

    Out-of-plane coupler

    What is new ?Other grating couplers

    long (>100µm)

    very narrow bandwidth

    couple in and out

    high efficiency (>50%)

    Our grating couplershort (10µm)

    bandwidth > 50nm possible

    couple in and out

    high efficiency ?

    Grating couplers :‘Second’ order grating (Λ=λ /neff)

    First order diffraction couples light out of the waveguide producing a surface normal propagating field

  • http://photonics.intec.UGent.be 71© intec 2004

    Fabricated DevicesAlternative: Grating couplers

    Waferscale testing

    Waferscale packaging

    High alignment tolerance

    deep trench

    shallow fibre coupler

    Towards optical circuit

    Single modefiber core

    From Fibre

    -40

    -35

    -30

    -25

    -20

    -15

    -10

    -51500 1520 1540 1560 1580 1600

    Tran

    smis

    sion

    [dB

    ]

    Wavelength [nm]

    ∆λ1dB = 35nm

  • http://photonics.intec.UGent.be 72© intec 2004

    -60

    -55

    -50

    -45

    -40

    -35

    -30

    -25

    -20

    -151500 1510 1520 1530 1540 1550 1560 1570 1580 1590 1600

    wavelength [nm]

    T [d

    B] passdrop

    bcb cladding, ring resonator with bend coupling, R=8µm

  • http://photonics.intec.UGent.be 73© intec 2004

    Fiber CouplersCoupling light from waveguide to optical fiber on top

  • http://photonics.intec.UGent.be 74© intec 2004

    Experimental results

    33% efficiency (4.8dB coupling loss)35-40nm 1dB bandwidth

    0.00

    0.10

    0.20

    0.30

    0.40

    1520 1560 1600 1640

    wavelength (nm)

    fiber

    cou

    plin

    g ef

    ficie

    ncy

    620nm period630nm period620nm theory630nm theory

  • http://photonics.intec.UGent.be 75© intec 2004

    2D grating fiber couplerFiber to waveguide interface for polarisationindependent photonic integrated circuit

    2D grating

    couples each fiber polarisationin its own waveguide

    in the waveguides the polarisation is the same (TE)

    Allows for polarisationdiversity approach

    patent

    Single modefiber core

  • http://photonics.intec.UGent.be 76© intec 2004

    Experimental resultsFabrication

    SOI: 220nm Si / 1000nm SiO2Etch depth: 90nm

    Square lattice of holes: 580nm period

  • http://photonics.intec.UGent.be 77© intec 2004

    Optical Ring ResonatorsOptical Ring Resonators

    For a given loss: trade-offHigh Q Low coupling

    ButLow coupling Low drop efficiency !!!

    Relevant Characteristics:• Free Spectral Range (Period)• Quality Factor

    Determined by:• Coupling ratio• Round-trip loss• Length (=radius)

  • http://photonics.intec.UGent.be 78© intec 2004

    Optical Ring Resonator

    1um

    5um

  • http://photonics.intec.UGent.be 79© intec 2004

    Optical Ring Resonators

    -60.00

    -50.00

    -40.00

    -30.00

    -20.00

    -10.00

    1520 1540 1560 1580

    Ring resonator demux4 rings in series

    Linearly increasing radius

    λc does not increase linearly as expected !!

    Fabrication problem: mask discretisation

    Solution: vary parameter which is less sensitive to fabrication

    Other:Peak splitting due to reflections

  • http://photonics.intec.UGent.be 80© intec 2004

  • http://photonics.intec.UGent.be 81© intec 2004

    Increasing Index ContrastLow Contrast - Fiber Matched

    (silica or polymer based)Bend Radius ~ 5 mmSize ~ several cm^2

    Medium Contrast (InP-InGaAsP)

    Bend Radius ~ 500µm

    5 mm

    Ulra-high Contrast (SOI based)

    Bend Radius < 50µm

    200

    µm

    5 cm

  • http://photonics.intec.UGent.be 82© intec 2004

    AWG DesignVarious devices were designed :

    ∆ν = 400GHzFSR = 8 x 400GHzw = 0.5µm# arms = 18 or 24R = 75µm ~ 150µmwi = 0.6µm ~ 1.0µmwg = 0.6µm ~ 1.0µmg = 0.2µm

    200µm To gratingcouplers

    R

    wg

    wiw

    All designs fabricated with different exposure doses (during litho)

    different actual waveguide widths

    wg

    wi

    g

  • http://photonics.intec.UGent.be 83© intec 2004

    200µmAWG ResultsAWG, 400GHz spacing, 8 channels

    ∆ν = 340GHz 360GHz (different exposure times)8dB on-chip loss

    -25

    -20

    -15

    -10

    -51500 1520 1540 1560 1580 1600

    Tran

    smis

    sion

    [dB

    ]

    Wavelength [nm]

  • http://photonics.intec.UGent.be 84© intec 2004

    O2

    -25

    -20

    -15

    -10

    -51500 1520 1540 1560 1580 1600

    200µmAWG Results5 x 8 AWG, 400GHz spacing, 8 Channels

    300µm x 300µm area8dB on-chip loss6-10 dB crosstalk

    Tran

    smis

    sion

    [dB

    ]

    Wavelength [nm]

    7dB

  • http://photonics.intec.UGent.be 85© intec 2004

    AWGYokohama Nat. University

    (Fukazawa, Ohno, Baba, Jap. J of Appl. Physics, ’04)

  • http://photonics.intec.UGent.be 86© intec 2004

    AWG Crosstalk OriginPossible reasons for crosstalk

    “Overspill” in star-coupler

    Reflections in star-coupler

    Phase errors in grating arms

  • http://photonics.intec.UGent.be 87© intec 2004

    AWG Crosstalk OriginPhase errors in Waveguide arms ?

    Assume standard deviation for phase-error given by :

    ic

    Lfi

    πσφ1

    =

    Cro

    ssta

    lk L

    evel

    [dB

    ]

    fc

    Calculated Crosstalk vs. fc

    Correlation length [µm]0.01 0.1 1 10 100 103 104

    Rou

    ghne

    ss[n

    m]

    5

    10

    fc=100

    fc

    fc=100

  • http://photonics.intec.UGent.be 88© intec 2004

    -25.00

    -20.00

    -15.00

    -10.00

    -5.00

    0.00

    1520.00 1530.00 1540.00 1550.00 1560.00 1570.0

    Cascaded MZ FilterExample: 5 stage CMZ

    3.2nm bandwidth

    17nm FSR

    coupling efficiency ~80%

    -10 dB crosstalk

    wavelength [nm]no

    rmal

    ized

    out

    put [

    dB]

    pass

    drop

    20µm 14µm 20µm 20µm 14µm 20µm

    ∆L = 32.8µm

    gap width = 220nm

    waveguide width= 535nm

    waveguide width= 565nm

  • http://photonics.intec.UGent.be 89© intec 2004

    PICCO04: Cascaded MZ FilterExample: 5 stage CMZ

    2.6nm bandwidth

    17nm FSR

    coupling efficiency ~100%

    -10 dB crosstalk-25.00

    -20.00

    -15.00

    -10.00

    -5.00

    0.00

    1520.00 1530.00 1540.00 1550.00 1560.00 1570.0

    wavelength [nm]no

    rmal

    ized

    out

    put [

    dB]

    pass

    drop

    26µm 14µm 20µm 20µm 14µm 26µm

    ∆L = 32.8µm

    gap width = 220nm

    waveguide width= 535nm

    waveguide width= 565nm

  • http://photonics.intec.UGent.be 90© intec 2004

    Optical ProximityEffectsOptical Lithography

    Images of neighbouringstructures interfere

    Effect can be additiveof subtractive

    = optical proximity effects

    Example:isolated line width: 565

    gap width: 220nm

    line width in coupling section: 535nm

    wg

    wi

    wc

    W. Bogaerts et al. to be published in JLT (Oct 2004)

  • http://photonics.intec.UGent.be 91© intec 2004

    ConclusionsSub-micron SOI-waveguides:

    Powerful platform for high-density Photonic circuitsWe have all basic building blocs (and no need for these complicated Photonic Crystals)

    Fabrication issues to be solvedOptical proximity (narrow lines, fine gaps)Phase-Errors, control central wavelength of devicesFurther reduction losses needed ?

    Next step: active functionality ?


Recommended