16313 Halewijn Presentation

Embed Size (px)

Citation preview

  • 7/28/2019 16313 Halewijn Presentation

    1/24

    23-10-2012 NXP

    MEMS resonator

    Presentation by

    Drs. Helger van Halewijn

    1

    Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan

  • 7/28/2019 16313 Halewijn Presentation

    2/24

    MEMS resonator overview package

    Resonatorsize500x700x150micron

    SinglesilicononCMOStechnology

    Highfrequencystability,lowtimejitter,lowtemperaturedrift.

    LowmotiondampingQ-factor>40000(50MHz)

    23-10-2012

    NXP2

  • 7/28/2019 16313 Halewijn Presentation

    3/24

    Overview of dogbone resonator

    23-10-2012 NXP

    actuationgap

    anchormass ofresonator

    spring

    3

  • 7/28/2019 16313 Halewijn Presentation

    4/24

    Schematic of dogbone resonator

    23-10-2012 NXP 4

  • 7/28/2019 16313 Halewijn Presentation

    5/24

    Some parameters studied in COMSOL Duringproductioninwaferfaballdimensionsvariationscaninfluencetheperformance

    Q-factor,resonancestability1. Dimensionsresonatorhead

    2. Anchorloss3. ThicknessSi

    4. Oxidationlayer

    5. Airpressure(vacuum)\

    6. Thermallosses.

    7. ViscousDrag

    8. Electrostaticactuation+fringing

    9. Modecoupling.

    23-10-2012 NXP 5

  • 7/28/2019 16313 Halewijn Presentation

    6/24

    Non linear resonance frequency Resonancefrequency

    1. kspringconstant

    2. mmasssystem

    3. Vdrivingvoltage

    4. wwidth

    5. hheight

    6. ggap

    23-10-2012 NXP

    2

    03

    sin( ) DCel AC

    DC

    res

    V xF V t

    g

    V whk

    f m mg

    = +

    =

    6

  • 7/28/2019 16313 Halewijn Presentation

    7/24

    Q-factor estimation Reciprocaladdition.

    Anchorloss:Qanchor104-107dimension Thermo-elasticlossQte10

    4-106material

    Surfaceloss:Qsur106-108debrisorcracks

    Airdamping:Qair102107vacuumquality,leakage

    ViscousdragQdrag105-107vacuumquality

    AllseperateQfactorsareestimatedwithCOMSOL.

    23-10-2012 NXP

    1

    1 1N

    itot iQ Q=

    =

    7

  • 7/28/2019 16313 Halewijn Presentation

    8/24

    Overview resonance mode at 56 MHz

    23-10-2012 NXP 8

  • 7/28/2019 16313 Halewijn Presentation

    9/24

    Acoustic loss at anchor.

    23-10-2012 NXP 9

  • 7/28/2019 16313 Halewijn Presentation

    10/24

    Acoustic loss: Gray domains, PML

    23-10-2012 NXP

    10

    Red arrowsindicate PML-layers. In PML completeacousticabsorption. Blue domainsrepresent normalmaterial properties. Eigenfrequencyanalyses.

  • 7/28/2019 16313 Halewijn Presentation

    11/24

    Acoustic loss: dimensionalvariations in the structure.

    23-10-2012 NXP 11

  • 7/28/2019 16313 Halewijn Presentation

    12/24

    23-10-2012 NXP 12

    Capacitances

    Areaofgatewithingreenlinesis41084m2(drainareaexcluded)

  • 7/28/2019 16313 Halewijn Presentation

    13/24

    Fringe effects, from simple to real system

    SimulatedCap

    =338fF

    Dominatedby

    drainbottom

    capacitance

    23-10-2012 NXP 13

  • 7/28/2019 16313 Halewijn Presentation

    14/24

    23-10-2012 NXP 14

    1. Simple hand calculation Drain C1 = 322 fF2. Gate C2 = 1270 fF and Source C3 = 18700 fF3. Capacitance C12=22.5 fF and C23=51 fF4. Using mathematical formula

    . Cgate = 1264 fF6. Comsol result is Cgate Comsol= 1266 fF OK7. Csource and Cdrain have similar accuracies.

    Calculation + Simulation Gate Capacitance

    312

    2312 11

    1

    CCC

    CCCgate

    +

    +

    ++=

  • 7/28/2019 16313 Halewijn Presentation

    15/24

    23-10-2012 NXP15

    Capacitance as function of resistivity

    1. ResultsfromComsol

    2. Includingeffectslikeairheightandsubstratesthickness

    3. Resonatoroperatesat0.31.2cm(

    equivalentto80330S/m)4. Capacitancevariationis

  • 7/28/2019 16313 Halewijn Presentation

    16/24

    Asymmetric amplitude underelectrical load (V bias)

    At10Volt,aharmonic

    behavior.

    At80Voltbiasan

    inharmonictermcaneasilybeseen.

    At56MHz,oscillation

    timeis40nsec.

    COMSOLtimestep

  • 7/28/2019 16313 Halewijn Presentation

    17/24

    Static stressdue to process cycles

    23-10-2012 NXP

    17

    Staticstressatperiferyduetocuring

    proceduresatvariousprocessconditions

    Stressconcentrationsinresonatorlegs.

    Ue 1.1 m

    Ue 7.1 m Ue 13.1 m

  • 7/28/2019 16313 Halewijn Presentation

    18/24

    Q-factor of resonator under Stress

    23-10-2012 NXP

    18

    0

    50000

    100000

    150000

    200000

    250000

    300000

    0.E+00 2.E-06 4.E-06 6.E-06 8.E-06 1.E-05

    Qf

    acto

    r

    Underetch [m]

    Q-factor as funct ion of Underetch and Stress

    0 MPa

    2.6 MPa

    5.3 MPa

    10.5 MPa

    21.0 MPa

    31.5 MPa

  • 7/28/2019 16313 Halewijn Presentation

    19/24

    Thermal losses in resonator

    Q-factorcanbewrittenas:

    E=YoungsModulus[Pa]

    =expansioncoefficient[1/m]

    To=ambienttemperature[K]

    =density[kg/m3]

    =frequency[rad/s]

    =thermalrelaxation[s]

    23-10-2012 NXP 19

    2

    2

    1

    1 ( )

    o

    therm p

    E T

    Q C

    =+

  • 7/28/2019 16313 Halewijn Presentation

    20/24

    Thermal losses in resonator

    23-10-2012 NXP

    20

    Temperaturefluctuations ofabout 100C, at56 MHz.This mechanismis due to materialproperties.

  • 7/28/2019 16313 Halewijn Presentation

    21/24

    23-10-2012NXP

    Thickness Oxidation: Q factor and frequency shift Simulationwithsingle

    layer.

    Meshisadaptedatlayerslowerthan20

    nm.

    Minimumlayer

    thickness2.5nm.

    InComsol,Solid

    mechanicsandshells

    arecombined.

    Q factor Straight resonator

    0

    2000000

    4000000

    6000000

    8000000

    10000000

    12000000

    14000000

    0 10 20 30 40 50 60

    SiO2 Layer thickness [nm]

    Qf

    actor

    55000000

    55200000

    55400000

    55600000

    55800000

    56000000

    56200000

    56400000

    0 20 40 60

    Frequency

    [Hz]

    SiO2 Layer thickness [nm]

    Frequency Straight reso nator

    21

  • 7/28/2019 16313 Halewijn Presentation

    22/24

    Squeezed film effect including stress

    23-10-2012 NXP22

    1000

    10000

    100000

    10 100 1000 10000 100000

    Q-factor

    Pressure [Pa]

    Q-factor compared with

    measurements KIM

    Q-factor 12nm

    Q-factor 20nm

    KIM

    Q > 40000 atappropriateconditions Sliding effect onlarge surfacenegligable Q-factor is slightlydependant on meshsize of the surfacewhere the filmdamping is applied.

  • 7/28/2019 16313 Halewijn Presentation

    23/24

    Overview: Influence of Polyamide Cappingabove resonator

    23-10-2012 NXP 23

    10,000

    100,000

    1,000,000

    10,000,000

    100,000,000

    -30.00 -20.00 -10.00 0.00 10.00 20.00 30.00 40.00 50.00

    Q-factor

    Frequency shift [ppm]

    Q factor and Capping stress, Polyimide only

    UnderEtch 5.4 [m]

    UnderEtch 1.4 [m]

    -10 MPa

    -100 MPa

    100 MPa

    31 MPa

    10MPa

    Permanent frequency shifts: 20 ppmWell within specs.

  • 7/28/2019 16313 Halewijn Presentation

    24/24

    Conclusion ProductionstepsinCMOStechnlogyhavetheirown

    influenceontheperformanceoftheresonator.

    COMSOLpavedthewaytobetterunderstandingtocontrol

    specsinproduction.(Q-factor,dimensions,stressand

    materiallossfactors)

    MEMSmodulewasusedtogetherwiththemechanical

    module.

    PrestressedAnalysis,eigenfrequency

    PrestressedAnalysis,frequencydomain.

    Heatmodule,Squeezedfilmandmuchmore.

    ThankstoDr.HvandeVlist(NXP,Nijmegen)

    Dr.J.vanBeek(NXP,Eindhoven)

    23-10-2012NXP

    24