wtheiss_ICCG6

Embed Size (px)

Citation preview

  • 8/13/2019 wtheiss_ICCG6

    1/4

    Model meets measurement thechallenges of optical spectrum simu-lation facing real thin film deposition

    W.TheissW.Theiss Hard- and Software, Aachen, Germany

    AbstractOptical spectrum simulation is a powerful technique

    providin valua!le information for the control of thin film

    deposition devices. The required inredients of this

    method are descri!ed. Successful optical models have

    ad"usta!le parameters which reflect the possi!le variations

    of the deposition hardware. The values of these parameters

    can !e determined analy#in spectra recorded !y an

    appropriate networ$ of spectrometers. This way the

    current state of the coatin deposition is $nown and - if

    necessary - corrections leadin to ideal production

    conditions can !e made.

    %eywords& optical spectrum, simulation, thin film,production control

    1. Introduction

    Optical spectroscopy is a powerful method to characteri#e

    thin films. Applied durin or directly after the deposition,

    it supplies information a!out the thic$ness and the

    comple' refractive inde' of the deposited material. The

    refractive inde' (or more eneral& the optical constants)

    reflect, in an indirect way, the composition of the film.

    Hence quantities li$e electrical conductivity or o'yen

    content can !e deduced from a quic$, non-destructive

    optical inspection.

    The !est technique to analy#e optical thin film spectra is

    to ad"ust the parameters of a suita!le optical model in

    such a way that compara!le simulated and measured

    spectra match. This parameter fittin method provides the

    fle'i!ility required in cases of floatin production

    conditions (due to, for e'ample, taret erosion in

    sputterin devices, as pressure variations, temperature

    drifts). *ependin on the wanted information, optical

    models can !e made very simple deliverin only a few $ey

    parameters li$e thic$nesses, or rather sophisticated,

    providin information a!out su!tle details of the

    produced coatins.

    Section + summari#es required features of a spectrum

    simulation software to !e used with comple' deposition

    devices (li$e, for e'ample, production lines for

    architectural lass coatins). Then, in the main section ,

    the systematic development of optical models is discussed.

    The article closes with some remar$s in section a!out

    the setup of an optical inspection networ$ providin the

    information required for deposition control.

    2. Optical spectrum simulation

    Optical inspection of thin film deposition requires thecom!ination of spectroscopic hardware and data analysis

    software. To !e equipped for cases of comple' coatin

    equipment for multilayer systems the applied spectrum

    simulation proram should have the followin features&

    fle'i!le optical constant models

    several layer stac$s, partially connected

    simultaneous computation of several spectra

    n the followin sections some details concernin these

    points are discussed.

    2.1 Optical constant models

    Althouh a lot of effort is usually put into the deposition

    hardware to produce thin films with sta!le properties, the

    optical constants in the various layers of a coatin cannot

    !e considered to !e the same all the time. This e'cludes

    the use of fi'ed ta!les of literature data for optical

    constants. nstead, one must employ fle'i!le models that

    can follow driftin material properties. The !est choice is

    to wor$ with physical models !ased on parameters which

    have a (more or less) intuitive meanin.

    /owerful optical constant models to !e applied inspectroscopic production control must !e availa!le from

    the 01 (wavelenth 2 33 nm) to the 45 (up to

    67833 nm). This spectral rane can !e covered !y fast

    array spectrometers. n some cases - if 9T5 (9ourier

    Transform nfra-5ed) spectrometers are used - also the

    mid-infrared ( +333 nm : : +3333 nm) is of interest.

    Typical material e'citations are electronic inter!and

    transitions and the acceleration of free chare carriers in

    conductive materials. Ta!le 7 ives an overview on

    appropriate models for the various types of e'citations ;

    the optical constants of a iven material are then o!tained

    !y the superposition of several suscepti!ility terms.

  • 8/13/2019 wtheiss_ICCG6

    2/4

    availa!le in the software. n most cases it is sufficient to

    use the simple Drueman mi'in formula ?E@.

    2.2 Layer stacks

    Once the optical constant models are set for all relevant

    materials, one has to define the sequence of these

    materials in the coatin.

    n order to ma$e use of measured spectra of incomplete

    coatins, ta$en in !etween two deposition steps, thesoftware should !e a!le to handle several layer stac$s

    within one optical model. t must !e possi!le to connect

    layer thic$nesses of different stac$s in order to use

    information o!tained for one stac$ in another.

    >iht propaation in thin films must !e computed with

    coherent superposition of partially reflected and

    transmitted waves. 9or thic$ su!strates li$e lass panes

    incoherent superposition (no interference patterns visi!le

    in the spectrum) should !e selected.

    2.3 Spectra

    9or each spectrometer in the optical inspection networ$there must !e its counterpart in the model, i.e. a simulated

    reflectance, transmittance or ellipsometry spectrum.

    *etails li$e anle of incidence and polari#ation of the

    incident radiation should match. This applies to the

    normali#ation as well& f the hardware records relative

    spectra, the software must do that as well.

    0sually there is a unique assinment of layer stac$s and

    spectra& 9or each spectrum there is a correspondin layer

    stac$. However, in the case of products with a

    microstructure on a lenth scale much smaller than the

    optical measurement spot, the spectrometers see a mi'ture

    of several layer stac$s. n order to analy#e such spectra thesimulation proram must !e a!le to averae the spectra of

    several layer stac$s, ta$in into account their individual

    area fractions.

    2.4 rediction o! coatin" per!ormance

    t can !e advantaeous to !e a!le to compute technical

    data li$e color coordinates ?8@, emissivities ?F@, 0 ?F@and

    ?@values for the currently produced coatin. This way a

    direct prediction of the final products performance is

    possi!le which simplifies decisions concernin the

    modification of deposition parameters.

    3 #e$elopin" optical models

    Good optical models used for production control can save

    a lare amount of money ; hence their development has

    to !e done as careful as possi!le. The most efficient way to

    wor$ is to use the same optical model for the desin of

    new coatins and their production control. This ensures

    that $nowlede a!out production pro!lems and

    phenomena (li$e mi'in of ad"acent layers) can !e ta$en

    into account in the desin phase already. 4ew desins, on

    the other hand, can !e turned into methods for

    production control most easily.

    3.1 Sin"le layers

    The !asis of any success are relia!le optical constants of

    the deposited materials. 9or dielectric materials (o'ides,

    nitrides) it is recommended to deposit thin and thic$

    sinle layers on typical su!strates. The spectra of the thic$

    layers should e'hi!it at least one interference ma'imum or

    minimum . This ensures a quite safe determination of the

    optical constants. =omparin measured spectra of the thin

    layer versions with simulated spectra o!tained from asimple thic$ness fit (Ifro#en optical constant model) one

    can easily chec$ if the o!tained optical constants wor$ for

    !oth thin and thic$ layers. f not, a systematic study of the

    variation of the optical constants with thic$ness miht !e

    useful.

    The test layers should !e produced with the final

    deposition equipment, even if it hurts to run an e'pensive

    production line for several sets of sinle layers.

    nformation o!tained directly from the factory may turn

    out to !e very valua!le, in particular if you allow a

    feed!ac$ of the production model to the one used for

    coatin desin.

    Whenever possi!le, one should use parameters which are

    reflectin the controls of the deposition equipment. f, for

    e'ample, the reactive sputterin of an o'ide is reulated !y

    the settin of a lam!da-pro!e and the value of the

    electrical power, there should !e two correspondin

    parameters in the model. f the applied optical constant

    model requires more parameters, it is useful to investiate

    how the model parameters depend on the machine

    settins. 9i. 7 shows how the ap enery of an O> model

    ?+@ varies with the value of a deposition parameter. The

    quite simple relation can easily !e e'pressed !y

    polynomial interpolation. f it is possi!le to e'press alloptical model parameters as simple functions of the

    machine parameters, these relations can !e implemented

    in the optical model itself& The machine parameters are

    then used as Imaster parameters in the model (here&

    lam!da-pro!e settin, electrical power), and the optical

    constants of the produced material are computed usin

    Islave parameters li$e ap enery or oscillator strenths.

    9i. 7. J'ample for a simple relation !etween a deposition

    parameter (lam!da-pro!e settin) and an optical

    parameter (ap enery of a reactively sputtered o'ide).

    ,.,

    ,.-

    ,.B

    ,.E

    ,.8

    3 7 + , - B E

    %achine settin" &m'(

    )apener"y&e'

    (

  • 8/13/2019 wtheiss_ICCG6

    3/4

    n luc$y cases a sinle parameter li$e a lam!da-pro!e

    settin determines the optical constants of the material,

    and an independent second parameter li$e the electrical

    power is responsi!le for the achieved layer thic$ness. 9i. +

    shows an e'ample.

    9i. +. Optical constants of a reactively sputtered o'ide&

    The upper curves show the real part of the comple'refractive inde' n, the lower curves the imainary part $.

    The solid lines ive the optical constants for E m1, the

    dashed ones for 3 m1.

    n real systems the correlation of deposition and model

    parameters may not !e perfect.

  • 8/13/2019 wtheiss_ICCG6

    4/4

    9i. & Schematic s$etch of a satisfyin silver layer model&

    The rouh interfaces of the Ireal system on the left are

    appro'imated !y thin effective-medium layers, and the

    reduction of electron mo!ility due to diffuse scatterin at

    the rouh !oundaries is ta$en into account !y an increase

    of the dampin constant in the *rude model.

    3.3 Application e*ample

    After careful development of a coatin model, ta$in into

    account realistic sinle layer results and the interaction

    effects of neih!oured layers in the stac$, it is possi!le to

    achieve a ood areement of model and measured coatin

    properties even for stac$s with many layers. 9i. shows

    the comparison of measured and simulated reflectance

    and transmittance spectra of a commercial solar control

    coatin. The development of this hih performance

    product has !een supported !y model computations,

    which helped sinificantly to reduce the amount of

    required e'pensive production of test samples in the plant.

    9i. & Simulated (dashed) and measured (solid) spectra of

    a commercial solar control coatin (SJeary et al., .Appl. /hys. F+, 8 (78) -3.

    ?@ =.=. %im et al., /hys. 5ev. D B (+3) (7+) 778-778E8.

    ?@ /.*rude, Ann. /hys. (733) E.

    ?B@ *