Thin Film Interference Filter and Bootstrap Method for Interference Filter Thin Film Deposition Process Control

Abstract

A thin film interference filter system includes a plurality of stacked films having a determined reflectance; a modeled monitor curve; and a topmost layer configured to exhibit a wavelength corresponding to one of the determined reflectance or the modeled monitor curve. The topmost layer is placed on the plurality of stacked films and can be a low-index film such as silica or a high index film such as niobia.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims benefit of U.S. Provisional Patent Application, Ser. No. 60/609,406, entitled “Bootstrap Method for Interference Filter Thin Film Deposition Process Control,” filed Sep. 13, 2004.

STATEMENT OF FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

Research described in this application was sponsored by the United States Air Force Research Laboratory grant number F22615-00-2-6059.

FIELD OF THE INVENTION

The present invention relates to thin film optical devices. More particularly, the invention relates to complex interference filters.

BACKGROUND OF THE INVENTION

Highly accurate optical interference filters can be manufactured using thin film deposition processes. These optical interference filters are used for multivariate optical computing, multiple-band-pass, and the like and can exhibit complex optical spectra defined over a range of wavelengths. These filters are typically constructed by depositing alternating layers of transparent materials where one layer possesses a much larger refractive index relative to the other layer. Theoretically, the proper choice of composition, thickness and quantity of layers could result in a device with any desired transmission spectrum.

Among the simplest devices is the single cavity bandpass filter; i.e., the thin-film form of an etalon. This device consists of three sets of layers. The first stack is a dielectric mirror, a next thicker layer forms a spacer, and a second stack forms another dielectric mirror. The mirror stacks are typically fabricated by depositing alternating transparent materials that have an optical thickness that is one quarter of the optical wavelength of light.

To achieve theoretical optical performance, each layer must possess a precise and specific physical thickness and refractive index. Any nonuniformity in the deposition of the layers can affect the spectral placement and transmission or reflection characteristics of the device. A design that requires very tight manufacturing tolerances over large substrate areas could result in the costly rejection of many devices. Given these manufacturing limits, it would be desirable to analyze the devices after construction and alter the devices that do not meet a predetermined optical transmission or reflection specification by some electrical or mechanical means. For example, if the peak transmission wavelength of a manufactured optical bandpass cavity filter was slightly out of tolerance, it would be desirable to have a mechanism or process for shifting the peak back to the desired spectral location. It is also desirable that the optical filters have precise rejection bands and passbands that are electrically or mechanically selectable.

Mechanical methods of achieving a variable transmission spectrum device are well known. This includes changing a prism or grating angle, or altering the optical spacing between mirrors of an etalon. To overcome the performance, size and cost disadvantages of using mechanical schemes, many have conceived of electrical methods for varying a transmission spectrum. For example, U.S. Pat. No. 5,150,236, issued Sep. 22, 1992 to Patel, discloses a tunable liquid crystal etalon filter. The liquid crystal fills the space between dielectric mirrors. Electrodes on the mirrors are used to apply an electric field, which changes the orientation of the liquid crystal that changes the optical length for tuning. The change in the optical length corresponds to a change in the location of the passband. U.S. Pat. No. 5,103,340, issued Apr. 7, 1992 to Dono et al., discloses piezoelectric elements placed outside the optical path that are used to change the spacing between cascaded cavity filters. Furthermore, U.S. Pat. No. 5,799,231, issued Aug. 25, 1998 to Gates et al., discloses a variable index distributed mirror. This is a dielectric mirror with half of the layers having a variable refractive index that is matched to other layers. Changing the applied field increases the index difference that increases the reflectance. The mathematics that describes the transmission characteristics of multilayer films composed of electro-optic and dielectric materials are well known.

Another electrically actuated thin film optical filter uses a series of crossed polarizers and liquid crystalline layers that allow electrical controls to vary the amount of polarization rotation in the liquid by applying an electric field in such a way that some wavelengths are selectively transmitted. However, these electrically actuated thin film optical filters have the characteristic that the light must be polarized and that the frequencies of light not passed are absorbed, not reflected. Another electrically actuated thin film optical device is the tunable liquid crystal etalon optical filter. The tunable liquid crystal etalon optical filter uses a liquid crystal between two dielectric mirrors.

The common cavity filter, such as the etalon optical filter, is an optical filter with one or more spacer layers that are deposited in the stack and define the wavelength of the rejection and pass bands. The optical thickness of the film defines the placement of the passband. U.S. Pat. No. 5,710,655, issued Jan. 20, 1998 to Rumbaugh et al., discloses a cavity thickness compensated etalon filter.

In the tunable liquid crystal etalon optical filter, an electric field is applied to the liquid crystal that changes the optical length between the two mirrors so as to change the passband of the etalon. Still another tunable optical filter device tunes the passband by using piezoelectric elements to mechanically change the physical spacing between mirrors of an etalon filter.

Bulk dielectrics are made by subtractive methods like polishing from a larger piece; whereas thin film layer are made by additive methods like vapor or liquid phase deposition. A bulk optical dielectric, e.g., greater than ten microns, disposed between metal or dielectric mirrors suffers from excessive manufacturing tolerances and costs. Moreover, the bulk material provides unpredictable, imprecise, irregular, or otherwise undesirable passbands. These electrical and mechanical optical filters disadvantageously do not provide precise rejection bands and passbands that are repeatably manufactured.

In an attempt to avoid some of the foregoing problems, modeling of interference filters can be conducted during on-line fabrication with in-situ optical spectroscopy of the filter during deposition. The current state of the art for on-line correction of the deposition involves fitting the observed spectra to a multilayer model composed of “ideal” films based on a model for each film. The resulting model spectra are approximations of the actual spectra. To use reflectance as an example: the measured reflectance of a stack of films can be approximately matched to a theoretical reflectance spectrum by modeling. Layers remaining to be deposited can then be adjusted to compensate for errors in the film stack already deposited, provided the film stack has been accurately modeled. However, films vary in ways that cannot be readily modelled using any fixed or simple physical model. Heterogeneities in the films that cannot be predicted or compensated by this method cause the observed spectra to deviate more and more from the model. This makes continued automatic deposition very difficult; complex film stacks are therefore very operator-intensive and have a high failure rate. To improve efficiency in fabrication, laboratories that fabricate these stacks strive to make their films as perfectly as possible so the models are as accurate as possible.

As outlined above, many thin films are usually designed in a stack to produce complex spectra and small variations in deposition conditions make it difficult to accurately model in situ film spectra for feedback control of a continuous deposition process because it is practically impossible to obtain full knowledge of the detailed structure of the stack from reflectance, transmittance, ellipsometry, mass balance or other methods. Thus, a thin film interference filter is needed that is less difficult to manufacture, which will address varying refractive indices of thin films and varying absorptions with deposition parameters.

BRIEF SUMMARY OF INVENTION

In general, the present invention is directed to a layered, thin film interference filter and related bootstrap methods. A bootstrap method according to one aspect of the invention permits a user to focus on a single layer of a film stack as the layer is deposited to obtain an estimate of the properties of the stack. Although the single layer model is a guideline and not a basis for compensating errors, only the most-recently-deposited layer—and not the already-deposited film stack—need be modeled according to an aspect of the present invention. Thus, the user can neglect deviations of the stack from ideality for all other layers. The single-layer model can then be fit exactly to the observed spectra of the film stack at each stage of deposition to allow accurate updating of the remaining film stack for continued deposition.

The present invention works with any type of films, whether absorbing or non-absorbing, and regardless of whether the control of the deposition conditions are state-of-the-art or not. The methods of the invention are relatively straightforward, and a resultant thin film interference filter is economical to produce and use.

According to a particular aspect of the invention, a method using experimental measurements to determine reflectance phase and complex reflectance for arbitrary thin film stacks includes the steps of determining reflectance of a stack of a plurality of films before depositing a topmost layer; considering a modeled monitor curve for a wavelength of a high-index layer; and discarding a plurality of monitor curves without maxima in their reflectance during the topmost layer deposition. In this aspect of the invention, the topmost layer can be a niobia layer.

The exemplary method can also include the steps of determining an anticipated standard deviation in φ_k for a plurality of monitor wavelengths in the niobia layer and discarding any with σ greater than 0.9 degrees. Another step in this aspect is computing expected error in δ for wavelengths with σ less than 0.9 degrees at a target thickness of the niobia layer. When no wavelengths have an error less than 0.9 degrees, a further step according to this method is to proceed with a full model deposition of the niobia layer. When there are no maxima in the reflectance of each of the plurality of monitor curves, another step according to the exemplary method is to use only the modeled monitor curve during the topmost layer deposition.

The method according to this aspect of the invention can also include the step of computing a value of δ for all wavelengths based on a value calculated for the monitor wavelength.

The method according to this aspect of the invention can further include the step of computing two possible values of phase angle for each wavelength other than the monitor wavelength.

Additional steps according to the exemplary method include using information extracted from the model for r_k at each wavelength and the computed best value of δ, and computing an estimated standard deviation of phase at all wavelengths except the monitor.

The method may further include the steps of using the computed phase closest to the model phase for r_k at each wavelength, measured R_f and R_k values and the computed best value of δ, and computing the estimated standard deviation of phase at all wavelengths for which the magnitude of r_k was estimated other than the monitor.

Further steps according to the exemplary method include the steps of determining if a phase error estimate is less than about 1.3 degrees and averaging calculated and modeled reflectance and phase values to obtain a new value for use in subsequent modeling at that wavelength.

In yet another aspect according to the exemplary method, the topmost layer can be a silica film. Accordingly, the method can include the step of replacing the magnitude of the amplitude reflectance at each wavelength with √{square root over (R_k)} whenever measuring a latest depositing silica film having an intensity reflectance greater than 9%. The method can also include the step of determining if a phase error estimate is less than about 1.3 degrees when the magnitude of the amplitude reflectance at each wavelength has been replaced with √{square root over (R_k)} and averaging calculated and modeled reflectance and phase values to obtain a new value for use in subsequent modeling at that wavelength.

In yet another aspect of the invention, a method for correcting thin film stack calculations for accurate deposition of complex optical filters can include the steps of determining phase angle φ_k at a monitor wavelength from |r′_k| and Rk using a first equation expressed as: $\cos \left(\pm {\varphi }_k\right)=\frac{{r_k^{\prime }}^2\left(1+R_k{r_2}^2\right)-{r_2}^2-R_k}{2r_2\sqrt{R_k}\left(1-{r_k^{\prime }}^2\right)};$

and estimating r′_k using a second equation $r_k^{\prime }=\frac{r_k-r_2}{1-r_2{r}_k};$
and obtaining a value for phase at a monitor wavelength.

Still another aspect of the invention includes a method for automated deposition of complex optical interference filters including the steps of determining from a measurement of intensity reflectance at a topmost interface a phase angle φ at an interface k according to an equation expressed as: $\cos \left({\varphi }_k\right)=\left(\frac{A\left(1+r_2^2\right)\sin \left(\delta \right)\pm B\cos \left(\delta \right)}{C}\right)$ $\begin{array}{c}A=R_f+r_2^4\left(R_f-R_k\right)-R_k+\\ 2r_2^2\left(\left(1-R_f\right)\left(1+R_k\right)\cos \left(2\delta \right)-\left(1-R_f{R}_k\right)\right)\end{array}$ $B=\sqrt{D\left(1+r_2^{12}\right)+F\left(r_2^2+r_2^{10}\right)+G\left(r_2^4+r_2^8\right)+{\mathrm{Hr}}_2^6}$ $C=\sin \left(\delta \right)\left(4r_2\left(1-R_f\right)R_k^{1/2}\left(2r_2^2\cos \left(2\delta \right)-1-r_2^4\right)\right)$ $D=-{\left(R_f-R_k\right)}^2$ $F=2\left(\begin{array}{c}{R}_k\left(2+R_k\right)+R_f^2\left(1+2R_k\right)+2R_f\left(1-5R_k+R_k^2\right)-\\ 2\left(1-R_f\right)\left(1-R_k\right)\left(R_f+R_k\right)\cos \left(2\delta \right)\end{array}\right)$ $\begin{array}{c}G=-6-4R_f-5R_f^2-4R_k+38R_f{R}_k-4R_f^2{R}_k-5R_k^2-\\ 4R_f{R}_k^2-6R_f^2{R}_k^2+8\left(1-R_f^2\right)\left(1-R_k^2\right)\cos \left(2\delta \right)-\\ 2{\left(1-R_f\right)}^2{\left(1-R_k\right)}^2\cos \left(4\delta \right)\end{array}$ $H=4\left(\begin{array}{c}\begin{array}{c}3+2R_f^2-10R_f{R}_k+2R_k^2+3R_f^2{R}_k^2-\\ 2\left(1-R_f\right)\left(1-R_k\right)\left(2+R_f+R_k+2R_f{R}_k\right)\cos \left(2\delta \right)+\end{array}\\ {\left(1-R_f\right)}^2{\left(1-R_k\right)}^2\cos \left(4\delta \right)\end{array}\right)$

According to this exemplary method, a process control for a deposition system is bootstrapped by detaching the deposition system from all but the topmost interface.

The method can further include the step of validating two resultant solutions according to the expression: $R_f=\frac{2r_2^2+R_k\left(1+r_2^4\right)+2r_{2}Q}{1+r_2^4+2r_2^2{R}_k+2r_{2}Q}$ $\begin{array}{c}Q=r_2^2{R}_k^{1/2}\cos \left(2\delta +{\varphi }_k\right)+R_k^{1/2}\cos \left(2\delta -{\varphi }_k\right)-\\ r_2\left(1+R_k\right)\cos \left(2\delta \right)-\left(1+r_2^2\right)R_k^{1/2}\cos \left({\varphi }_k\right)\end{array}$

The method can also include the step of averaging calculated and modeled reflectance and phase values to obtain a new value to be used in all future modeling at a given wavelength.

According to another aspect of the invention, a thin film interference filter system includes a plurality of stacked films having a determined reflectance; a modeled monitor curve; and a topmost layer configured to exhibit a wavelength corresponding to one of the determined reflectance or the modeled monitor curve, the topmost layer being disposed on the plurality of stacked films. The topmost layer according to this aspect can be a low-index film such as silica or a high index film such as niobia.

Other aspects and advantages of the invention will be apparent from the following description and the attached drawings, or can be learned through practice of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

A full and enabling disclosure of the present invention, including the best mode thereof to one of ordinary skill in the art, is set forth more particularly in the remainder of the specification, including reference to the accompanying figures in which:

FIG. 1 is a schematic, cross sectional view of a stack of films according to an aspect of the invention;

FIG. 2 is a schematic, cross sectional view of a stack of films similar to FIG. 1 showing an internal interface in the stack of films according to another aspect of the invention;

FIG. 3 is a schematic, cross sectional view of a stack of films according to another aspect of the invention similar to FIG. 1 but showing amplitude reflectance of a buried interface;

FIG. 4 is a representation of regions of permissible values of Rk and Rmax, particularly showing lowest values of σ_φk for silica according to an aspect of the invention;

FIG. 5 is similar to FIG. 4 but for niobia according to another aspect of the invention;

FIG. 6 is a histogram for silica as in FIG. 4;

FIG. 7 is a histogram for niobia as in FIG. 5;

FIG. 8 is a perspective plot showing an error logarithm versus Rmax and δ in accordance with an aspect of the invention; and

FIG. 9 is a histogram of FIG. 8 data.

DETAILED DESCRIPTION OF THE INVENTION

Detailed reference will now be made to the drawings in which examples embodying the present invention are shown. Repeat use of reference characters in the drawings and detailed description is intended to represent like or analogous features or elements of the present invention.

The drawings and detailed description provide a full and detailed written description of the invention and the manner and process of making and using it, so as to enable one skilled in the pertinent art to make and use it. The drawings and detailed description also provide the best mode of carrying out the invention. However, the examples set forth herein are provided by way of explanation of the invention and are not meant as limitations of the invention. The present invention thus includes modifications and variations of the following examples as come within the scope of the appended claims and their equivalents.

Turning now to the figures, FIG. 1 shows a thin film interference filter 10, which broadly includes a substrate 12 upon which a stack of films 14A-X is deposited (where x represents a theoretically infinite number of film layers). As shown, the last (alternatively, final, top or topmost) deposited film is designated by the alphanumeral 14A while previously deposited or lower level films are designated 14B-X. An incoming ray 18 is shown in FIG. 1 being reflected at an interface 16 (also referred to herein as top or top surface and when mathematically referenced as k). The reflected ray is designated by the number 20. For simplicity, any contribution from multiple incoherent reflections in the substrate 12 is ignored in the following discussion and only reflections with respect to the film stack 14A-X are described.

Typically, reflectance of the top surface 16 is obtained using a matrix calculation that is in turn built from the characteristic matrices of each of the preceding films 14A-X. As shown in FIG. 1, a computed value of an electric field amplitude reflectance r_k is obtained by optimizing the thickness of the topmost film 14A to provide the best fit over the fall spectrum consistent with an understanding of the existing film stack 14B-X. The calculated value of r_k can be considered an estimate of the actual value of r_k exhibited by the film stack 14A-X. Since r_k is a complex value, it cannot be measured directly.

Normal Matrx-type Calculations:

For standard calculations, the complex reflectance of a stack of films is computed using the admittances of the incident medium (often air), and the first interface of the stack. For s and p polarization, this reflectance is: $\begin{array}{cc}{r}_s^{\left(f_1\right)}=\frac{-{\eta }_s^{\left(\mathrm{inc}\right)}-{\eta }_s^{\left(f_1\right)}}{-{\eta }_s^{\left(\mathrm{inc}\right)}+{\eta }_s^{\left(f_1\right)}}{r}_p^{\left(f_1\right)}=\frac{{\eta }_p^{\left(\mathrm{inc}\right)}-{\eta }_p^{\left(f_1\right)}}{{\eta }_p^{\left(\mathrm{inc}\right)}+{\eta }_p^{\left(f_1\right)}}& 6.\end{array}$
where the subscript s or p indicates for s or p-polarized light, η is a complex admittance, the superscript “inc” indicates the incident medium and (f₁) indicates the first interface the light strikes when coming from the incident medium.

For s and p polarized light, the admittances of the incident media are written: $\begin{array}{cc}{\eta }_s^{\left(\mathrm{inc}\right)}=\sqrt{{\varepsilon }_{\mathrm{inc}}-{\varepsilon }_{\mathrm{inc}}{\sin }^2\left({\theta }_{\mathrm{inc}}\right)}{\eta }_p^{\left(\mathrm{inc}\right)}=\frac{{\varepsilon }_{\mathrm{inc}}}{\sqrt{{\varepsilon }_{\mathrm{inc}}-{\varepsilon }_{\mathrm{inc}}{\sin }^2({\theta }_{\mathrm{inc}})}}& 7.\end{array}$
where ∈ indicates a complex dielectric constant, and θ_inc is the angle of incidence in the medium of incidence.

The problematic part of the calculation is how to express the admittance of the initial interface. The matrix calculation proceeds by relating the admittance of the initial interface to that of the second interface, the admittance of the second to the third, etcetera, through a series of 2×2 matrices, until the calculation is related to the final interface. At the final interface, the admittance (ratio of magnetic to electric fields) is equal to the admittance of the exit medium, which is simple to compute because there is only a single ray (the transmitted ray), rather than rays propagating in two different directions.

For a single layer stack the admittance of the initial interface is related to the exit medium admittance according to the following equations: $\begin{array}{cc}\begin{array}{c}\left(\begin{array}{c}{E}_x^{\left(f_1\right)}\\ H_y^{\left(f_1\right)}\end{array}\right)=\left(\begin{array}{cc}{s}_{11}& s_{12}\\ s_{21}& s_{22}\end{array}\right)\left(\begin{array}{c}{E}_x^{\left(f_2\right)}\\ H_y^{\left(f_2\right)}\end{array}\right)\\ =\left(\begin{array}{cc}{s}_{11}& s_{12}\\ s_{21}& s_{22}\end{array}\right)\left(\begin{array}{c}{E}_x^{\left(3\right)}\\ H_y^{\left(3\right)}\end{array}\right)\therefore \left(\begin{array}{c}{E}_x^{\left(f_1\right)}/E_x^{\left(3\right)}\\ H_y^{\left(f_1\right)}/E_x^{\left(3\right)}\end{array}\right)\\ =\left(\begin{array}{cc}{s}_{11}& s_{12}\\ s_{21}& s_{22}\end{array}\right)\left(\begin{array}{c}1\\ -{\eta }_s^{\left(3\right)}\end{array}\right)\therefore {\eta }_s^{\left(f_1\right)}\\ =\frac{H_y^{\left(f_1\right)}}{E_x^{\left(f_1\right)}}=\frac{s_{11}+s_{12}\left(-{\eta }_s^{\left(3\right)}\right)}{s_{21}+s_{22}\left(-{\eta }_s^{\left(3\right)}\right)}\end{array}& 8.\end{array}$

In these equations, the superscript (3) indicates the exit medium. The admittance of the exit medium for s-polarized light is given by η_s⁽³⁾=√{square root over (∈₃−∈_inc sin²(θ_inc))}. The negative sign in front of η_s⁽³⁾ in the second line results from defining light as propagating in the negative z direction. For s-polarized light, the magnetic and electric fields are of opposite signs in this case. For p-polarized light, they are of the same sign. The 2×2 matrix for the single film is described below.

For p-polarized light, the admittance of the initial interface is arrived at in the same way: $\begin{array}{cc}\begin{array}{c}\left(\begin{array}{c}{E}_y^{\left(f_1\right)}\\ H_x^{\left(f_1\right)}\end{array}\right)=\left(\begin{array}{cc}{p}_{11}& p_{12}\\ p_{21}& p_{22}\end{array}\right)\left(\begin{array}{c}{E}_y^{\left(f_2\right)}\\ H_x^{\left(f_2\right)}\end{array}\right)\\ =\left(\begin{array}{cc}{p}_{11}& p_{12}\\ p_{21}& p_{22}\end{array}\right)\left(\begin{array}{c}{E}_y^{\left(3\right)}\\ H_x^{\left(3\right)}\end{array}\right)\therefore \left(\begin{array}{c}{E}_y^{\left(f_1\right)}/E_y^{\left(3\right)}\\ H_x^{\left(f_1\right)}/E_y^{\left(3\right)}\end{array}\right)\\ =\left(\begin{array}{cc}{p}_{11}& p_{12}\\ p_{21}& p_{22}\end{array}\right)\left(\begin{array}{c}1\\ {\eta }_p^{\left(3\right)}\end{array}\right)\therefore {\eta }_p^{\left(f_1\right)}\\ =\frac{H_x^{\left(f_1\right)}}{E_y^{\left(f_1\right)}}=\frac{p_{11}+p_{12}{\eta }_p^{\left(3\right)}}{p_{21}+p_{22}{\eta }_p^{\left(3\right)}}\end{array}& 9.\end{array}$

For p-polarized light, the admittance of the exit medium is written as η_p⁽³⁾=∈₃/√{square root over (∈₃−∈_inc sin²(θ_inc))}. The 2×2 matrices for s and p polarizations are defined by: $\begin{array}{cc}{s}_{12}=s_{22}=p_{11}=p_{22}=\cos \left({\delta }_{\mathrm{film}}\right)s_{12}=\frac{-i}{{\eta }_s^{\left(\mathrm{film}\right)}}\sin \left({\delta }_{\mathrm{film}}\right)s_{21}=-i{\eta }_s^{\left(\mathrm{film}\right)}\sin \left({\delta }_{\mathrm{film}}\right)p_{12}=\frac{i}{{\eta }_p^{\left(\mathrm{film}\right)}}\sin \left({\delta }_{\mathrm{film}}\right)p_{21}=i{\eta }_p^{\left(\mathrm{film}\right)}\sin \left({\delta }_{\mathrm{film}}\right)& 10.\end{array}$

In these equations, the admittances of the film are written in the same form as the admittances of the exit medium given above, but with ∈film replacing ∈3. The value δfilm is the phase thickness of the film, given by $\begin{array}{cc}{\delta }_{\mathrm{film}}=\frac{2\pi d_{\mathrm{film}}\sqrt{{\varepsilon }_{\mathrm{film}}-{\varepsilon }_{\mathrm{inc}}{\sin }^2\left({\theta }_{\mathrm{inc}}\right)}}{{\lambda }_0}& 11.\end{array}$
where dfilm is the physical thickness of the film and λ0 is the free-space wavelength of the incident light.

If there are multiple films, the matrix for a stack of films is obtained from $\begin{array}{cc}S=\left(\begin{array}{cc}{s}_{11}& s_{12}\\ s_{21}& s_{22}\end{array}\right)=\prod _{\mathrm{inc}}^{\mathrm{exit}}{S}_{\mathrm{films}}& 12.\end{array}$
where the product is over the 2×2 matrices of each individual film from the entrance to the exit. The final product matrix is used as though it described a single equivalent layer.

Notably in the preceding calculation, the matrices describing the film are used as “transfer” matrices. This permits propagation of the calculation of the admittance of the initial interface down through a stack of films. The downward propagation is stopped at the substrate because, once there are no longer rays propagating in both directions, a simple form (the admittance of the exit medium) can be written. Thus, the matrices allow an impossible calculation to be related to a simplified calculation via a 2×2 matrix.

A critical piece of understanding results from the following discussion. Referring to Equation 8 above for s polarization, −η_s⁽³⁾, is the admittance of the final interface (thus the basis for preserving the negative sign). In comparison, p polarization, η_p⁽³⁾, in Equation 6 is the admittance of the final interface. Thus, both the s and p calculation of admittance for the initial interface can be written in a generic form: $\begin{array}{cc}{\eta }^{\left(f_1\right)}=\frac{H^{\left(f_1\right)}}{E^{\left(f_1\right)}}=\frac{m_{11}+m_{12}{\eta }^{\left(f_{\omega }\right)}}{m_{21}+m_{22}{\eta }^{\left(f_{\omega }\right)}}& 13.\end{array}$
where the matrix elements are for either the s or p matrices, and η^(fω) is the admittance of the final interface. The final interface is always chosen because a simple expression for its admittance can be written in terms of the admittance of the exit medium.

A Bootstrap Method according to an aspect of the invention depends on finding experimental values for the complex reflectance at a given interface in a film stack. Thus, an initial matter of using the amplitude reflectance of a film surface alone to complete the matrix calculation for the films above the film surface in question will be described.

Solving the Top of the Stack

Turning now to a problem illustrated in FIG. 2, reflectance at some interface 160 (mathematically, k′) below a film stack 115A-X is assumed. Earlier layers 114A-X are shown “grayed out” and with diagonal lines to indicate a vague idea of what those layers 114A-X are. The reflectance of the “known” layer 114A differs in value from the reflectance described above (thus, a “prime” symbol on k at the interface 116 indicates the different value). However, it is still desirable to be able to compute the reflectance spectrum of the stack of films 115A-X layered on top of the “known” layer 114A.

It is possible to compute the spectrum of a film stack when the amplitude reflectance at one interface at the bottom is known. To appreciate how this calculation is done, it is useful to review the known method for calculating reflectance and understand what the assumptions are.

If the amplitude reflectivity for an interface k (which can be any interface, including the final interface) is known, an equivalent expression to Equation 13 can be obtained in terms of the admittance of that interface in lieu of carrying the calculation all the way down to the substrate. The admittance for the k^th interface can be written as follows: $\begin{array}{cc}\mathrm{for}s\text{:}\left\{\begin{array}{c}\left(\begin{array}{c}{E}_x^{\left(f_k\right)}\\ H_y^{\left(f_k\right)}\end{array}\right)=\left(\begin{array}{c}\left(r_k^{\prime }+1\right)E_x^{\left(i_2\right)}\\ \left(r_k^{\prime }-1\right){\eta }_s^{\left(2\right)}{\hat{E}}_x^{\left(i_2\right)}\end{array}\right)\\ {\eta }_s^{\left(f_k\right)}={\eta }_s^{\left(2\right)}\frac{\left(r_k^{\prime }-1\right)}{\left(r_k^{\prime }+1\right)}\end{array}\right\}\mathrm{for}p\text{:}\left\{\begin{array}{c}\left(\begin{array}{c}{E}_y^{\left(f_k\right)}\\ H_x^{\left(f_k\right)}\end{array}\right)=\left(\begin{array}{c}\left(1+r_k^{\prime }\right)E_y^{\left(i_2\right)}\\ \left(1-r_k^{\prime }\right){\eta }_p^{\left(2\right)}{E}_y^{\left(i_2\right)}\end{array}\right)\\ {\eta }_p^{\left(f_k\right)}={\eta }_p^{\left(2\right)}\frac{\left(1+r_k^{\prime }\right)}{\left(1-r_k^{\prime }\right)}\end{array}\right\}& 14.\end{array}$

In this expression for the admittance of the known interface, “2” indicates the admittance of the film deposited directly on the known interface. This can then used as the starting point for computing the reflectance of a stack of films above the known interface. The admittance of the top interface can be written as: $\begin{array}{cc}{\eta }_s^{\left(f_1\right)}=\frac{H_y^{\left(f_1\right)}}{{\hat{E}}_x^{\left(f_1\right)}}=\frac{s_{21}^{\prime }+s_{22}^{\prime }{\eta }_s^{\left(f_k\right)}}{s_{11}^{\prime }+s_{12}^{\prime }{\eta }_s^{\left(f_k\right)}}=\frac{s_{21}^{\prime }+s_{22}^{\prime }{\eta }_s^{\left(2\right)}\left(\frac{r_k^{\prime }-1}{r_k^{\prime }+1}\right)}{s_{12}^{\prime }+s_{12}^{\prime }{\eta }_s^{\left(2\right)}\left(\frac{r_k^{\prime }-1}{r_k^{\prime }+1}\right)}{\eta }_p^{\left(f_1\right)}=\frac{H_x^{\left(f_1\right)}}{{\hat{E}}_y^{\left(f_1\right)}}=\frac{p_{21}^{\prime }+p_{22}^{\prime }{\eta }_p^{\left(f_k\right)}}{p_{11}^{\prime }+p_{12}^{\prime }{\eta }_p^{\left(f_k\right)}}=\frac{p_{21}^{\prime }+p_{22}^{\prime }{\eta }_p^{\left(2\right)}\left(\frac{1-r_k^{\prime }}{1+r_k^{\prime }}\right)}{p_{11}^{\prime }+p_{12}^{\prime }{\eta }_p^{\left(2\right)}\left(\frac{1-r_k^{\prime }}{1+r_k^{\prime }}\right)}& 15.\end{array}$

In addition to the change in definition for the terminal interface of the calculation, another difference is the matrix elements come from a modified 2×2 matrix for the film stack. The modified matrix is computed as: $\begin{array}{cc}{S}^{\prime }=\left(\begin{array}{cc}{s}_{11}^{\prime }& s_{12}^{\prime }\\ s_{21}^{\prime }& s_{22}^{\prime }\end{array}\right)=\prod _{\mathrm{inc}}^{k+}{S}_{\mathrm{films}}{P}^{\prime }=\left(\begin{array}{cc}{p}_{11}^{\prime }& p_{12}^{\prime }\\ p_{21}^{\prime }& p_{22}^{\prime }\end{array}\right)=\prod _{\mathrm{inc}}^{k+}{P}_{\mathrm{films}}& 16.\end{array}$
where the product is taken in the order of incident light penetrating the stack as before, but the calculation ends with the film deposited directly onto the known interface. The symbol k+ in Equation 16 is used to indicate that the product terminates with a layer 115A directly above the known interface 114A. The change is a significant one, in that the 2×2 matrices for any of the films below the known interface 114A no longer have to be computed.

Returning to Equation 6, the following can be expressed: $\begin{array}{cc}{r}_s^{\left(f_1\right)}=\frac{\begin{array}{c}{\eta }_s^{\left(\mathrm{inc}\right)}{s}_{11}^{\prime }\left(1+r_k^{\prime }\right)-{\eta }_s^{\left(\mathrm{inc}\right)}{\eta }_s^{\left(2\right)}{s}_{12}^{\prime }\left(1-r_k^{\prime }\right)+\\ s_{21}^{\prime }\left(1+r_k^{\prime }\right)+{\eta }_s^{\left(2\right)}{s}_{22}^{\prime }\left(1-r_k^{\prime }\right)\end{array}}{\begin{array}{c}{\eta }_s^{\left(\mathrm{inc}\right)}{s}_{11}^{\prime }\left(1+r_k^{\prime }\right)-{\eta }_s^{\left(\mathrm{inc}\right)}{\eta }_s^{\left(2\right)}{s}_{12}^{\prime }\left(1-r_k^{\prime }\right)-\\ s_{21}^{\prime }\left(1+r_k^{\prime }\right)+{\eta }_s^{\left(2\right)}{s}_{22}^{\prime }\left(1-r_k^{\prime }\right)\end{array}}{r}_p^{\left(f_1\right)}=\frac{\begin{array}{c}{\eta }_p^{\left(\mathrm{inc}\right)}{p}_{11}^{\prime }\left(1+r_k^{\prime }\right)+{\eta }_p^{\left(\mathrm{inc}\right)}{\eta }_p^{\left(2\right)}{p}_{12}^{\prime }\left(1-r_k^{\prime }\right)-\\ p_{21}^{\prime }\left(1+r_k^{\prime }\right)-{\eta }_p^{\left(2\right)}{p}_{22}^{\prime }\left(1-r_k^{\prime }\right)\end{array}}{\begin{array}{c}{\eta }_p^{\left(\mathrm{inc}\right)}{p}_{11}^{\prime }\left(1+r_k^{\prime }\right)+{\eta }_p^{\left(\mathrm{inc}\right)}{\eta }_p^{\left(2\right)}{p}_{12}^{\prime }\left(1-r_k^{\prime }\right)+\\ p_{21}^{\prime }\left(1+r_k^{\prime }\right)+{\eta }_p^{\left(2\right)}{p}_{22}^{\prime }\left(1-r_k^{\prime }\right)\end{array}}& 17.\end{array}$

Note that these equations feature a complex quantity called r′_k, which, as mentioned above, is not the same as r_k, the amplitude reflectance of the top of the film stack before the topmost layer was added. The two things are related to one another, however, as evident in the following discussion.

Amplitude Reflectance of a Buried Interface

As discussed above, the magnitude of the reflectance of the interface 116 (mathematically, k) in air can be learned by measuring its intensity reflectance but not the phase of the reflectance in the complex plane. With reference to FIG. 3, when the interface 116 (k) is covered by another material 115, the known reflectance is changed. If how the reflectance changes cannot be computed in a simple way, then having learned anything about that reflectance is of no use. Fortunately, there is a straightforward way to relate this to the amplitude reflectance of the interface in air.

In Optical Properties of Thin Solid Films (Dover Publications, Inc., Mineola, USA, 1991), O. S. Heavens gives an expression for the amplitude reflectance of a film in terms of the reflectance of the two interfaces of the film: $\begin{array}{cc}{r}_{\mathrm{film}}=\frac{r_{\mathrm{top}}+r_{\mathrm{bot}}{e}^{-\mathrm{i2\delta }}}{1+r_{\mathrm{top}}{r}_{\mathrm{bot}}{e}^{-\mathrm{i2\delta }}}& 18.\end{array}$
where δ is the optical phase change $\delta =\frac{2\pi d\sqrt{{\varepsilon }_2-{\varepsilon }_{\mathrm{inc}}{\sin }^2\left({\theta }_{\mathrm{inc}}\right)}}{{\lambda }_0},$
directly proportional to the physical thickness of the film. If rtop is replaced with r2 (the Fresnel coefficient for reflectance off an infinite slab of the film material with dielectric constant ∈2), and rbot is allowed to be r′_k, the reflectance of the multilayer stack when the entrance medium is an infinite slab of film material, then the reflectance of the film's top interface can be written as: $\begin{array}{cc}r\left(f_1\right)=\frac{r_2+r_k^{\prime }{e}^{-{\mathrm{i2\delta }}_2}}{1+r_2{r}_k^{\prime }{e}^{-{\mathrm{i2\delta }}_2}}& 19.\end{array}$

When the thickness of the film goes to zero, the exponential equals 1 and the film reflectance must be identical to r_k. This allows r′_k to be solved in terms of r_k as: $\begin{array}{cc}{r}_k^{\prime }=\frac{r_k-r_2}{1-r_2{r}_k}& 20.\end{array}$

This provides an estimate of r′_k that is partially independent of the preceding film stack, since r₂ does not depend on it at all and r_k has been modified, keeping only the phase determined by the film stack calculation.

The Bootstrap Method.

In light of the foregoing introduction, a Bootstrap method for film deposition and refinement is described in the following sections; more particularly, steps to perform Bootstrap refinement of the optical model of a thin film stack are provided as follows.

Step 1. Determine the reflectance of an existing film stack prior to the deposition of a new layer.

The reflectance of a film stack provides some information regarding the complex amplitude reflectance that can be used to refine the model of the reflectance, and that is totally independent of any modeling. If one does not measure reflectance directly, it can be obtained by noting that transmission plus reflectance for an absorption-free thin film stack is unity.

The relationship between the amplitude and intensity reflectance is that the intensity reflectance is the absolute square of the amplitude reflectance. Considering the amplitude reflectance for a moment, it will be clear that it can be expressed in standard Cartesian coordinates on a complex plane, or in complex polar coordinates: $\begin{array}{cc}{r}_k=a+ib=r_ke^{{\mathrm{i\varphi }}_k}r_k=\sqrt{a^2+b^2}{\varphi }_k={\tan }^{-1}\left(\frac{b}{a}\right).& 1\end{array}$

If the amplitude reflectance is expressed in polar coordinates, it is the magnitude of the amplitude reflectance that is provided by a measure of intensity reflectance, |r_k|=√{square root over (R_k)}.

It is possible at this point to replace the magnitude of the amplitude reflectivity in Equation 1, |r_k|, with the square root of the intensity reflectance. This can be done whenever the anticipated error in future reflectance values due to errors in this initial measurement of R_k is expected to be small. How to obtain this relationship is shown in the following.

Calculating the Worst-case Future Reflectance

The standard deviation of the magnitude of the amplitude reflectance is given by Equation 3: $\begin{array}{cc}r_k=\sqrt{R_k}{\varepsilon }_{r_k}=\frac{1}{\sqrt{R_k}}{\varepsilon }_{R_k}{\sigma }_{r_k}=\frac{{\sigma }_{R_k}}{\sqrt{R_k}}.& 3\end{array}$
In the worst-case scenario, Rk is a minimum (the amplitude reflection is on the real axis nearest the origin). This would make the error in magnitude relatively larger. Further, the next layer could result in this vector being advanced by π, crossing the real axis at the furthest point from the origin, producing a reflectance maximum. Again, in the worst case scenario, the maximum reflectance that could be generated as a result of the observed Rk is: $\begin{array}{cc}{R}_{\max }\left(\max \right)={\left(\frac{\left(1+r_2^2\right)R_k^{1/2}-2r_2}{1+r_2^2-2r_2{R}_k^{1/2}}\right)}^2.& 4\end{array}$

To assure that replacing the reflectance amplitude with a measured value does not affect future reflectance measurements by more than the standard deviation of the reflectance measurement, the worst-case scenario must be known. This is: $\begin{array}{cc}{\sigma }_{R_{\max }}\left(\max \right)=\frac{{\left(r_2^2-1\right)}^2\left(1+r_2^2-2r_2{R}_k^{-1/2}\right)}{{\left(1+r_2^2-2r_2{R}_k^{1/2}\right)}^3}{\sigma }_{R_k}.& 5\end{array}$
Solving Equation 5 for a factor of 1σ is not easy, and the result is very complicated. However, a numerical solution is straightforward. For silica (r2=−0.2), this value is about Rk=0.19 or 19% reflectance. For niobia (r2=−0.4), the value is about 9% reflectance. In other words, when about to deposit a silica layer, values of |r_k| should not be adjusted when the measured reflectance is less than 19%. When about to deposit a niobia layer, values should not be replaced when the measured reflectance is less than 9%. Instead, assume the modeled reflectance is more accurate in these cases, although it is not necessary to do this often. In the following sections, reasons are discussed to perform bootstrapping only on high-index layers, so by extension, this step is recommended only when a low-index layer is completed.

• • Step 2: Replace the magnitude of the amplitude reflectance at each wavelength with √{square root over (R_k)} whenever measuring a freshly completed silica film with an intensity reflectance greater than 9% or a low-index film with an intensity reflectance greater than the limiting value of the high-index material.

While the intensity measurement provides useful information (most of the time) about the magnitude of reflectance, it unfortunately provides no information about the phase angle in the complex plane, φ. Much of the remainder of the present description relates to how to obtain these phase angles in at least some circumstances.

Monitor Curves Can Give Non-Redundant Calculations of Phase

In most cases, the magnitude of r_k at the base of a niobia layer can be obtained from the measured reflectance of the film stack terminating in a fresh silica layer. The phase of the amplitude reflectance is more difficult to ascertain, but there are two general approaches. The first is to consider what values of phase are consistent with the final value of reflectance after the next layer is added. To use this information, the optical thickness of the next layer must be known. This is sometimes a redundant calculation since the estimation of optical thickness is usually based on an understanding of the initial reflectance. This is, in fact, a weakness of the usual matrix modeling approach—the calculation is somewhat redundant.

Without additional information, redundant calculation would normally be the only option. However, monitor curves are usually recorded during deposition, and those curves contain all the information necessary to compute the phase, φ_k, without the need for redundant calculations. This involves the use of reflectance maxima in the monitor curves.

• • Step 3. Consider the modeled monitor curve for each wavelength of a niobia (high-index) layer. Discard any monitor curves without maxima in their reflectance during the niobia layer deposition. If none meet this criterion, deposit the layer using a pure model approach.

Based on Equation 14 above, the maxima and minima of a monitor curve can be shown to depend solely on |r′_k|, the magnitude of the buried interface's reflectance, and not at all on its phase. The maximum and minimum reflectance during the monitor curve are given by Equation 21: $\begin{array}{cc}{R}_{\max }={\left(\frac{r_k^{\prime }+r_2}{1+r_2r_k^{\prime }}\right)}^2{R}_{\min }={\left(\frac{r_k^{\prime }-r_2}{1-r_2r_k^{\prime }}\right)}^2.& 21\end{array}$

Therefore, R_max or R_min can be used in the monitor curve to convey the magnitude of the buried reflectance. This magnitude can be related to the reflectance as follows. $\begin{array}{cc}r_k^{\prime }=\frac{\sqrt{R_{\max }}-r_2}{1-r_2\sqrt{R_{\max }}}\mathrm{or}r_k^{\prime }=\frac{\sqrt{R_{\min }}+r_2}{1+r_2\sqrt{R_{\min }}};\frac{r_2-\sqrt{R_{\min }}}{1-r_2\sqrt{R_{\min }}}.& 22\end{array}$
Thus, from a monitor curve covering at least a quarter wave at the monitor curve wavelength, |r′_k| can be determined.

A caveat to using these equations is as follows. The R_min expression has two failings. First, there are two possible solutions for |r′_k| based on R_min depending on whether |r′_k| is less than or greater than |r₂|. If it is less than |r₂|, the right-hand solution is appropriate. If it is greater than |r₂|, the left-hand solution is appropriate. The R_max expression also has two solutions in principle but can be discarded because it provides nonphysical results. The second problem with the equation from R_min is the issue of experimental error. The error expected in the estimation of |r′_k| is related to the error in measurement of R_min and R_max by Equation 23: $\begin{array}{cc}{\varepsilon }_{r_k^{\prime }}=\frac{\left(1-r_2^2\right)}{2\sqrt{R_{\min }}{\left(1+r_2\sqrt{R_{\min }}\right)}^2}{\varepsilon }_{R_{\min }}\approx \frac{{\varepsilon }_{R_{\min }}}{2\sqrt{R_{\min }}}{\varepsilon }_{r_k^{\prime }}=\frac{\left(1-r_2^2\right)}{2\sqrt{R_{\max }}{\left(1-r_2\sqrt{R_{\max }}\right)}^2}{\varepsilon }_{R_{\max }}\approx \frac{{\varepsilon }_{R_{\max }}}{2\sqrt{R_{\max }}}.& 23\end{array}$
In other words, the error goes up as the key reflectance diminishes. Since the minimum is, by definition, smaller than the maximum, the error expected in estimating |r′_k| goes up accordingly. Thus, for both reasons, the calculation of magnitude from a maximum reflectance (i.e., a minimum in the transmission monitor curve) is preferred.
Choosing the Best Monitor Wavelength

The phase angle φk at the monitor wavelength can be determined from |r′_k| and the Rk as: $\begin{array}{cc}\cos \left(\pm {\varphi }_k\right)=\frac{{r_k^{\prime }}^2\left(1+R_k{r_2}^2\right)-{r_2}^2-R_k}{2r_2\sqrt{R_k}\left(1-{r_k^{\prime }}^2\right)}.& 25\end{array}$
This, of course, provides 2 solutions. Once the phase angle φk from Equation 25 is determined, r′_k can be computed using Equation 20 and a monitor curve can be generated if desired for comparison with the actual to help determine which solution is better. Afterwards, a value for phase at the monitor wavelength should have been obtained that is as correct as possible. It depends, of course, on accurately measuring the maximum reflectance value and Rk. Thus, not all wavelengths are created equal as potential monitor wavelengths. A full-spectrum monitor (acquiring many monitor wavelengths) is the best solution, but if only a single wavelength is available, then there is a systematic approach to choosing the best.

Equation 25 depends, ultimately, on only two measurements: the measurement of the initial reflectance and the measurement of the maximum reflectance. For those wavelengths that exhibit a maximum reflectance during deposition, these can be evaluated quantitatively as possible monitor wavelengths.

It can be shown that the anticipated standard deviation of the phase calculation can be written as: $\begin{array}{cc}{\sigma }_{{\varphi }_k}=\sqrt{A{\sigma }_{R_k}^2+B{\sigma }_{R_{\max }}^2}A=\frac{{\left(\left(1+r_2^2\right)R_k+2r_2\left(1+R_k\right)\sqrt{R_{\max }}+\left(1+r_2^2\right)R_{\max }\right)}^2}{4R_k^2\left(\begin{array}{c}4r_2^2{R_k\left(1-R_{\max }\right)}^2-\\ {\left(\begin{array}{c}2r_2\sqrt{R_{\max }}+\left(1+r_2^2\right)R_{\max }-\\ R_k\left(1+r_2^2+2r_2\sqrt{R_{\max }}\right)\end{array}\right)}^2\end{array}\right)}B=\frac{{\left(1-R_k\right)}^2{\left(r_2+\left(1+r_2^2\right)\sqrt{R_{\max }}+r_2{R}_{\max }\right)}^2}{{R_{\max }\left(1-R_{\max }\right)}^2\left(\begin{array}{c}4r_2^2{R_k\left(1-R_{\max }\right)}^2-\\ {\left(\begin{array}{c}2r_2\sqrt{R_{\max }}+\left(1+r_2^2\right)R_{\max }-\\ R_k\left(1+r_2^2+2r_2\sqrt{R_{\max }}\right)\end{array}\right)}^2\end{array}\right)}.& 27\end{array}$

This equation is helpful selecting the best monitor wavelength for the purpose of determining the phase of the amplitude reflectivity at the monitor wavelength.

FIG. 4 shows a representation of the regions of permissible values of Rk and Rmax, with a color code for the lowest values of σ_φk for silica (assuming r2=−0.2 and the standard deviation of the reflectance measurements is 0.003). The lowest value possible under these conditions is 0.0262 radians (1.5 degrees). The lower axis, Rmax, represents possible values of Rmax, while the left axis, Rk, gives possible values of Rk. Note that large regions of reflectance are not possible—there are many combinations of Rk and Rmax that cannot coexist. On the boundaries of those disallowed regions, the error in estimating the phase angle, φk, becomes infinite.

The same plot for niobia, assuming r₂=−0.4, is given in FIG. 5. The minimum value of error here under the same conditions is 0.00875 radians (0.5 degrees)—a much better phase calculation.

It is possible to develop a histogram of the number of combinations of allowed Rmax and Rk that provide a specific level of error in φk. This is accomplished first by considering the range of possible Rmax values when depositing a layer: it cannot, as FIGS. 4 and 5 illustrate, be less than the reflectance of the thin film material being deposited itself (e.g., Rmax for niobia in FIG. 5 cannot be less than −0.4²=0.2 as shown in a leftmost portion of FIG. 5 in dark yellow). Possible values of Rk can be evaluated for each value of Rmax, and the values of phase precision those values provide can be determined using Equation 27. This is accomplished by dividing the phase thickness of the top layer into increments and computing the reflectance at each increment (this is done because the reflectances at the turning points are more likely than those in between the turning point values). In the end, a histogram of the resulting precision values can be formed and a determination made as to how likely each will appear for a given film material. This has been done for silica (assuming r₂=−0.2, and reflectance standard deviations of 0.003) and for niobia (assuming r₂=−0.4) as shown respectively in FIGS. 6 and 7.

For silica, 30% of all observed combinations will have a phase error given by Equation 24 that is less than 2.4 degrees. No values less than about 1.5 degrees error in φk is possible for silica under these conditions. For niobia, the same fraction will have φk errors less than 0.9 degrees, as illustrated in the FIG. 7, also derived from a full numeric simulation.

Thus, the calculated phase error of possible monitor wavelengths will tend to be considerably better for niobia films than for silica. If a limit of 0.9 degrees phase error is placed on the monitors before this calculation is performed, only niobia will give possible monitor wavelengths, and 30% of all wavelengths (overall) will meet this criterion. On some layers, it is possible that no wavelengths will meet this criterion, while on others many may do so.

• • Step 4. For the remaining possible monitor wavelengths in a niobia layer deposition, determine the anticipated standard deviation in φk. Discard any with cy greater than 0.9 degrees (0.016 radians). If none remain, proceed with a pure model deposition.

For low-index layers, bootstrapping is not recommended. For layers with larger magnitudes of r₂ (such as niobia), the precision of the bootstrap is almost always better, but there is no guarantee that a specific layer will include a set of R_k and R_max values anticipated to provide excellent precision in calculating the phase φk. If no wavelength with an anticipated reflectance maximum meets this criterion, modeling alone should be relied upon to deposit the layer until a suitable bootstrap layer is reached.

An ingenious characteristic about this calculation is that it provides φk—and thus also φ′_k—that is consistent with the monitor curve and is independent of δ, the phase thickness of the film. Once a valid solution for rk is determined, the valid solution for r′_k can be obtained. Thus, the final transmission value at the monitor curve wavelength can be used to determine what value of δ is most accurate for the monitor wavelength.

If a wavelength meets the criterion specified above, then consideration should be given as to whether the anticipated end of the layer will have a reflectance, Rf, suitable for estimating δ, the phase thickness of the layer with some precision. This calculation is performed with Equation 26, where δ depends on Rf and rk′. $\begin{array}{cc}\delta =\frac{{\varphi }_k^{\prime }}{2}\pm \frac{1}{2}{\cos }^{-1}\left(\frac{R_f+R_f{r_2}^2{r_k^{\prime }}^2-{r_2}^2-{r_k^{\prime }}^2}{2r_2r_k^{\prime }\left(1-R_f\right)}\right).& 26\end{array}$

This equation provides fairly unique solutions for δ that can be used to correct the reflectance at all wavelengths. The possible solutions for δ that are obtained can be tested against the observed monitor curve to determine which is correct.

For a given monitor curve with a given expectation of Rmax, and with a known value of r2, the sensitivity of δ to errors in Rf can be determined according to Equation 28: $\begin{array}{cc}{\sigma }_{\delta }=\frac{\left(1-r_2^2\right)\left(1-R_{\max }\right)}{2\left(R_f-1\right)\sqrt{A}}{\sigma }_{R_f}A=\left(R_{\max }-R_f\right)\left(\begin{array}{c}2r_2^2\left(R_{\max }{R}_f+R_f-R_{\max }-2\right)-\\ \left(1+r_2^4\right)\left(R_{\max }-R_f\right)+\\ 4\left(r_2+r_2^3\right)\left(1-R_f\right)\sqrt{R_{\max }}\end{array}\right).& 28\end{array}$

If the type of numerical analysis above using Equation 28 is repeated, a plot as shown in FIG. 8 is obtained (shown as the logarithm of error vs. Rmax and δ because otherwise the scale would be difficult to see). This plot is made over the entire range of possible values of Rmax between r₂² and 1 (on the receding axis) and δ angle between 0 and 2π (the front axis).

The data in FIG. 8 can also be rendered as a histogram as shown in FIG. 9. The histogram in FIG. 9 implies that the error in phase thickness is usually satisfactory compared to the error in φk. At a cutoff of 0.9 degrees error in δ, about 51% of the remaining monitor wavelengths for niobia, for instance, should be usable. Thus, given a good monitor wavelength for determining the phase angle φk, there is a good chance of having one that also provides a good precision in δ.

To conserve time, and when modeling is running fairly well, it is reasonable to only select monitor wavelengths for niobia (in a silica/niobia stack), and only when they meet these two criteria (standard deviation of φk<0.9 degrees, and standard deviation of δ<0.9 degrees at the end of the layer). When no monitor wavelengths meet these criteria, it is reasonable to proceed with a pure model matrix approach to depositing the next layer.

• • Step 5. Compute the expected error in δ for the remaining wavelengths at the target thickness of the niobia layer. If no wavelengths have an error less than 0.9 degrees, proceed with a full model deposition of the layer. If some do meet this criterion, select the lowest error in this category.
    Reflectance of a New Film and Determination of the Old Phase at Wavelengths Other than the Monitor Wavelength.

Why bother with determining the φk and δ from a monitor curve as precisely as possible? First, if a monitor wavelength for bootstrapping has been selected successfully, all connection to the previous dependence on the matrix calculation for the monitor wavelength may be avoided. For all other wavelengths, there is at least the opportunity to determine the magnitude of rk, but only the original modeled estimate of phase. The question arises: how to “repair” the phases of all other wavelengths? Since in a single-channel monitor there are no monitor curves at those wavelengths, the phase at each wavelength cannot be directly obtained. (If there was spectrograph recording all wavelengths all of the time, such as with an FTIR system, the steps provided below would not be needed). However, to reach this point in bootstrapping, there must be a good value for δ for the monitor wavelength, plus good values of Rmax and Rk. With this, at least some of the other measurements can be “fixed” to accord with measurements already taken. This will restrict errors to those of a single layer at the worst for those wavelengths where the following calculation is possible.

From δ for the monitor wavelength, the physical thickness of the layer consistent with the modeled refractive index of the film material can be estimated. This physical thickness and the modeled refractive index of the film can be used to estimate δ for all other wavelengths.

• • Step 6. Compute the value of δ for all wavelengths based on the value calculated for the monitor wavelength.

Returning to Equation 16 and replacing rk′ with the definition in Equation 20, the following is obtained: $\begin{array}{cc}{r}_f=\frac{r_2\left(1-r_2{r}_k\right)+\left(r_k-r_2\right)\left(\cos \left(2\delta \right)-i\sin \left(2\delta \right)\right)}{1-r_2{r}_k+r_k\left(r_k-r_2\right)\left(\cos \left(2\delta \right)-i\sin \left(2\delta \right)\right)}.& 29\end{array}$

In Equation 29, the exponential has been replaced with a trigonometric expression using Euler's relation. If r_k is replaced with a+i b, a conventional expression for r_f can be obtained. The complex conjugate of r_f can be formed and the product taken of the two. This provides the intensity reflectance of the top interface in terms of values from the new film, plus a and b. The following can then replace a and b:
a=cos (φ_k)√{square root over (R_k)}
b=sin (φ_k)√{square root over (R_k)}  (30.)
where use is made of the intensity reflectance measured in vacuum for interface k to represent the magnitude of the amplitude reflectance vector at the interface k in vacuum.

The resulting expression can be simplified as Equation 31: $\begin{array}{cc}{R}_f=\frac{2r_2^2+R_k\left(1+r_2^4\right)+2r_{2}Q}{1+r_2^4+2r_2^2{R}_k+2r_{2}Q}Q=r_2^2{R}_k^{1/2}\cos \left(2\delta +{\varphi }_k\right)+R_k^{1/2}\cos \left(2\delta -{\varphi }_k\right)-r_2\left(1+R_k\right)\cos \left(2\delta \right)-\left(1+r_2^2\right)R_k^{1/2}\cos \left({\varphi }_k\right).& 31\end{array}$

In this expression, everything is known EXCEPT the phase angle φ at interface k, allowing it to be determined from the measurement of intensity reflectance at the subsequent interface.

Dispensing with numerical solution methods, this expression can be solved for the Cosine of the angle: $\begin{array}{cc}\cos \left({\varphi }_k\right)=\left(\frac{A\left(1+r_2^2\right)\sin \left(\delta \right)\pm B\cos \left(\delta \right)}{C}\right)A=R_f+r_2^4\left(R_f-R_k\right)-R_k+2r_2^2\left(\begin{array}{c}\left(1-R_f\right)\left(1+R_k\right)\cos \left(2\delta \right)-\\ \left(1-R_f{R}_k\right)\end{array}\right)B=\sqrt{D\left(1+r_2^{12}\right)+F\left(r_2^2+r_2^{10}\right)+G\left(r_2^4+r_8^2\right)+H{r}_6^2}C=\sin \left(\delta \right)\left(4r_2\left(1-R_f\right)R_k^{1/2}\left(2r_2^2\cos \left(2\delta \right)-1-r_2^4\right)\right)D=-{\left(R_f-R_k\right)}^{2}F=2\left(\begin{array}{c}{R}_k\left(2+R_k\right)+R_f^2\left(1+2R_k\right)+2R_f\left(1-5R_k+R_k^2\right)-\\ 2\left(1-R_f\right)\left(1-R_k\right)\left(R_f+R_k\right)\cos \left(2\delta \right)\end{array}\right)G=-6-4R_f-5R_f^2-4R_k+38R_f{R}_k-4R_f^2{R}_k-5R_k^2-4R_f{R}_k^2-6R_f^2{R}_k^2+8\left(1-R_f^2\right)\left(1-R_k^2\right)\cos \left(2\delta \right)-2{\left(1-R_f\right)}^2{\left(1-R_k\right)}^2\cos \left(4\delta \right)H=4\left(\begin{array}{c}3+2R_f^2-10R_f{R}_k+2R_k^2+3R_f^2{R}_k^2-\\ 2\left(1-R_f\right)\left(1-R_k\right)\left(2+R_f+R_k+2R_f{R}_k\right)\cos \left(2\delta \right)+\\ {\left(1-R_f\right)}^2{\left(1-R_k\right)}^2\cos \left(4\delta \right)\end{array}\right)& 32\end{array}$

In principle, Equation 32 can be used to solve for the phase angle. By doing so, the deposition system process control is effectively “bootstrapped” by detaching the system completely from everything that came before the last layer. Equation 32 provides four (4) solutions for the phase angle; two come from the +/− portion of the calculation; two more from the fact that cosine is an even function, so positive and negative angles both work equally well. However, only two of these solutions are consistent with the measured value of R_f. Thus, solutions should be checked via Equation 31 for validity. Only two solutions should be left after this process is complete.

• • Step 7. Compute the two possible values of phase angle for each wavelength other than the monitor wavelength.
    Estimating Error for Non-Monitor Wavelengths

The phase angle is dependent on three reflectivity measurements (R_max, R_f and R_k), those being scrambled together in Equation 32. While this works well for a hypothetical system with no noise, a real spectrometer exhibits errors in measurement of the intensity transmittance.

Based on work already done, the analysis of Equation 32 is fairly straightforward for extending phase information to other wavelengths. The following expression can be constructed from it: $\begin{array}{cc}{\varepsilon }_{{\varphi }_k}=\frac{-1}{\sin \left({\varphi }_k\right)}\left(\left(\frac{\partial \cos \left({\varphi }_k\right)}{\partial R_k}\right){\varepsilon }_{R_k}+\left(\frac{\partial \cos \left({\varphi }_k\right)}{\partial R_f}\right){\varepsilon }_{R_f}+\left(\frac{\partial \cos \left({\varphi }_k\right)}{\partial \delta }\right){\varepsilon }_{\delta }\right){\sigma }_{{\varphi }_k}\approx \frac{{\sigma }_R}{\sin \left({\varphi }_k\right)}{\left({\left(\frac{\partial \cos \left({\varphi }_k\right)}{\partial R_k}\right)}^2+{\left(\frac{\partial \cos \left({\varphi }_k\right)}{\partial R_f}\right)}^2\right)}^{1/2}.& 33\end{array}$

In Equation 33 error in phase thickness has been omitted from the calculation since it is of minor concern at this stage. The phase thickness was settled previously for the sake of argument; therefore, making that approximation, computing the sine of the angle and the sensitivity of the cosine functions to errors in R_k and R_f remain. There is a subtle issue to be considered at this point. What angle should be focused on—the modeled angle or the computed angle using Equations 31/32? In principle, both values are available.

Thus, the conservative answer to the foregoing question is “both”. Assume, for example, that the model has a complex reflectance rk at a given wavelength from which the phase angle φk and the predicted value of Rk can be obtained. Using a best estimate of the phase thickness at the wavelength, Rf can be computed using Equation 31. Next, compute the phase angle using Equation 32 after varying Rk and Rf each by a small amount—e.g., by 10⁻⁵, 1/100^th of a percent transmission. Since the modeled phase is known exactly, it will be trivial to identify which results are the ones closest to the modeled phase; therefore, the sensitivities are computed as: $\begin{array}{cc}\frac{\partial \cos \left({\varphi }_k\right)}{\partial R_k}\approx \frac{\cos \left({\varphi }_k\left(R_{k,m},R_{f,m},\delta \right)\right)-\cos \left({\varphi }_k\left(R_{k,m}-{10}^{-5},R_{f,m},\delta \right)\right)}{{10}^{-5}}& 34\\ \frac{\partial \cos \left({\varphi }_k\right)}{\partial R_f}\approx \frac{\cos \left({\varphi }_k\left(R_{k,m},R_{f,m},\delta \right)\right)-\cos \left({\varphi }_k\left(R_{k,m},R_{f,m}-{10}^{-5},\delta \right)\right)}{{10}^{-5}}.& \end{array}$

From the model, the sine of the phase angle that appears in Equation 33 is trivial to obtain.

Thus, the standard deviation of the phase calculation can be estimated from the model. If the model were absolutely trustworthy, this would be sufficient; however, this is not the case.

• • Step 8. Using information extracted from the model for r_k at each wavelength and the computed best value of δ, compute the estimated standard deviation of phase at all wavelengths except the monitor.

Since the model is not completely trustworthy, the process is repeated using values taken from the calculated value of phase closest to the model value of phase. In other words, for the two possible values of φk, the one closest numerically to the model phase is selected. Now, keying on that value, compute the sensitivities according to Equation 34 using the measured values of R_k, R_f and the estimate of δ for the wavelength being tested.

• • Step 9. Using the computed phase closest to the model phase for r_k at each wavelength, the measured R_f and R_k values and the computed best value of δ, compute the estimated standard deviation of phase at all wavelengths for which the magnitude of r_k was estimated (Step 2) other than the monitor.

Now, compute the estimated error in phase as follows:
σ_φk(estimated)=(σ_φk²(model)+σ_φk²(calculated))^1/2  35.

What are the probable limits to this calculation? A bit of experimentation suggests that the value of standard deviation in phase angle could be between about that of the monitor measurement (on the low end) to nearly infinite. A sufficient approach is to replace the modeled values with calculated values only if a rather conservative cutoff is obtained. If the error estimates are less than about 1.3 degrees from Equation 35 (square root of two times 0.9), and if the reflectance was greater than 9 percent at the start of the layer (Step 2), then the model and calculated values are averaged together to obtain a new estimate. If either the phase error OR the reflectance criteria are not met, the estimate is returned alone to the model.

• • Step 10. If the phase error estimate is less than 1.3 degrees AND the criterion of Step 2 is met, average the calculated and modeled reflectance and phase values to obtain a new value that will be used in all future modeling at that wavelength.

A further refinement of this approach is to fit the phase values obtained above to a Kramers-Kronig model of the phase to fill in values of the phase that have not been determined previously.

While preferred embodiments of the invention have been shown and described, those of ordinary skill in the art will recognize that changes and modifications may be made to the foregoing examples without departing from the scope and spirit of the invention. Furthermore, those of ordinary skill in the art will appreciate that the foregoing description is by way of example only, and is not intended to limit the invention so further described in such appended claims. It is intended to claim all such changes and modifications as fall within the scope of the appended claims and their equivalents.

Claims

1. A method using experimental measurements to determine reflectance phase and complex reflectance for arbitrary thin film stacks, the method comprising the steps of:

determining reflectance of a stack of a plurality of films before depositing a topmost layer;

considering a modeled monitor curve for a wavelength of a high-index layer; and

discarding a plurality of monitor curves without maxima in their reflectance during the topmost layer deposition.

2. The method as in claim 1, wherein the topmost layer is a niobia layer.

3. The method as in claim 2, further comprising the steps of determining an anticipated standard deviation in φk for a plurality of monitor wavelengths in the niobia layer and discarding any with σ greater than 0.9 degrees.

4. The method as in claim 3, further comprising the step of computing expected error in δ for wavelengths with σ less than 0.9 degrees at a target thickness of the niobia layer.

5. The method as in claim 4, further comprising the step proceeding with a full model deposition of the niobia layer when no wavelengths have an error less than 0.9 degrees.

6. The method as in claim 1, further comprising the step of using only the modeled monitor curve during the topmost layer deposition when there is no maxima in the reflectance of each of the plurality of monitor curves.

7. The method as in claim 1, further comprising the step of computing a value of δ for all wavelengths based on a value calculated for the monitor wavelength.

8. The method as in claim 1, further comprising the step of computing two possible values of phase angle for each wavelength other than the monitor wavelength.

9. The method as in claim 1, further comprising the steps of using information extracted from the model for rk at each wavelength and the computed best value of δ, and computing an estimated standard deviation of phase at all wavelengths except the monitor.

10. The method as in claim 9, further comprising the steps of using the computed phase closest to the model phase for rk at each wavelength, measured Rf and Rk values and the computed best value of δ, and computing the estimated standard deviation of phase at all wavelengths for which the magnitude of rk was estimated other than the monitor.

11. The method as in claim 9, further comprising the steps of determining if a phase error estimate is less than about 1.3 degrees and averaging calculated and modeled reflectance and phase values to obtain a new value for use in subsequent modeling at that wavelength.

12. The method as in claim 1, wherein the topmost layer is a silica film.

13. The method as in claim 12, further comprising the step of replacing the magnitude of the amplitude reflectance at each wavelength with √{square root over (Rk)} whenever measuring a latest depositing silica film having an intensity reflectance greater than 9%.

14. The method as in claim 13, further comprising the step of determining if a phase error estimate is less than about 1.3 degrees when the magnitude of the amplitude reflectance at each wavelength has been replaced with √{square root over (Rk)} and averaging calculated and modeled reflectance and phase values to obtain a new value for use in subsequent modeling at that wavelength.

15. A method for correcting thin film stack calculations for accurate deposition of complex optical filters, the method comprising the steps of:

determining phase angle φk at a monitor wavelength from |r′k| and Rk using a first equation expressed as:

cos ⁡ ( ± ϕ k ) =  r k ′  2 ⁢ ( 1 + R k ⁢  r 2  2 ) -  r 2  2 - R k 2 ⁢  r 2  ⁢ R k ⁢ ( 1 -  r k ′  2 );

and

estimating r′k using a second equation

r k ′ = r k - r 2 1 - r 2 ⁢ r k;

and

obtaining a value for phase at a monitor wavelength.

16. A method for automated deposition of complex optical interference filters, the method comprising the steps of:

determining from a measurement of intensity reflectance at a topmost interface a phase angle φ at an interface k according to an equation expressed as:

  ⁢ cos ⁡ ( ϕ k ) = ( A ⁡ ( 1 + r 2 2 ) ⁢ sin ⁡ ( δ ) ± B ⁢   ⁢ cos ⁡ ( δ ) C )   ⁢ A = R f + r 2 4 ⁡ ( R f - R k ) - R k + 2 ⁢ r 2 2 ⁡ ( ( 1 - R f ) ⁢ ( 1 + R k ) ⁢ cos ⁢ ( 2 ⁢ δ ) - ( 1 - R f ⁢ R k ) )   ⁢ B = D ⁡ ( 1 + r 2 12 ) + F ⁡ ( r 2 2 + r 2 10 ) + G ⁡ ( r 2 4 + r 8 2 ) + H ⁢   ⁢ r 6 2   ⁢ C = sin ⁡ ( δ ) ⁢ ( 4 ⁢ r 2 ⁡ ( 1 - R f ) ⁢ R k 1 / 2 ⁡ ( 2 ⁢ r 2 2 ⁢ cos ⁡ ( 2 ⁢ δ ) - 1 - r 2 4 ) )   ⁢ D = - ( R f - R k ) 2   ⁢ F = 2 ⁢ ( R k ⁡ ( 2 + R k ) + R f 2 ⁡ ( 1 + 2 ⁢ R k ) + 2 ⁢ R f ⁡ ( 1 - 5 ⁢ R k + R k 2 ) - 2 ⁢ ( 1 - R f ) ⁢ ( 1 - R k ) ⁢ ( R f + R k ) ⁢ cos ⁡ ( 2 ⁢ δ ) ) G = - 6 - 4 ⁢ R f - 5 ⁢ R f 2 - 4 ⁢ R k + 38 ⁢ R f ⁢ R k - 4 ⁢ R f 2 ⁢ R k - 5 ⁢ R k 2 - 4 ⁢ R f ⁢ R k 2 - 6 ⁢ R f 2 ⁢ R k 2 + 8 ⁢ ( 1 - R f 2 ) ⁢ ( 1 - R k 2 ) ⁢ cos ⁡ ( 2 ⁢ δ ) - 2 ⁢ ( 1 - R f ) 2 ⁢ ( 1 - R k ) 2 ⁢ cos ⁡ ( 4 ⁢ δ )   ⁢ H = 4 ⁢ ( 3 + 2 ⁢ R f 2 - 10 ⁢ R f ⁢ R k + 2 ⁢ R k 2 + 3 ⁢ R f 2 ⁢ R k 2 - 2 ⁢ ( 1 - R f ) ⁢ ( 1 - R k ) ⁢ ( 2 + R f + R k + 2 ⁢ R f ⁢ R k ) ⁢ cos ⁡ ( 2 ⁢ δ ) + ( 1 - R f ) 2 ⁢ ( 1 - R k ) 2 ⁢ cos ⁡ ( 4 ⁢ δ ) )

17. The method as in claim 16, wherein a process control for a deposition system is bootstrapped by detaching the deposition system from all but the topmost interface.

18. The method as in claim 16, further comprising the step of validating two resultant solutions according to the expression:   ⁢ R f = 2 ⁢ r 2 2 + R k ⁡ ( 1 + r 2 4 ) + 2 ⁢ r 2 ⁢ Q 1 + r 2 4 + 2 ⁢ r 2 2 ⁢ R k + 2 ⁢ r 2 ⁢ Q Q = r 2 2 ⁢ R k 1 / 2 ⁢ cos ⁡ ( 2 ⁢ δ + ϕ k ) + R k 1 / 2 ⁢ cos ⁡ ( 2 ⁢ δ - ϕ k ) - r 2 ⁡ ( 1 + R k ) ⁢ cos ⁡ ( 2 ⁢ δ ) - ( 1 + r 2 2 ) ⁢ R k 1 / 2 ⁢ cos

19. The method as in claim 16, further comprising the step of averaging calculated and modeled reflectance and phase values to obtain a new value to be used in all future modeling at a given wavelength.

20. A thin film interference filter system, comprising:

a plurality of stacked films having a determined reflectance;

a modeled monitor curve; and

a topmost layer configured to exhibit a wavelength corresponding to one of the determined reflectance or the modeled monitor curve, the topmost layer being disposed on the plurality of stacked films.

21. The thin film interference filter system as in claim 20, wherein the topmost layer is a silica film.

22. The thin film interference filter system as in claim 20, wherein the topmost layer is a niobia film.