User:Nandktech/Forouhi-Bloomer Equations

Overview
The dispersion of light in any medium can be quantified by two parameters, the extinction coefficient k, and  index of refraction n, which represent the absorption and refraction of light respectively. In 1986 Rahim Forouhi and Iris Bloomer published expressions for n and k (known as the F-B Dispersion Equations), as a function of photon energy E=ħω, that are applicable to amorphous semiconductors and dielectrics thin films. Their work was subsequently extended to the case of crystalline materials in 1988.

Upon publishment, the Forouhi-Bloomer relation was the first model to accurately describe n and k based on quantum mechanics over a wide wavelength range. In practice, the F-B model has been universally applied to describe the n and k of various materials including: insulators, semiconductors, metals, transparent conductors, and polymers, whether amorphous, poly-crystalline, or crystalline.

Equations
The optical dispersion (n(λ) and k(λ)) of an amorphous material can be generally described as a single broad absorption peak due to the short-range order of the atoms that comprise the solid. Conversely, a crystalline material by nature exhibits some degree of long range order which causes discrete absorption peaks within the optical dispersion.

Amorphous Materials
The Forouhi-Bloomer (F-B) equations for n and k of amorphous materials are given as:

$$ k(E) = A\frac{(E-E_g)^2}{E^2-BE+C} \ $$

$$ n(E) = n(\infty)+\frac{(B_0 E + C_0 )}{E^2-BE+C} \ $$

where
 * A is proportional to position matrix element squared
 * B is related to the difference in energies between preferential states in the conduction band and valence band
 * C is related to the lifetime of the excited electron state
 * Eg represents the optical band gap energy
 * n(∞) is the value of n(E) as E → ∞.

The parameters B0 and C0 for n(E) are not independent parameters, but depend on A, B, C, and Eg. They are given by:

$$ B_0 = \frac{A}{Q} \ \left (\frac{-B^2}{2} \ + E_g B - {E_g}^2 + C \right) $$

$$ C_0 = \frac{A}{Q} \ \left ({E_g}^2 + C) \frac{B}{2} \ - 2E_g C \right) $$

where

$$ Q = \frac{1}{2} \ (4C - B^2 )^{0.5} $$

Thus, according to the F-B formalism a total of five parameters are sufficient to fully describe the dependence of both n and k on E or wavelength (λ) since λ can be related to E by Planck's constant.

Crystalline Materials
For materials with multiple absorption peaks in their n and k spectra (i.e. crystalline and polymeric materials) Forouhi-Bloomer obtained:

$$ k(E) = \sum_{i=1}^q \left [ \left (\frac{A_i}{E^2-B_iE+C_i} \ \right) (E - E_{g})^2 \ \right] $$

$$ n(E) = n(\infty)+\sum_{i=1}^q \left [\frac{B_{0i}E+C_{0i}}{E^2-B_iE+C_i} \ \right] $$

Where q is equal to the number of peaks/shoulders in the n and k spectra of the material.

In addition, each term in the sum has its own values of A, B, C, B0, and C0.

Derivation
The F-B equations were obtained from first principles quantum mechanics and solid state physics by deriving an analytical expression for k(E) based on photon absorption of electrons by an amorphous medium, and then incorporating the expression for k(E) into the Kramers-Kronig relationship to determine n(E).

k(E) for Amorphous Materials
The extinction coefficient k can be expressed in terms of the absorption coefficient α(ω) according to:

$$ k=\alpha(\omega)\frac{c}{2\omega} $$

Alternatively, the extinction coefficient can also be expressed as:

$$ \alpha(\omega)=\frac{\phi(\omega)\theta(\omega)}{I_0} $$

where


 * Φ is the probability of an electron transition having finite lifetime τ
 * θ signifies the number of possible transitions per unit volume in a layer of thickness Δx
 * I0 signifies the incident photon intensity

According to the the F-B formalism:

$$ \phi(\omega)=\frac{4\pi}{3c}e^2h\omega I_0 \times \left\vert \left \langle \sigma^*\left\vert \vec{x} \right\vert \sigma \right \rangle \right\vert ^2 \times \frac {\gamma}{(E_{\sigma*}-E_\sigma-h\omega)^2+(h^2\gamma ^2/4)} $$

where


 * e is the electronic charge
 * γ is the inverse of the lifetime (τ) of the excited state
 * σ⟩ represent initial preferred electronic state in the amorphous medium
 * ⟨σ*| represent final preferred electronic state in the amorphous medium

The term │⟨σ*│ ⃗x│σ⟩│2 is the position matrix squared between the states σ and σ* and is related to the strength of the electronic transition from │σ⟩ and ⟨σ│, which in turn relates to the measured amplitude in the k spectrum.

Assuming a complete lack of momentum conservation, the number of possible transitions from the valence to conduction band, θ, will depend on the number of occupied states in the valence band,

$$ \eta_v(E_v)f(E_v)d_v $$

and on the number of unoccupied states in the conduction band,

$$ \eta_c(E_c)[1-f(E_c)]dE_c $$

where


 * ηv(Ev) are the density of states in the valence bands
 * ηc(Ec) are the density of states in the conduction bands
 * f represents the Fermi function

Then,

$$ \theta \propto \int \eta_{\nu}(E_{\nu})f(E_{\nu})\eta_c(E_{\nu}+h\omega)[1-f_c(E_{\nu}+h\omega)dE_{\nu} $$

F-B applied the following assumptions:


 * At ambient temperature, f(Ev) is unity and fc (Ev+ħω) is equal to zero.
 * Density of states in the valence and conduction bands are parabolic in form and can be expressed in terms of photon energy:

$$ \eta_{\nu}(E) \propto \sqrt{(E_{top}-E)} $$

$$ \eta_c(E) \propto \sqrt{(E_{bottom}-E)} $$

where Etop and Ebottom represent the energies at the top and bottom of the valence and conduction bands and their difference, is the optical energy band gap, Eg. Combining the equation for θ with stated assumption yields:

$$ \theta \approx [h\omega - (E_{bottom}-E_{top})^2=[h\omega - E_g]^2 $$

Substituting the equations above yields:

$$ k(E)=Const.\times \frac{2\pi}{3c}e^2h^2 \left\vert \left \langle \sigma^*\left\vert \vec{x} \right\vert \sigma \right \rangle \right\vert ^2 \times \frac {\gamma}{(E_{\sigma*}-E_\sigma-h\omega)^2+(h^2\gamma ^2/4)} \times (E-E_g)^2 $$

By letting

$$ A =Const.\times \frac{2\pi}{3c}e^2h^2 \left\vert \left \langle \sigma^*\left\vert \vec{x} \right\vert \sigma \right \rangle \right\vert ^2 \times \gamma $$

$$ B=2(E_{\sigma*}-E_\sigma) $$

$$ C=(E_{\sigma*}-E_\sigma)^2+(h^2\gamma ^2/4) $$

the analytical form for k(E) is obtained:

$$ k(E) = A\frac{(E-E_g)^2}{E^2-BE+C} \ $$

n(E) for Amorphous Materials
The formulation for n(E) in the F-B dispersion equations was determined from the Kramers-Kronig Dispersion Relation which states that

$$ n(E) - n(\infty) = \frac{2}{\pi} P \int\limits_{0}^{\infty}\frac{k(E')-k(\infty)}{E'-E}dE' $$

where the symbol P represents the principle value of the integral, and n(∞) and k(∞) are the values of n and k in the limit E→∞. The Kramers-Kronig relationship utilizes the concept that n(E) and k(E) represent the real and imaginary parts of the complex index of refraction N(E) stated as:

$$ N(E) = n(E) - ik(E) $$

and N(E) can be continued to complex values of ''E = Ereal + iEimaginary

The integral above can be evaluated along a contour that extends from -∞ to +∞ along the real E axis in the lower half of the complex E plane, closed by an infinite semicircle. The residue theorem can then be invoked which equates the integral on the right hand side to the sum of residues within the contour plus the sum of the residues evaluated at the poles on the real axis (as shown below):

$$ n(E) - n(\infty) = \frac{1}{\pi} P \oint\frac{k(E')-k(\infty)}{E'-E}dE' = -2{\pi}i{\sum}Residues \text{ } - \text{ } i{\pi}{\sum}Residues\text{ }of\text{ }Poles\text{ }on\text{ }the\text{ }Real\text{ }Axis $$

Thus,

$$ \frac{1}{\pi} P \oint\frac{k(E')-k(\infty)}{E'-E}dE' = \frac{1}{\pi} P \int\limits_{-\infty}^{\infty}\frac{k(E')-k(\infty)}{E'-E}dE' + \int\frac{k(E')-k(\infty)}{E'-E}dE' $$

The integral component on the righthand side of the above equation

$$ \int\frac{k(E')-k(\infty)}{E'-E}dE' $$

is evaluated over a semicircle with radius ∞ and vanishes by incorporating the F-B Equation for k(E) and letting E = limE→∞|E|eiθ to obtain,

$$ \frac{1}{\pi} P \oint\frac{k(E')-k(\infty)}{E'-E}dE' = \frac{1}{\pi} P \int\limits_{-\infty}^{\infty}\frac{k(E')-k(\infty)}{E'-E}dE' $$

The integral in the equation above can then be divided into two parts along the real E axis: -∞ to 0 and 0 to ∞ obtaining:

$$ \frac{1}{\pi} P \int\limits_{-\infty}^{\infty}\frac{k(E')-k(\infty)}{E'-E}dE' = \frac{1}{\pi} P \int\limits_{-\infty}^{0}\frac{k(E')-k(\infty)}{E'-E}dE' + \frac{1}{\pi} P \int\limits_{0}^{\infty}\frac{k(E')-k(\infty)}{E'-E}dE'$$

Forouhi and Bloomer discarded the integral along the negative real axis, in which case,

$$ \frac{1}{\pi} P \oint\frac{k(E')-k(\infty)}{E'-E}dE' = \frac{1}{\pi} P \int\limits_{-\infty}^{\infty}\frac{k(E')-k(\infty)}{E'-E}dE' $$

However, it should be noted that by convention, in order to deal with the integral along the negative real axis, N(-E) is contended to be equal to N†(E), where N† is the Hermitian conjugate of N(E); in which case n(-E)= n(E) and k(-E)= -k(E). This convention places a strong restriction on the symmetry of n(E) and k(E). Since k(E) does not satisfy this convention, F-B chose to discard the contribution to n(E) along the negative real axis.

The residue in the lower half complex plane is given as

$$ R = \frac{B}{2}-iQ $$

and the residue at the pole on the real axis is at E.

By combining the equations above the following equation for n(E) is obtained,

$$ n(E) = n(\infty)+\frac{(B_0 E + C_0 )}{E^2-BE+C} \ $$

It is significant to note that F-B do not regard n(∞) to be equal to 1, but rather greater than 1, as opposed to classical dispersion theory which contends that n(∞)=1. However, n(∞)=1 implies no interaction between light and the medium as E→∞. Thus, n(∞)>1 is a result of absorption, that is k(E)≠0, in the limit E→∞.

Furthermore, n(∞)>1 is consistent with John S. Toll’s generalized dispersion relation which is based on the principle of limiting distance. The principle of limiting distance asserts that no signal ever precedes the light cone of its source, and is a generalization of the principle of causality which states that no signal can be transmitted through a medium at a speed greater than that of light in vacuum. According to Toll, mathematics dictates that n(∞)=1+ca where c is the speed of light in vacuum and a is a positive constant.

n(E) and k(E) for Crystalline Materials
The structural peaks and shoulders observed in the optical parameters, k(E) and n(E), of crystalline materials, are thought to be due to the presence of long range order in the solid. To account for the structure, F-B noted that in the amorphous case, the maximum of Φ(ω) (defined as the probability for an electron transition with a finite lifetime) which occurs when ħω≅Eσ* and Eσ= B/2, is very close to the maximum for k(E). F-B stated that a peak in k(E) occurs approximately when Φ(ω) has a local maximum; in which case, the transition corresponds to

$$ h\omega \approxeq E_c(\mathbf{\overrightarrow{k}}_{crit})-E_v(\mathbf{\overrightarrow{k}}_{crit}) $$

where

$$E_c(\mathbf{\overrightarrow{k}}_{crit})$$

$$E_v(\mathbf{\overrightarrow{k}}_{crit})$$

are the energies associated with critical states in the conduction and valence bands of the crystalline solid, and are sources of the prominent structure in the n and k spectra. Symmetry analysis of the Brillouin Zone of the solid will determine which particular states are involved in transitions which produce structure in the spectra (F-B did not specify these states).

F-B argued that k(E) is simply a sum of terms, with each term having the form given by the amorphous case and where the number of terms in the sum is equal to the number of peaks and shoulders in the k(E) spectrum. Thus for the crystalline case

$$ k(E) = \sum_{i=1}^q \left [ \left (\frac{A_i}{E^2-B_iE+C_i} \ \right) (E - E_{g})^2 \ \right] $$

where

$$ A_i=Constant\times\frac{2\pi}{3c}e^2h^2\left\vert \left \langle \psi^c_{crit}\left\vert \vec{x} \right\vert \psi^v_{crit}\right \rangle \right\vert ^2_i \times \gamma_i $$

$$ B_i=2[E_c(\vec{k}_{crit})-E_v(\vec{k}_{crit})]_i $$

$$ C_i=[E_c(\vec{k}_{crit})-E_v(\vec{k}_{crit})]^2_i+\frac{h^2\gamma^2_i}{4} $$

and $$\psi^c_{crit}$$ and $$\psi^v_{crit}$$ denote the electron states in the conduction and valence bands when the wave vector, $$\vec{k}$$ is equal to $$\vec{k}_{crit}$$.

Applying the same arguments given in the derivation of n(E) for amorphous materials, F-B determined for crystalline materials

$$ n(E) = n(\infty)+\sum_{i=1}^q \left [\frac{B_{0i}E+C_{0i}}{E^2-B_iE+C_i} \ \right] $$

where

$$B_{0_i} = \frac{A_i}{Q_i}(-\frac{B^2_i}{2}+EgB_i-Eg^2+C_i)$$

$$C_{0_i} = \frac{A_i}{Q_i} \left [(Eg^2+C_i)\frac{B_i}{2}-2EgC_i \ \right]$$

$$Q_i=\frac{1}{2}(4C_i-B^2_i)^{\frac{1}{2}}$$

where the number of terms, q, in the sum is the number of peaks and shoulders in k(E). Each term in the sum has its own values of A, B, C, B0, and C0. As in the amorphous case, n(∞)>1.

Applying F-B Equations for Characterization of Thin Films by Spectroscopic Reflectometry
From a practical perspective, characterization of thin films plays an important role in semiconductor device fabrication, since thin film layers of varying materials and thicknesses (angstroms to microns), are used as functional constituents of such devices. Characterization of these thin film constituents is crucial for proper control of the many interdependent process steps involved in fabricating semiconductor devices, as well as for development of new functional materials and more efficient device structures. A fundamental characteristic of thin film materials is their optical properties, the index of refraction, n, and the extinction coefficient, k, as a function of the wavelength of light λ. Other primary characteristics of thin films include thickness, surface roughness, interface roughness, and Energy band gap (Eg), which is related to k(λ) and the onset of absorption. In principle, all these film characteristics can be established by the interaction of the film with incident light.

The n and k spectra and Eg of any film are intrinsic characteristics and are unique to the given film. Other characteristics (such as thickness and roughness) are considered extrinsic properties of a film. Variations of a film’s material characteristics (caused by variations in processing conditions) can be directly related to variations in the film’s n and k spectra. However, the n and k spectra cannot be measured directly.

As result, a material’s n and k spectra must be determined indirectly from measurable quantities that depend on n(λ) and k(λ). Both spectroscopic reflectometry (measurement of reflectance as a function of λ), and spectroscopic ellipsometry (measurement of Ψ(λ) and Δ(λ)) are non-destructive techniques that utilize measurable quantities for determining n(λ) and k(λ). For films deposited on transparent substrates (i.e. quartz or sapphire), the transmittance of the film stack T(λ) is another non-destructive measurable quantity that can be used in determining n(λ) and k(λ).

Reflectance R(λ), transmittance T(λ), and ellipsometric parameters (Ψ(λ) and Δ(λ)), depend not only on n(λ) and k(λ) of every film in a stack of films, but also on film thickness, surface and interface roughness, as well as n(λ) and k(λ) of the substrate. All three of these measurement techniques can be used to determine film thickness t in addition to n(λ) and k(λ). Furthermore, surface and interface roughness can also be determined from the scattered, non-specular component of reflectance.

Spectroscopic Reflectance
Prior to the development of the F-B dispersion equations, the industry consensus was that reflectometry could not be used to simultaneously and unambiguously determine n, k and thickness t of a thin amorphous film (due to one measurement quantity and three unknown variables). However, in principle, spectroscopic reflectometry could be used to characterize thin films if a functional form for reflectance in terms of n, k, and t were known.

Based on Fresnel coefficients, general expressions for reflectance of a thin film have been established and provide Rtheoretical in terms of nfilm, kfilm, tfilm, nsubstrate, ksubstrate symbolically expressed as:

$$ R_{theoretical} \equiv R_{theoretical} \left [n_{film}, k_{film}, t_{film}, n_{substrate}, k_{substrate} \right ] \equiv R_{specular} $$

However, the actual form for theoretical reflectance stated above is incomplete because it does not furnish the spectral dependence of n and k as functions of λ. This missing information limited the use of reflectance for thin film characterization. However, provided a physically valid analytical expressions of n(λ) and k(λ), a complete theoretical expression for R(λ) can be established as:

$$ R_{theoretical} \equiv R_{theoretical} \left [n_{film}(\lambda), k_{film}(\lambda), t_{film}, n_{substrate}(\lambda), k_{substrate}(\lambda) \right ] \equiv R_{specular} $$

By fitting the complete theoretical expression for reflectance stated above to measured reflectance, nfilm(λ), kfilm(λ), and tfilm could be simultaneously obtained (typically the optical properties of the substrate are known a priori and are not considered unknowns in the fitting procedure). Furthermore, when the measurement involves multiple films in a stack, the above equation must be expanded to include n(λ), k(λ), and tfilm of each film present. However, in order for spectroscopic reflectometry to function as a viable thin film characterization technique, a set of three conditions must be met:


 * Condition 1: Measured reflectance should cover a wide range of wavelengths that includes deep UV to near infrared wavelengths (190 – 1000 nm)
 * Condition 2: Optimized signal-to-noise ratio of the measured reflectance
 * Condition 3: Measured reflectance must be analyzed using a valid physical model with valid dispersion equations for n(λ) and k(λ) over the wavelengths measured

Prior to 1986, the dependence of n and k on wavelength was only developed for a narrow spectral range, and thus did not satisfy the above requirements and was inadequate for unambiguously determining optical properties and thickness simultaneously. However, Forouhi and Bloomer demonstrated that the F-B dispersion equations could be applied to a variety of thin films measured via spectroscopic reflectometry. This was done by fitting the measured reflectance (collected by near-normal incident spectroscopic reflectometry) to the theoretical reflectance (determined by the physically valid F-B dispersion equations and Fresnel equations). As result, all three conditions stated above are satisfied and the following quantities can be obtained simultaneously:


 * Film thickness
 * n(λ) and k(λ) of the film from 190 – 1000 nm
 * Eg (which is determined from k(λ))
 * Surface and interface roughness

It should be noted that there is no correspondence to the observed shape (peaks and valleys) of Rmeasured with the shape of nfilm(λ) and kfilm(λ). This is because the shape of Rmeasured is dependent on not only the optical properties, but the thickness and roughness of the film too.

Methodology
Characterization of a thin film by spectroscopic reflectometry involves determining the contribution of the following factors to measured reflectance Rmeasured(λ):


 * Material properties, n(λ) and k(λ), of each film comprising a film stack
 * n(λ) and k(λ) of the substrate
 * Eg (determined from k(λ))
 * Thickness of each film t
 * Surface and interface roughness

Single Layer Film Stacks
The methodology for characterizing thin films is based on regression analysis utilizing the Levenberg-Marquardt nonlinear least squares method to fit Rtheoretical to Rmeasured. To help with the convergence of the Levenberg-Marquardt method, nominal values of the parameters to be derived are typically incorporated as starting points for the regression (nominal values are based on the processing conditions the films underwent).

The analytical expressions for n(λ) and k(λ) needed to express Rtheoretical are furnished by the F-B dispersion equations, and are such that the resultant expression for Rtheoretical is highly nonlinear. When dealing with complex films and film structures, the regression may not converge to a set of unique values, due to the non-linear nature of Rtheoretical. Therefore, it may require that the n and k spectra, and/or the thickness, of one or more of the films in the sample to be fixed, while other parameters are allowed to vary in the regression.

Multi-Layer Film Stacks
The measured reflectance of multi-layer film stack depends on the n and k spectra and thickness of each film in the stack. For example, consider the theoretical reflectance of a multilayer stack consisting of three films:

$$ R_{theoretical}\equiv R_{theoretical}[n_{film 1}(\lambda), k_{film 1}(\lambda), t_{film 1}, n_{film 2}(\lambda), k_{film 2}(\lambda), t_{film 2}, n_{film 3}(\lambda), k_{film 3}(\lambda), t_{film 3}, n_{substrate}(\lambda), k_{substrate}(\lambda)] $$

If the thicknesses and n and k spectra of all the films are unknown, then the regression may not converge. To eliminate some of the unknowns, the n and k spectra of one or more of the films may be held fixed during the regression. This is often possible, by incorporating previously measured n and k values of the one or more of the films.

For example, it may be possible to deposit each film individually as a single layer in order to characterize n(λ) and k(λ) independently. Alternatively, a film in the multilayer stack may be well known and the n and k spectra could be obtained from published literature. In either case, the elimination of some of the unknowns can assist the regression in converging on unique, accurate values for the parameters of interest.

Multi-Spectral Analysis of Reflectance
In many instances with the above methodology, the parameters cannot be resolved uniquely due to the non-linearity of Rtheoretical. That is, there are cases whereby the fit to the measured data does not provide unique spectra of nfilm(λ) and kfilm(λ), and other parameters of interest. In order to overcome this problem and constrain the solution to a set of unique values, a technique involving multi-spectral analysis can be used. In the simplest case, this entails depositing the film on two different types of substrates. Then the reflectance spectra of the film on the two different substrates is measured and simultaneously analyzed utilizing the F-B equations and the fitting methodology described above to obtain a unique. The capability of the F-B equations to constrain the parameters and achieve a unique result is a confirmation of their validity.

Determination of Optical Energy Band Gap
The optical energy band gap is distinguished from the electrical energy band gap and is defined as the point where the photon energy is at the onset of absorption. In practice, for thin films there is no absolute energy point, but rather a range of energies, which can be identified as the onset of absorption. Therefore, for practical purposes, and for all the examples presented later, Eg is identified as the point where a film’s absorption coefficient, α, has the value:

$$ \alpha = \frac{4 \pi k}{hc} E = 10^4cm^{-1} $$, E = Eg

where k is determined from the methodology described above.

Determination of Surface and Interface Roughness
The contribution of surface and interface roughness to reflectance, is obtained by introducing a roughness factor, Rroughness, defined by the non-specular component of reflectance, that takes into account the fact that a rough film will scatter light and suppress the measured reflectance. This roughness factor is combined with Rtheoretical, whereby the complete theoretical reflectance is

$$ R_{theoretical} = R_{specular} \times R_{roughness} $$

Rtheoretical is fit to Rmeasured to obtain an average value of roughness over the spot size of the incident beam of light.

Thin Film Examples
The measurement examples contained in this section were chosen based on their significance in the semiconductor industry. These examples showcase the versatility of the F-B equations in characterizing thin films and includes dielectric, semiconductor, transparent conductor, and polymer films deposited on various substrates.

For all examples, the spot size of the incident beam is 50 um for both reflectance and transmittance and is collected across a wavelength range of 190 - 1000 nm. Also, the n(λ) and k(λ) spectra were determined simultaneously along with the optical band gap energy (Eg) and thickness by incorporating the F-B equations into the theoretical expression for reflectance (and transmittance when applicable), and then fitting Rtheoretical and Rmeasured with the methodology described above.

Silicon-Rich Silicon Oxide on Silicon Substrate
Measured (Rexp) and theoretical (Rcalc) reflectance, from 190 - 1000 nm, of single layer Silicon-rich Silicon Oxide deposited on a Silicon substrate is shown as a function of wavelength. The optical dispersion, n(λ) and k(λ) (displayed next to the reflectance spectra), thickness, and Eg were all determined simultaneously from the F-B equations and theoretical expression for reflectance. Using the F-B model, the SiOx film was found to have a thickness of 33 nm with a Eg of 1.55.

Unlike pure silicon dioxide (SiO2) whereby k(λ)=0 from 190 – 1000 nm, for silicon-rich silicon oxide, k(λ) is found to be non-zero in the deep UV wavelength range. This non-zero part of the k(λ) spectrum can be attributed to the extra amount of silicon of the silicon-rich film. In fact, as the silicon concentration increases in the film, so does the value of k in the deep UV region.

Silicon-Rich Silicon Nitride on Silicon Substrate
Similar to the silicon-rich silicon oxide, silicon-rich silicon nitride (SiNx) exhibits a non-zero k value in the DUV. Contained in the graphic is the measured and theoretical reflectance, as well as the results for n(λ), k(λ), thickness, and Eg. These quantities were obtained simultaneously by incorporating the F-B equations into the theoretical expression for reflectance with the methodology described above. In R(λ) plot, interference fringes can be observed in the visible wavelengths indicating a thicker film than described in the silicon-rich SiO2 example. Using the F-B model, the SiNx film was found to have a thickness of 1185 nm with a Eg of 1.53.

Optical Dispersion of Silicon-Rich Silicon Nitride and Oxide on Silicon Substrate Extended into the Vacuum UV
As can be seen in the previous examples, the broad maximum, expected for amorphous materials does not appear in the measured spectral range of 190 – 1000 nm. This is because the maximum occurs in the vacuum UV (VUV) at wavelengths less than 190 nm. This is demonstrated in the graphic, whereby the analytical expressions of n(λ) and k(λ) for the silicon-rich SiOx and silicon-rich SiNx have been extended to VUV wavelengths. In essence, once the F-B dispersion factors (A,B,C,Eg, and n(∞)) have been determined for a film, the resultant expressions for n(λ) and k(λ) can be interpolated to any wavelength range, including the VUV.

Amorphous Silicon on Oxidized Silicon-Substrate
In this example, amorphous silicon (a-Si) has been deposited on an oxidized silicon substrate to form a multilayer film stack consisting of a-Si on SiO2 on silicon substrate. In order to obtain n(λ) and k(λ) of the a-Si layer, the optical properties of the SiO2 were fixed based on literature values (a valid assumption since the SiO2 was thermally grown), and only the thicknesses of both layers and the optical properties of the a-Si were allowed to vary. Using the F-B model, the a-Si film was found to have a thickness of 1147 nm with a Eg of 1.67.

As expected, one board maximum is observed in the n and k spectra. As the long range order of a film increases (i.e. the film becomes more crystalline), the broad maximum gives way to several sharper peaks in the n and k spectra. This is demonstrated in the graphic comparing the optical dispersions of a-Si (no crystallinity), typical poly-Si film (composed of crystalline and amorphous regions), and pure crystalline-Si. With the absence of long range order, the discrete absorption peaks observed in the structure of n and k spectra of the crystalline silicon become washed out and form the single broad curve observed with a-Si.

Titanium Nitride on Aluminum-Substrate
For this example, a TiN film was selected to demonstrate the ability of the F-B equations to determine the optical properties for a material in which k(λ) is non-zero over the entire measured spectral range. Thus, spectroscopic ellipsometry is not a suitable technique for characterizing TiN, unless the thickness of the TiN film has been determined by some alternate technique. However, the F-B dispersion equations are capable of determining the thickness, and optical properties of the TiN film simultaneously without prior knowledge of the film's thickness. In order to determine the optical properties of the TiN film, the optical properties of the Al substrate were fixed based on literature values, and the thickness and n and k spectra of the TiN was allowed to vary. The thickness of the TiN fim was measured to be 30 nm with a Eg of X.

Measurement of Surface Roughness on Polycrystalline Silicon Deposited on Oxidized Silicon
It is expected that poly-Si deposited on oxidize silicon will be rough, because of grain boundaries inherent in the film. Therefore, it is reasonable to include a roughness factor when analyzing poly-Si. The calculated reflectance is determined from specular reflectance combined with a roughness factor based on the scattered light produced by the roughness represented symbolically as,

$$R_{calculated}=R_{specular}\times R_{roughness} $$

In the graphic, the results obtained for Rcalculated=Rspecular×Rroughness labled as “Rough poly-Si” are compared to results obtained without the roughness factor labled as “Smooth poly-Si” (in which case Rcalculated=Rspecular. The fit of the theoretical reflectance to the measured reflectance spectrum is good in both cases, but the behavior of the n and k spectra for the smooth poly-Si is not physically real.  When the roughness factor is not included, the behavior of the n and k spectra in the deep UV wavelength regime obtained for “smooth” poly-Si is inaccurate (circled on the graphic) and is not realistic.  Also, there is a difference in the thickness values obtained for the rough and smooth film.

Note that the n and k spectra of the SiO2 layer and its thickness were held constant in the calculation. Also, the value given for roughness represents an average value over the spot-size of the incident beam which was 50 μm.

Measurement of 193 nm Photoresist Deposited on Silicon
Polymers consist of long chains of molecules which do not form a crystallographic structure. However, their n and k spectra exhibits sharp peaks rather than a broad maximum expected for non-crystalline materials. Thus, the measurement results for a polymer are based on the formulation for crystalline materials.

Contained in the graphic is a measurement example of a photoresist (polymer) material used for 193 nm microlithography. Six terms were needed in the F-B equations in order to accurately describe the polymer's complex optical properties in the DUV. The film was found to have a thickness of 500 nm with a Eg value of X.

Measurement of Indium-Tin-Oxide (ITO) on Glass Substrate
Indium-Tin-Oxide (ITO) is a conducting material with the unusual property that it is transparent in the visible wavelengths (450 - 750 nm). Due to its conductive nature, ITO is widely used in the flat panel display industry. The results presented in the graphic were obtained by first simultaneously measuring reflectance and transmittance from 190 – 1000 nm of the uncoated glass substrate, to determine nglass(λ) and kglass(λ). Then reflectance and transmittance from 190 – 1000 nm of ITO deposited on the same glass substrate were measured simultaneously, and analyzed based on the methodology described above.

As expected, kITO(λ)=0 in the visible wavelength range, since ITO is transparent. The behavior of kITO(λ) in the near-infrared (NIR) to infrared regime resembles that of a metal in that it is non-zero at the NIR range (750 nm <λ< 1000 nm) and reaches a maximum value in the IR range (λ>1000 nm). The behavior of kITO(λ) in the NIR - IR range is shown by extending the analytical form of kITO(λ). Once the factors A, B, C, Eg, and n(∞) have been determined for any film, the resultant expressions for n(λ) and k(λ) can be interpolated to any wavelength range (including the IR).

Measurement of Gallium Nitride with Adhesion Layer on Sapphire Substrate
An important film used for manufacturing light emitting diodes (LED) is gallium-nitride (GaN). The band gap of the film can be controlled by mixing of In or Al to obtain InGaN and AlGaN respectively. Depending on the ratio of In/Al to the GaN, LED manufacturers can control the emitting color. Contained in the graphic are the results of measuring GaN deposited on 0.1 mm sapphire substrate. An adhesion layer which is present for this particular of thin film structure permits the GaN film to adhere to the sapphire substrate and must be taken into account.

The results presented were obtained first by simultaneously measuring reflectance and transmittance from 190 – 1000 nm of the uncoated sapphire substrate to determine nsapphire(λ) and ksapphire(λ). Then reflectance and transmittance from 190 – 1000 nm of GaN/Adhesion Layer/Sapphire-Sub sample are measured simultaneously and analyzed based on the methodology described above. The thickness (640 nm) of both the GaN and Adhesion layer (16 nm) along with the n and k spectra (190 – 1000 nm) of both these layers were simultaneously determined.

Multi-Spectral Analysis of Ge40%Se60% Deposited on Both a Silicon Substrate and Oxidized Silicon Substrate
The single measurement of reflectance from 190 - 1000 nm of Ge40%Se60%/Si-Sub does not provide unique n and k spectra. A unique solution can be achieved by also depositing the Ge40%Se60% film on another substrate, namely oxidized silicon. Then analysis, based on the F-B equations, of the measured reflectance from both Ge40%Se60%/Si-Sub and Ge40%Se60%/SiO2/Si-Sub simultaneously and uniquely determines:


 * Thickness of the Ge40%Se60% film on the silicon substrate
 * Thickness of the Ge40%Se60% film on the on the oxidized silicon substrate
 * Thickness of SiO2 (The n and k spectra of SiO2 is held fixed)
 * n and k spectra (from 190 - 1000 nm) of Ge40%Se60%