Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Optical Properties of Thin Transparent Conducting Oxide Films on Glass for Photovoltaic Applications Mark Stockett Professor John H Scofield, Advisor April 15, 2006 Abstract This project was motivated by my earlier work with John Scofield on thin-film copper indium diselenideide (CIS) photovoltaic devices. We were attempting to fabricate transparent conducting films of tantalum oxide (Ta2 O5 ) for use as a window layer in these devices. Exploring different deposition parameters and doping techniques, many samples were produced that needed characterization to determine their quality and properties. Properties of interest include the film thickness, the real and imaginary parts of the index of refraction, the bandgap energy, and the carrier density in doped films. As a result of this project, we are now able to obtain information on all of these quantities through the study of a single measurement, namely the transmission spectrum. This document summarizes the fabrication work done by me and previous students and details the method I developed for using transmission measurements to characterize the samples. 1 Contents 1 Introduction 1.1 Solar Cell Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 CIS Device Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Alternative Window Materials . . . . . . . . . . . . . . . . . . . . . . . . . . 5 5 5 6 2 Fabrication 2.1 DC Magnetron Reactive Sputtering from Metallic Targets 2.2 Deposition Parameters . . . . . . . . . . . . . . . . . . . . 2.3 Reactive Doping . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Target Doping . . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Routine Characterization . . . . . . . . . . . . . . . . . . . . . . . . 7 7 7 8 9 9 . . . . . 11 11 11 12 13 14 . . . . . . . . 17 17 18 19 20 21 22 22 24 . . . . . 25 25 25 26 27 28 3 Theory 3.1 Statement of Problem . . . . . . . . . . . . . . . 3.2 Maxwell’s Equations and Electromagnetic Waves 3.3 Reflection and Transmission . . . . . . . . . . . . 3.4 Propagation . . . . . . . . . . . . . . . . . . . . . 3.5 Multiple Reflections . . . . . . . . . . . . . . . . . 4 Measurement and Analysis 4.1 UV-Vis-NIR Spectrophotometer 4.2 Refractive Index of Substrate . 4.3 Interference Envelope . . . . . . 4.4 Refractive Index of Film . . . . 4.5 Film Thickness . . . . . . . . . 4.6 Absorption Coefficient . . . . . 4.7 Band Edge . . . . . . . . . . . . 4.8 Absorption in Doped Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Results and Discussion 5.1 Reconstructed Curves . . . . . . . . . . . . . . . . . 5.2 Refractive Indices . . . . . . . . . . . . . . . . . . . 5.3 Optical Versus Mechanical Thickness Measurements 5.4 Bandgap Energies . . . . . . . . . . . . . . . . . . . 5.5 Carrier Densities . . . . . . . . . . . . . . . . . . . 6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 2 List of Figures 1.1 1.2 1.3 2.1 2.2 2.3 2.4 3.1 3.2 3.3 4.1 4.2 4.3 4.4 4.5 The layers of a typical CIS thin film photovoltaic device[1]. The window layer (highlighted in green) was the focus of my research. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The p-n junction. Circled charges are fixed lattice ions. Non-circled charges are free carriers. Both p- and n-type regions are charge neutral. An electron-hole pair created in either region may be split by the electric field in the depletion region and lead to a photocurrent[2]. . . . . Solar blackbody energy density spectrum B(λ) (smooth red curve). The vertical lines indicate the wavelengths corresponding to the bandgap energies of Ta2 O5 , ZnO and CIS. Horizontal arrows show transparancy ranges of Ta2 O5 and ZnO window layers on CIS absorber layer. The Ta2 O5 transmits 11% more of the sun’s energy. . . . . . . . . . . . . . . . . . . . . . . . Left: Top view of vacuum chamber. The positioner can be moved to locate the substrate over either sputter gun. Oxygen is fed in through the front side of the chamber while argon is fed up through the guns. Right: The geometry of the sputter gun (shown in cross-section in lower portion) results in a nonuniform thickness profile (graph at top of frame). Lines show possible paths for metal atoms ejected from the target ring (cathode). . . . . . . . . . . . . . . . . . . Left: X-Ray diffraction (XRD) pattern for a polycrystalline Ta2 O5 film[3]. The location of the peaks is related to the lattice parameters of the crystallites. Right: XRD pattern for an amorphous Ta2 O5 film (sample WT-148). Note that both vertical scales are arbitrary. The pattern for the amorphous film is likely on the order of the background noise of the polycrystalline pattern. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Schematic diagram of four probe resistivity measurement[2]. This setup eliminates errors from lead wire and contact resistances. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Hall effect causes a potential difference VH across a two-dimensional semiconductor related to the carrier density. The applied magnetic field B is out of the page and the current I is indicated by the arrow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Interface between two media with different indices of refraction, showing incident, reflected and transmitted EM waves and angles of incidence and refraction. The z-axis points into the page . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Schematic diagram of a thin film (thickness d) between two infinite media, showing contribution to reflected and transmitted wavefronts from rays reflected within the film multiple times. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The physical situation for thin film measurements in the lab. The film (index n2 ) rests on top of a thick but finite substrate (index n3 ) bounded on either side by air (index n1 ). . . . . . . The “chopper” disk alternates the path of the measurement beam as it rotates. It has three sections: transparent (1; white), mirror (2; grey) and matte (3; black). . . . . . . . . . . . . . Simplified schematic of spectrophotometer. The chopper sends the measurement beam alternately through the sample and not through the sample. . . . . . . . . . . . . . . . . . . . . . Example of raw data from spectrophotometer showing blank substrate (flat upper curve) and ZnO coated sample (oscillating lower curve). Sample Zn-005. . . . . . . . . . . . . . . . . . . Left: Example transmission spectrum of blank substrate. Right: Calculated refractive index of blank substrate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Simulated transmission curve (red) for a 2000 nm thick film, showing interference envelope functions TM (Equation 4.4a) and Tm (Equation 4.4b) in blue. . . . . . . . . . . . . . . . . . 3 5 6 7 8 9 10 10 12 14 15 17 17 18 19 20 4.6 Left: Comparison of interference envelopes built from first (red), second (blue) and third (green) order interpolations. Dashed line is a measured transmission spectrum for sample WT-152. Right: Example of a measured spectrum (green) with linear interference envelope (blue). Sample Zn-005. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.7 Example of extracted refractive (yellow) index with Cauchy fit (red, Equation 4.8). The blue dashed lines show the 3% relative uncertainty in n. Sample Zn-005. . . . . . . . . . . . . . . . 4.8 Graphical method for determining film thickness, interference order `/2 is plotted against n/λ. The slope of the line is twice the film thickness. Sample Zn-005, with m0 = 3/2 and d = 617 nm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.9 Left: Example absorption coefficients calculated from the refractive index n (yellow) and the Cauchy fit to n (red). Blue dashed lines show uncertainty. Right: Example absorption coefficient near band edge. The substrate itself begins absorbing below 250 nm, so data in this region should be ignored. Note the difference in scales on the two plots. Sample Zn-005. 4.10 Left: Plot of α2 vs hν with fit to linear region. The intercept is the bandgap energy. Right: Plot of ln α vs hν with fit to linear region. The slope is the inverse of the Urbach disorder parameter. Sample Zn-005, with Eg = 3.28 eV and E0 = 75 meV. . . . . . . . . . . . . . . . . 4.11 Left: Example transmission spectra for a doped ZnO film (blue) showing infrared absorption and uncoated glass microscope slide (red). Right: By fitting the infrared absorption coefficient (blue) with a λ2 dependence (red), one can determine the ratio of the carrier density N to the scattering time τ . Sample Zn-044 with N/τ = 7.1 × 1035 cm−3 s−1 as determined through Hall effect measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Example of reconstructed transmission curve. The blue curve is raw data and the red curve is the reconstruction. Sample Zn-005. Compare Figure 4.3. . . . . . . . . . . . . . . . . . . . 5.2 Left: If n is taken as a constant, the reconstructed transmission curve (blue curve) fails to capture the variation in the amplitude of the interference fringes present in the data (red curve). Right: If n is allowed to vary, better agreement is achieved, particularly in the fringe amplitude. Sample WT-148. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Close-up of Ta2 O5 transmission curve near band edge (lower curve). The proximity of the Ta2 O5 bandgap to that of the substrate (upper curve) complicates analysis in this region. Sample WT-152. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4 Example of reconstructed transmission curve for a doped film. The blue curve is raw data and the red curve is the reconstruction. Sample Zn-044, with N/τ = 7.1 × 1035 cm−3 s−1 as determined through optical measurements. Compare Figure 4.11. . . . . . . . . . . . . . . . . 20 21 22 23 23 24 25 26 27 28 List of Tables 5.1 5.2 5.3 5.4 Real part of the index of refraction n as extracted with Equation 4.6 and literature values. Uncertainties in parentheses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Comparison of optical (dopt ) and mechanical (dmech ) thickness measurements for Ta2 O5 films. Uncertainties in optical measurements in parentheses. . . . . . . . . . . . . . . . . . . . . . . Measured and literature values of energy bandgap Eg , and Urbach slope E0 for ZnO and In2 O3 films. Uncertainties in parentheses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Values of N/τ as extracted from optical data ((N/τ )opt ) and from Hall effect measurements ((N/τ )Hall ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 26 27 27 28 1 1.1 Introduction Solar Cell Background Solar cells (photovoltaic devices) transform light from the sun into electrical work. Cells based on crystalline silicon have dominated the industry since their development in 1954[2]. In recent years, polycrystalline devices based on materials such as copper indium diselenide (CIS) have gained increased attention from the scientific community. Unlike silicon, these materials may be made into light-weight thin films which are flexible and do not degrade in direct sunlight[2]. Dr Scofield has at times collaborated with scientists at the National Aeronautics and Spaces Administration’s (NASA) Glenn Research Center. NASA’s interest in these devices stems from the significant cost of launching satellites equipped with heavy silicon modules. The efficiency (the fraction of the sun’s energy converted to work) of CIS devices has been shown to be comparable to that for silicon cells (10-15%)[4]. 1.2 CIS Device Components Figure 1.1 shows the layers of a typical CIS solar cell. As thin films, these materials do not have the mechanical strength to be self-supporting, so they must be deposited on some sort of substrate. I sued Corning 2947 plain soda glass microscope slides. On top of the glass substrate, a metallic molybdenum (Mo) layer forms the back contact. Then follows the CIS itself, called the absorber layer, a cadmium sulfide (CdS) buffer layer, and the zinc oxide (ZnO) window layer. ZnO is one of the most common widow layer materials and has been studied extensively by Dr Scofield and his students, but other possibilities may be viable. The top two layers are the aluminum-nickel (Al-Ni) contact grid and a magnesium fluoride (MgF2 ) anti-reflective coating. The Mo back contact and the Al-Ni contact grid define the electrical circuit through which the work done by the photovoltaic device is extracted. Figure 1.1: The layers of a typical CIS thin film photovoltaic device[1]. The window layer (highlighted in green) was the focus of my research. Dr Scofield and his student have studied and fabricated many of these layers. My research focused on the window layer. A solar cell is essentially a p-n junction (see Figure 1.2 and [5]), where a “p-type” layer is joined to an “n-type” layer. At the junction there is a “depletion region” where the oppositely charged carriers annihilate each other and in equilibrium the now unscreened lattice ion cores establish an electric field. Absorption of a photon in the (usually) p-type absorber layer creates an electron-hole pair, which are accelerated in opposite 5 Figure 1.2: The p-n junction. Circled charges are fixed lattice ions. Non-circled charges are free carriers. Both p- and n-type regions are charge neutral. An electron-hole pair created in either region may be split by the electric field in the depletion region and lead to a photocurrent[2]. directions by the depletion region electric field. If the absorber is p-type, the electron from the electronhole pair may diffuse into the depletion region and be accelerated into the n-type window layer, causing an excess negative charge in this layer with an equal and opposite excess charge in the other material from the hole that got left behind. This charge buildup can be collected through the grid and back contacts as a “photocurrent”. In an actual device, one would like the absorption event to occur spatially as close as possible to the junction to maximize the probability that the excited carriers will be collected. This is achieved by putting the window layer on top of the absorber layer so that solar photons passing completely through the window layer will be absorbed as soon as they enter the absorber layer at the junction. This requires that the window layer be highly transparent to photons which the absorber layer can absorb, ie those with energy greater than the absorber layer’s bandgap energy. Creating an effective p-n junction requires that the window layer be highly conducting, which is unfortunately at odds with the transparency requirement, as free charge carriers lead to absorption at infrared wavelengths. Any window layer will also absorb photons with energy greater than its own bandgap energy, so materials with wide bandgaps are needed for efficient window layers. 1.3 Alternative Window Materials Between September 2004 and May 2005, I worked investigating tantalum oxide (Ta2 O5 ) as a possible window layer for CIS solar cells. Ta2 O5 is frequently used for its high resistivity as a dielectric material in capacitors and memory devices[6]. With a wide (4.2 eV[3]) bandgap, a window layer of Ta2 O5 would allow a greater fraction of the sun’s blackbody spectrum to pass through to the absorber layer than would a ZnO (3.3 eV) window (see Figure 1.3). The solar spectrum peaks in the visible but has substantial power in the ultraviolet. Assuming perfect transmission, the wider bandgap of Ta2 O5 would lead to an 11% increase in cell efficiency. Much of this increase would likely be wiped out by ultraviolet absorption in the Earth’s atmosphere, but for space applications this advantage could be significant. While I was fabricating and studying these films, we became interested in their optical properties and thin-film interference effects. Since our motivation for studying this material was the possibility of increased optical transparency, quantifying and understanding the optical properties of our samples was very important. The optical analysis turned out to be complicated and tedious, so I developed a computerized analysis method 6 Figure 1.3: Solar blackbody energy density spectrum B(λ) (smooth red curve). The vertical lines indicate the wavelengths corresponding to the bandgap energies of Ta2 O5 , ZnO and CIS. Horizontal arrows show transparancy ranges of Ta2 O5 and ZnO window layers on CIS absorber layer. The Ta2 O5 transmits 11% more of the sun’s energy. which allowed us to extract all of the relevant parameters in a matter of minutes. This is the subject of Section 4. The most widely used window layer material for silicon solar cells is tin doped indium oxide (ITO). I studied the optical properties of a few indium oxide (In2 O3 ) films, as well as several ZnO films, all deposited by Dr Scofield and his students. 2 2.1 Fabrication DC Magnetron Reactive Sputtering from Metallic Targets Most of the samples I studied were fabricated using the Sloan sputter system located in Wright Laboratory at Oberlin. The system consists of a vacuum chamber equipped with two DC magnetron sputter guns and a moving substrate holder with a heating element. Thin metal-oxide films were deposited onto glass microscope slides using high purity metallic targets (tantalum, zinc, indium etc.) in an argon/oxygen reactive mixture. After the chamber pressure is reduced to 10−6 Torr using a diffusion pump, argon is fed into the base of the gun, while oxygen and other gasses are fed directly into the vacuum chamber. The layout of the vacuum chamber is shown in Figure 2.1 Left. A large negative voltage is applied to an anode disk in the sputter gun, which is mounted in the base plate of the vacuum chamber . This creates an electric field between the anode and the target, an electrically grounded metallic ring 1/8 inch thick with a 3 inch inner diameter, which is a few millimeters larger than the anode disk’s diameter. This field is strong enough to ionize the argon gas. The Ar ions are accelerated into the target ring where they eject metal (in my case tantalum) atoms up towards the substrate, which is held above the gun in the positioner (see Figure 2.1 Right). This is different from many DC sputter systems which use flat targets and sputter downwards. 2.2 Deposition Parameters The deposition process is a dynamic one, with many parameters affecting the resulting film. For making metal-oxides, one of the most important variables is the ratio of argon (Ar) to oxygen (O2 ) in the reactive 7 Figure 2.1: Left: Top view of vacuum chamber. The positioner can be moved to locate the substrate over either sputter gun. Oxygen is fed in through the front side of the chamber while argon is fed up through the guns. Right: The geometry of the sputter gun (shown in cross-section in lower portion) results in a nonuniform thickness profile (graph at top of frame). Lines show possible paths for metal atoms ejected from the target ring (cathode). gas mixture[3]. A higher oxygen fraction leads to higher oxygen incorporation in the film, ie a film that is more metal-oxide than metal. Transparent Ta2 O5 films were made at a 5:1 Ar:O2 flow ratio. The current delivered by the power supply primarily affects the sputter rate, with higher currents leading to higher deposition rates[4]. Sputtering too fast can lead to oxygen-poor, opaque films. The sputter guns are powered by a power supply configured to run in constant-current mode, but the plasma can become unstable at high voltages (>400 V for our system). This presented a problem for depositing Ta2 O5 which has the unusual property that increasing the oxygen flow rate leads to an increase in the sputter voltage for a constant current. The solution to this problem came by adjusting yet another parameter, the total pressure. By increasing both the O2 and Ar flow rates while maintaining their ratio, the overall chamber pressure can be varied. At a 5:1 ratio, Ta2 O5 films were made on the Sloan system at 20 mTorr total pressure with 175 mA current and 375-400 V voltage. Deposition rates varied from 5-10 nm/min. Other deposition parameters (substrate temperature, anode material and size, etc) were varied throughout the research, but their effects were not systematically studied. 2.3 Reactive Doping Transparent oxide films such as Ta2 O5 are typically highly insulating with resistivities on the order of 106 Ωm[7]. To be useful in photovoltaic devices, they must be doped to increase conduction. We attempted to dope the Ta2 O5 films by substituting carbon or fluorine for oxygen atoms in the lattice. This was done by adding gasses containing these elements to the reactive gas mixture. For carbon doping the oxygen gas normally used was replaced with carbon dioxide (CO2 ), which was used in the same ratio as the O2 [6]. In the case of fluorine doping, small amounts of carbon tetra-fluoride (CF4 ) were added to the Ar-O2 mixture through a third independent gas line. This approach is appealing as it allows the doping rate to be varied simply by changing the concentration of the doping gas. Unfortunately we were not able to appreciably increase the conduction of our samples in this way. 8 2.4 Target Doping The more traditional method of doping metal-oxide semiconductors is to add metallic impurities[8]. Previous students accomplished this by adding small amounts of aluminum (Al) to their zinc and indium targets. Many of these films conducted, and I studied the effect of this form of doping on the optical properties of the films. 2.5 Routine Characterization After depositing a film, several tests are usually performed to characterize its properties. Thickness measurements were performed using a Dektak IIA stylus profilometer. The sputtering geometry used results in a nonuniform thickness profile (see Figure 2.1 Right). Structural properties were assessed using a Philips X’Pert X-ray diffractometer. This system measures the intensity of X-rays reflected off of a thin film as a function of the angle of incidence θ at a given wavelength λx . In a crystalline or polycrystalline sample, one expects to see sharp peaks in this spectrum at angles given by the Bragg condition, 2a sin θ = mλx m = 1, 2, 3, ..., (2.1) where a is the spacing of crystal planes perpendicular to the plane of incidence of the X-rays[5]. See, for example, the polycrystalline pattern in Figure 2.2 Left. This film was fabricated by another group using the same sputtering method, but was subjected to a post-deposition annealing process which led to the formation of crystal structure[3]. The absence of sharp peaks in the spectra of the Ta2 O5 films I studied (see Figure 2.2 Right) suggests that they were amorphous, which is in line with observations of non-annealed films made by[3]. Resistivities were measured using a four probe method and two lock-in amplifiers. This method (shown schematically in Figure 2.3) ensures that the lead wire and probe contact resistances are not included in the measured sample resistance. The current I is provided by the reference channel of the first lock-in and is measured as the voltage drop across a 10 Ω resistor. The ballast resistance RB is chosen to be large enough so that I is unaffected by the sample, lead and contact resistances. The second lock-in measures the voltage drop ∆V across a second pair of probes. If the film thickness d is much less than the probe spacing, then Figure 2.2: Left: X-Ray diffraction (XRD) pattern for a polycrystalline Ta2 O5 film[3]. The location of the peaks is related to the lattice parameters of the crystallites. Right: XRD pattern for an amorphous Ta2 O5 film (sample WT-148). Note that both vertical scales are arbitrary. The pattern for the amorphous film is likely on the order of the background noise of the polycrystalline pattern. 9 Figure 2.3: Schematic diagram of four probe resistivity measurement[2]. This setup eliminates errors from lead wire and contact resistances. the resistivity ρ is given by ρ=d π ∆V , log 2 I (2.2) where log is the natural logarithm[9]. This technique provides precise measurements for small resistivities, but for the insulating films I studied it is less illuminating. Another important electrical property of semiconductors is the carrier density. This is typically measured using the Hall effect, whereby a magnetic fields establishes a potential gradient across a two-dimensional sample which is perpendicular to the direction of the field (see Figure 2.4). The size of this potential difference (called the Hall voltage VH ) is a measure of the carrier density N . If a current I is applied to a sample of thickness d and a magnetic field B is directed into it, the carrier density is given by N= BI , edVH (2.3) where e is the carrier charge[10]. The sign of VH determines the charge of the carriers. If both positively and negatively charged carriers are present, the N in Equation 2.3 is the difference in the positive and negative carrier densities, with the sign determining the charge of the dominant carrier. Optical transmission measurements were taken using a UV-Vis-NIR spectrophotometer (see Section 4). Figure 2.4: The Hall effect causes a potential difference VH across a two-dimensional semiconductor related to the carrier density. The applied magnetic field B is out of the page and the current I is indicated by the arrow. 10 3 3.1 Theory Statement of Problem One of the most valuable pieces of information used to characterize our samples is the optical transmission spectrum. This measurement gives the ratio of the intensity of light transmitted through the sample to the incident intensity as a function of wavelength. The purpose of this section is to derive an equation for the transmission spectrum in terms of the optical properties of the material. The sample consists of a thin dielectric film with an unknown permeability µ and permittivity deposited on a thick but finite substrate. Both media are considered to be infinite slabs of uniform thickness with plane-parralell sides with air on either side. Maxwell’s equations, with appropriate boundary conditions, can be applied to find the coefficients of reflection and transmission at each interface (air/film, film/substrate and substrate/air)and then find the net transmission and reflection of the composite system. 3.2 Maxwell’s Equations and Electromagnetic Waves Maxwell’s equations (in CGS units) for linear dielectric media are[11] ~ = ∇ · E ~ = 4πρ, ∇·D (3.1a) ~ = µ∇ · H ~ = 0, ∇·B (3.1b) ~ ~ = − µ ∂H , ∇×E c ∂t (3.1c) and ~ ~ + ∂E . ~ = 4πσ E (3.1d) ∇×H c c ∂t ~ r, t), the electric field E(~ ~ r, t), the magnetic flux density B(~ ~ r, t) They relate the electric displacement D(~ 10 ~ and the magnetic field H(~r, t) where the speed of light c = 3.0 × 10 cm/s. For media with zero conductivity or free charge density (σ = ρ = 0), the wave equations ~− ∇2 E ~ µ ∂ 2 E =0 2 2 c ∂t (3.2a) ~ − ∇2 H ~ µ ∂ 2 H =0 2 c ∂t2 (3.2b) and can be derived. √ √ The phase velocity v = c/ µ = c/n, where n = µ is called the index of refraction of the medium. For √ most materials, the permeability µ in CGS units is practically unity and can be ignored, leaving n = . ~ = E~0 ei(~k·~r−ωt) where the The plane wave solution to Equation 3.2a is the real part of the exponential E wavevector |~k| = 2πn/λ points in the direction of propagation. Substituting this solution into Equation 3.1c and integrating with respect to time gives ~ = H √ ~ = n(ŝ × E), ~ (ŝ × E) (3.3) where ŝ is a unit vector in the direction of propagation[12]. For non-conducting media, , n and k are all purely real. 11 3.3 Reflection and Transmission ~ A ei(k·r−ωt) is incident When an electromagnetic wave with an electric field vector given by the real part of E ~ A (which may be complex if on the boundary between two homogeneous dielectric media, it’s amplitude E k iφ there is a phase offset given by e ) may be resolved into a component EA polarized parallel to the plane of ⊥ incidence (defined by the dashed lines in Figure 3.1) and one perpendicular to it EA . Referencing Figure 3.1 ~ i(kR ·~ r −ωt) ~ we see a reflected wave with an electric field vector given by the real part of ER e and a transmitted i(k~T ·~ r −ωt) ~ wave with an electric field vector given by the real part of ET e , the amplitudes of both of which k k (which may be complex) can also be resolved into components parallel to the plane of incidence (ER , ET ) ⊥ and perpendicular to the plane of incidence (ER , ET⊥ ). One can also find the parallel and perpendicular components of the magnetic field vectors associated with each of these electric field vectors using Equation 3.3. Figure 3.1: Interface between two media with different indices of refraction, showing incident, reflected and transmitted EM waves and angles of incidence and refraction. The z-axis points into the page If a coordinate system is established with the plane of incidence in the xz plane and the interface between the two media in the yz plane, the total components of the electric and magnetic field vectors in the x and y directions can be found by adding the parallel and perpendicular components of the incident, reflected and transmitted waves. In the first medium k k E1x = (EA + ER ) cos φ1 , ⊥ ⊥ E1y = EA + ER (3.4a) and ⊥ ⊥ H1x = n1 (−HA + HR ) cos φ1 , k k H1y = n1 (HA − HR ). (3.4b) In the second medium k E2x = ET cos φ2 , E2y = ET⊥ (3.5a) and H2x = −n2 HT⊥ cos φ2 , k H2y = n2 HT . (3.5b) Our definition of the plane of incidence fixes Ez = Hz = 0. The boundary conditions on Maxwell’s equations require that both components of the the electric and magnetic field vectors must be continuous if there is no surface charge at the interface[12]. One may then solve for the amplitudes of the transmitted and reflected waves in terms of that of the incident wave[11]: 12 k = n1 cos φ2 − n2 cos φ1 ≡ rk , n1 cos φ2 + n2 cos φ1 (3.6a) = 2n1 cos φ1 ≡ tk , n1 cos φ2 + n2 cos φ1 (3.6b) ⊥ ER n1 cos φ1 − n2 cos φ2 = ≡ r⊥ , ⊥ n EA 1 cos φ1 + n2 cos φ2 (3.6c) ET⊥ 2n1 cos φ1 = ≡ t⊥ . ⊥ n1 cos φ1 + n2 cos φ2 EA (3.6d) ER k EA k ET k EA and The above relations define the Fresnel reflection and transmission coefficients r⊥ , rk , t⊥ and tk . Note that the derivation does not depend on whether the refractive indices n1 and n2 are real or complex. For normal incidence (when the incident wave is normal to the interface), φ1 = φ2 = 0 and the distinction between parallel and perpendicular components disappears. The Fresnel coefficients for normal incidence upon an interface between any two media i and j are given by rij = ni − nj ni + nj tij = 2ni . ni + nj (3.7) The Fresnel transmission coefficient gives the ratio of the amplitude of the transmitted wave to that of the incident wave. Similarly, the reflection coefficient gives the ratio of the amplitude of the reflected wave to that of the incident wave. In the experimental setting, the quantities of interest are the ratios of the intensities of the transmitted and reflected waves to that of the incident wave. The intensity of an EM wave is proportional to the square of its amplitude, so the ratios of the intensities of the reflected and transmitted waves to that of the incident wave (called simply the reflection and transmission) are are given by Rij = |rij |2 = (ni − nj )2 (ni + nj )2 Tij = nj nj 4n2i . |tij |2 = ni ni (ni + nj )2 (3.8) The factor of nj /ni in the transmission reflects the fact that the energy carried by an EM wave is proportional to its phase velocity as given by the index of refraction. To deal with multiple interfaces, the Fresnel coefficients should be combined in the appropriate way and then the result squared as above to obtain the reflection or transmission. 3.4 Propagation Since light is an electromagnetic wave, it’s propagation is profoundly affected by the electrical properties of the medium. In a metal or semiconductor where there may be some electrical conductivity (σ 6= 0), the wave equation for the electric field vector becomes[12] ~ = ∇2 E ~ ~ 4πµσ ∂ E µ ∂ 2 E + 2 . 2 2 c ∂t c ∂t (3.9) ~ The ∂∂tE term implies that the wave is damped and will be attenuated as it propagates. If time dependence of the form e−iωt is assumed, one finds ~ + k̃ 2 E ~ = 0. ∇2 E (3.10) This expression is equivalent to Equation 3.2a, but the wavevector |k̃| = 2π λ ñ is now complex. The complex index of refraction ñ = n − iκ where κ is called the extinction coefficient and n, κ are real. For our 13 purposes it will be useful to write κ = −αλ/4π where α ≥ 0 is called the absorption coefficient. One may α α then write |k̃| = 2π λ n + i 2 = |k| + i 2 . The plane wave solution to Equation 3.10 for propagation in the x direction is ~ E(x, t) = Re[E~0 ei(kx−ωt) e−αx/2 .] 3.5 (3.11) Multiple Reflections Consider the configuration in Figure 3.2 where a single film (thickness d) is “sandwiched” between two media which are assumed to be non-absorbing. A normally incident wave propagating in the x direction with an ~ A ei(kx−ωt) will be split into transmitted and reflected waves with amplitudes given electric vector given by E ~ T = t12 E ~ A and E ~ R = r12 E ~ A . If the film is absorbing, its index of refraction by the Fresnel coefficients as E is complex with n2 = n − iκ. If the first interface is taken to be at x = 0 and the second at x = d, then ~ A e−i(ωt−φ/2) e−αd/2 where using Equation 3.11 the amplitude at the second interface is found to be t12 E φ/2 = 2π λ nd is the phase acquired traversing the film and α is the same as above. The phase and attenuation 1 factors can be combined by defining the complex phase β ≡ 2π λ n2 d = 2 (φ + iαd). The amplitude at the ~ A e−iωt eiβ . For the rest of this discussion, we will ignore the time second interface can then be re-written as E part since the quantity of interest is simply the ratio of the transmitted intensity to the incident intensity. Figure 3.2: Schematic diagram of a thin film (thickness d) between two infinite media, showing contribution to reflected and transmitted wavefronts from rays reflected within the film multiple times. To calculate the intensity of light transmitted through the film, the contribution from rays reflected multiple times must be calculated[11]. The ratio of the amplitude of a wave passing directly through the film to the amplitude of the incident wave is equal to the product of the Fresnel transmission coefficients at the two interfaces and the attenuation term: t12 t23 eiβ . For a wave transmitted through the first interface, reflected off the second and then again off the first and then transmitted through the second interface, this ratio would further be multiplied by the Fresnel reflection coefficients of the two interfaces yielding t12 t23 r23 r21 e3iβ , where the r21 subscript denotes reflection into medium 2 and the factor of 3 in the exponential denotes the fact that the light passes through the film three times. Light can bounce back and forth in the film any number of times, so the ratio t123 of the total transmitted amplitude to that of the incident wave is given by 2 2 5iβ t123 = t12 t23 eiβ + t12 t23 r23 r21 e3iβ + t12 t23 r23 r21 e + ... = t12 t23 eiβ ∞ X `=0 This sum is a simple geometric series and can be written as 14 [r21 r23 e2iβ ]` . (3.12) t123 = t12 t23 eiβ . 1 − r21 r23 e2iβ (3.13) A similar calculation gives the ratio r321 of the reflected amplitude of a wave incident from the substrate side to that of the incident wave: r321 = r32 + r21 e2iβ . 1 + r32 r21 e2iβ (3.14) Equation 3.13 gives the ratio of the amplitude of the transmitted wave to that of the incident wave. The transmission (the ratio of the transmitted intensity to the incident intensity) T123 can be calculated from ∗ Equation 3.8. Accordingly, T123 = (n3 /n1 )t∗123 t123 . Noting that |eiβ |2 = e−iβ eiβ = e−i(φ−iαd)/2 ei(φ+iαd)/2 = e−αd , one finds T123 = n3 |t12 t23 |2 e−αd . n1 1 − 2r21 r23 e−αd cos φ + |r21 r23 |2 e−2αd (3.15) ∗ The ratio of the intensity of reflected light to incident light is given by R321 = r321 r321 , or R321 = |r32 |2 + 2r32 r21 e−αd cos φ + |r21 |2 e−2αd . 1 + 2r32 r21 e−αd cos φ + |2r32 r21 |2 e−2αd (3.16) The transmission in Equation 3.15 does not accurately portray the physical situation under which thin film spectra are obtained in the laboratory. Figure 3.3 shows the configuration correctly, with the thin film on top of a thick (but finite) substrate, usually a glass microscope slide. For a thick dielectric slab, Equation 3.15 breaks down because the multiple reflections in the slab do not interfere as they do in the film. Figure 3.3: The physical situation for thin film measurements in the lab. The film (index n2 ) rests on top of a thick but finite substrate (index n3 ) bounded on either side by air (index n1 ). The above analysis assumes that the light is monochromatic and coherent and can thus be expressed as a simple plane wave. For optical transmission measurements, the wavelength of light is on the order of the film thickness, which itself is small compared to the other dimensions of the problem. Over such short length scales the light is approximately coherent. The substrate thickness is several hundreds of thousands of times greater than that of the film. Over such long length scales, the finite bandwidth of the spectrometer comes into play and the light can not be considered monochromatic, a prerequisite for interference[13]. Interference effects can be averaged out (just as they are by the spectrometer) by integrating |t123 |2 (Equation 3.12) over the effective bandwidth ∆λ [14]. One finds 15 Z ∞ X |t12 t23 eiβ [r21 r23 e2iβ ]` |2 dλ = ∆λ Z ∞ X ∞ X dλ . ∆λ ∆λ ∆λ `=0 `=0 `0 =0 (3.17) In a thick sample, the real part of β must be much greater than the imaginary part for the reflected light to be detectable at all. This means that the factors of e2iβ` in Equation 3.17 oscillate rapidly with λ and vanish when averaged over ∆λ. The only terms in the sum that remain are then the terms where ` = `0 and the real parts of β cancel. One then has T = T = |t12 t23 |2 e−αd ∞ X |t12 t23 |2 e−αd [|r21 r23 |2 e−2αd ]` = `=0 0 ∗ ∗ ` 2i(β`−β (r21 r23 )` (r21 r23 ) e |t12 t23 |2 e−αd . 1 − |r21 r23 |2 e−2αd ∗ 0 `) (3.18) Equation 3.18 gives the ratio of the transmitted intensity to the incident intensity for a thick slab between two infinite media. To account for the presence of the film the transmission and reflection coefficients for the first interface are replaced by those for the film from Equations 3.13 and 3.14 (this is why the reflected intensity was calculated from the substrate side). The result is given by T = (1 − R13 )T123 , 1 − R13 R321 (3.19) where R13 = |r13 |2 is the reflection from the glass/air interface. Substituting Equations 3.15 and 3.16 and simplifying yields T = n3 (1 − |r13 |2 )|t12 t23 |2 e−αd . 1 − |r13 r32 |2 + 2r21 r32 (1 − |r13 |2 )e−αd cos φ + |r21 |2 (|r32 |2 − |r13 |2 )e−2αd (3.20) One can now finally substitute the Fresnel coefficients from Equation 3.7 with n1 = 1 for air, n2 = n and the substrate index n3 = s. Also plugging in the above expression for φ, Equation 3.21 is obtained, which is the same expression quoted by Swanepoel[15]. T = (n + 1)3 (n + s2 ) − 2(n2 − 1)(n2 16n2 se−αd . 3 2 −2αd − s2 )e−αd cos ( 4π λ nd) + (n − 1) (n − s )e (3.21) The use of a purely real index of refraction n2 = n may seem inappropriate as Equation 3.20 was derived with the assumption that n2 was complex. A more rigorous expression would use n2 = n − iκ. Over most of the spectrum, however, κ is nonzero but is several orders of magnitude smaller than the real part n. The effect of the imaginary part enters through the exponential terms. In regions of strong absorption, however, this approximation breaks down and different expressions must be used. The phenomenon of thin film interference is buried in the φ dependence of Equation 3.20. The transmission oscillates between extreme values when φ takes on values which are integer multiples of π. To determine the nature of these extrema, simply take the second derivative of Equation 3.21 and determine its sign when φ = 2mπ, or when 2nd = mλ m = 1/2, 1, 3/2, 2, 5/2, ... (3.22) This is the classic expression for thin film interference. The nature of the extremes turns out to depend on the sign of (n − s). If n < s, the interference order m = 1, 2, 3, ... for transmission maxima and m = 1/2, 3/2, 5/2, ... for minima. If n > s, the opposite is true. Note that this equation gives information on the product nd based on the measured wavelengths of the interference fringes and cannot give either n or d independently. It is only through use of the full transmission equation that n may be extracted without knowledge of d. 16 4 4.1 Measurement and Analysis UV-Vis-NIR Spectrophotometer I made transmission measurements using the Varian Cary 5e UV-Vis-Nir spectrophotometer located in the Chemistry department. It is configured to split the measurement beam in a 3-phase “chopper” cycle. The chopper itself is a disk (see Figure 4.1) which spins at 30 Hz and is divided into transparent, mirror and matte sections. Figure 4.1: The “chopper” disk alternates the path of the measurement beam as it rotates. It has three sections: transparent (1; white), mirror (2; grey) and matte (3; black). The chopper is placed between the light source and the sample chamber (see Figure 4.2) and as it spins, the beam from the source is incident on each of the three section in turn. Each cycle of the chopper disk then has three “phases”. In the transparent phase (the phase during which light is passing through the transparent section of the disk) the light beam from the source follows a path taking it through the sample and to the detector on the other side of the sample chamber. During the mirror phase the beam is reflected off the mirror portion of the disk onto a path taking it through the sample chamber but not through the sample. The intensity of light measured by the detector during this phase is equal to the intensity of light incident on the surface of the sample. The spectrometer software divides the transparent phase intensity by the mirror phase (incident) intensity to obtain the transmission. Figure 4.2: Simplified schematic of spectrophotometer. The chopper sends the measurement beam alternately through the sample and not through the sample. During the third phase the beam from the source is absorbed by the matte black portion of the disk and no light from the source is incident on the detector. This is appropriately called the “no-measurement” phase. In his phase a stepper motor rotates a diffraction grating to change the wavelength of the beam. This no-measurement phase wavelength stepping eliminates errors associated with the motion of the grating and increases resolution. 17 Before every measurement, a baseline scan with no sample in the beam line is performed to calibrate the system. This eliminates errors associated with differences between the two beam paths. A scan of an uncoated microscope slide is performed next. Once these measurements are made the samples of interest (the coated slides) can be scanned. All measurements are made from 200 - 2000 nm. The spectral bandwidth of the beam (full width at half height) was 2 nm and the scanning rate was 600 nm/min. An example of the resulting raw data for a ZnO sample are presented in Figure 4.3. The uncertainty in the transmission measurements is estimated by looking at the spectrum of a blank substrate. Over small wavelength ranges the transmission should be approximately constant. By observing the statistical noise in the data, I found the error to be less than 0.1%. Note that by using the classic interference condition, Equation 3.22, for the 1 mm thick glass substrate, the wavelength spacing between adjacent maxima can be shown to be less than 0.1 nm, which can not be resolved at this bandwidth. This justifies our use of Equation 5.2. Figure 4.3: Example of raw data from spectrophotometer showing blank substrate (flat upper curve) and ZnO coated sample (oscillating lower curve). Sample Zn-005. 4.2 Refractive Index of Substrate Before the optical properties of the thin film can be determined, those of the thick substrate on which it rests must be found. The absorption coefficient is assumed to be zero, though absorbing substrates can in principle be treated as in [16]. The thickness of the substrate does not enter into Equation 3.21, so all that remains is to calculate the index of refraction s. One begins by measuring the transmission of a blank substrate (see Figures 4.3 and 4.4 Left). The transmission for this simple case can be found using Equation 3.18 and is given by 2s . (4.1) s2 + 1 Inverting this expression allows us to determine s as a function of wavelength from the measured transmission Ts . Note from Figure 4.3 that there is a strong absorption feature in the short wavelength region of the measured spectrum. This is caused by a sudden increase in the absorption coefficient of the substrate, not the real part of the refractive index. To account for this effect, s is taken to be constant for wavelengths below 500 nm, although s probably increases by about a factor of 2 in this region. The increase in the imaginary part, by comparison, is likely a factor of 100. The result is shown in Figure 4.4. The substrate’s refractive index is seen to vary by less than 5% across the measurement region. Though s could be assumed to be a constant across the whole spectrum, allowing it to vary gives more accurate results. Ts = 18 Figure 4.4: Left: Example transmission spectrum of blank substrate. Right: Calculated refractive index of blank substrate. From Equation 4.1 and the rules of error analysis[17] one can see that the relative uncertainty in s is the same as the relative uncertainty in Ts , or about 0.1%. 4.3 Interference Envelope Equation 3.21 may be more succinctly written as T = Ax , B − Cx cos φ + Dx2 (4.2) where A ≡ 16n2 s, (4.3a) B ≡ (n + 1)3 (n + s2 ), (4.3b) C ≡ 2(n2 − 1)(n2 − s2 ), (4.3c) D ≡ (n − 1)3 (n − s2 ), (4.3d) x ≡ e−αd , (4.3e) 4π n. λ (4.3f) and φ≡ Using this form, the interference envelope consisting of the functions that pass through the extreme points in the spectrum (cos φ = ±1) are defined by TM ≡ Ax B − Cx + Dx2 and (4.4a) Ax . (4.4b) B + Cx + Dx2 The upper bounding function TM passes through the maxima of the spectrum and Tm through the minima, as in Figure 4.5, where the envelope for a simulated transmission curve is shown. In practice these functions are built by creating interpolating functions between the extrema in the transmission data. Figure 4.6 Left shows a comparison between first, second and third order interpolations Tm ≡ 19 Figure 4.5: Simulated transmission curve (red) for a 2000 nm thick film, showing interference envelope functions TM (Equation 4.4a) and Tm (Equation 4.4b) in blue. of an example data set. The higher order interpolations tend to give spurious results especially if the film is thin and few interference fringes are visible. A first order (linear) interpolation, while crude, is the most likely to capture the smoothly varying interference envelope (see Figure 4.6 Right). Figure 4.6: Left: Comparison of interference envelopes built from first (red), second (blue) and third (green) order interpolations. Dashed line is a measured transmission spectrum for sample WT-152. Right: Example of a measured spectrum (green) with linear interference envelope (blue). Sample Zn-005. 4.4 Refractive Index of Film Following Swanepoel[15], note that 1 2C 1 − = . Tm TM A The quantity on the right hand side is independent of the thickness of the film. Solving for n yields (4.5) n = [N + (N 2 − s2 )1/2 ]1/2 , (4.6) TM − Tm s2 + 1 + . T M Tm 2 (4.7) where N = 2s 20 Using the previously calculated values for s(λ) and the linear interference envelope described in the previous section, the index of refraction of the film can be extracted without knowledge of the film thickness. As n depends on s1/2 , the relative uncertainty in s (about 0.1%) leads to a relative uncertainty in n of less than 0.05%[17]. This is small compared to the error arising from our use of interpolated interference envelopes. According to Swanepoel, a 1% error in the envelopes leads to a 1-3% error in n. I conservatively estimate the error in n to be about 3%. For many of the films studied, the index of refraction was found to fit well to a second order Cauchy dispersion relation[18] of the form n2 n1 + 4. (4.8) λ2 λ This empirical fit was derived by Cauchy based on a theory of light propagation later proven to be false. Nevertheless, it is often used by optical spectroscopists for its simplicity[18]. An example of an extracted n(λ) and it’s accompanying Cauchy fit are shown in Figure 4.7. n(λ) ≈ n0 + Figure 4.7: Example of extracted refractive (yellow) index with Cauchy fit (red, Equation 4.8). The blue dashed lines show the 3% relative uncertainty in n. Sample Zn-005. 4.5 Film Thickness Now that the index of refraction is known, the classic thin film interference equation (Equation 3.22) may be used to determine the thickness of the film. This is done using a simple graphical method. First, Equation 3.22 is rewritten as 2nd = (m0 − `/2)λ, (4.9) where ` = 0, 1, 2, ... and m0 is the unknown order of the first interference fringe in the measured spectrum (ie the extreme that occurs at the longest wavelength). Then a plot is made of the straight line `/2 = 2d(n/λ) − m0 . (4.10) The value of m0 is chosen to be the nearest integer or half-integer to yield a good fit to the data, bearing in mind that if n > s then an integer-valued m0 implies that the the first interference extremum is a maximum, with a half-integer implying a minimum. The slope of the line is 2d where d is the film thickness. An example of this method is shown in Figure 4.8. Sometimes the points corresponding to higher order extrema will deviate significantly from the linear relation in Equation 4.10. This is due to absorption effects as the 21 imaginary part of the refractive index becomes non-negligable near the band-edge. These points should be ignored when determining the thickness. Uncertainty in the slope of the line (and hence the thickness) is determined through standard weighted lest-squares methods[17]. If the extracted n(λ) is fit well by Equation 4.8 than using this fit in Equation 4.10 can reduce dispersion and lead to more accurate thickness measurements. Figure 4.8: Graphical method for determining film thickness, interference order `/2 is plotted against n/λ. The slope of the line is twice the film thickness. Sample Zn-005, with m0 = 3/2 and d = 617 nm. 4.6 Absorption Coefficient If one now adds the reciprocals of Equations 4.4a and 4.4b, one can solve for x = e−αd [15] and find x= F − [F 2 − (n2 − 1)3 (n2 − s4 )]1/2 , (n − 1)3 (n − s2 ) (4.11) where F = 4n2 s T M + Tm . TM Tm (4.12) The uncertainty in α is found through error propagation and for the sample in Figure 4.9 Right varies from 3-4×10−5 nm−1 across the measurement region. This uncertainty is most sensitive too the error in Tm , which is largest in the long-wavelength regime where there are insufficient data to construct accurate interpolations. 4.7 Band Edge Equation 4.11 is only valid when our assumption n κ holds. At wavelengths corresponding to photon energies near the bandgap energy of the film, this assumption fails as α increases by orders of magnitude. This strong absorption washes out interference effects and, since x is very small, Equation 4.2 can be rewritten as T0 ≈ Ax/B. The values of A and B however are not the ones given in Equation 4.3. The exponential nature of this relation, however, allows a simple fit to the the transmission data near the band edge. Such a fit is shown in Figure 4.9 Right. The value of the absorption coefficient near the band edge provides key information about the film’s band structure. The energy gap Eg can be found from the well known relation[19] 22 Figure 4.9: Left: Example absorption coefficients calculated from the refractive index n (yellow) and the Cauchy fit to n (red). Blue dashed lines show uncertainty. Right: Example absorption coefficient near band edge. The substrate itself begins absorbing below 250 nm, so data in this region should be ignored. Note the difference in scales on the two plots. Sample Zn-005. α ∝ (E − Eg )1/2 , (4.13) where E = hν = hc/λ is the photon energy (Planck’s constant h = 4.136×10−15 eV s). This expression assumes a direct bandgap and parabolic valence and conduction bands. Ta2 O5 , ZnO and In2 O3 all have direct bandgaps[3][18][20]. A linear fit to a plot of α2 vs E has an E-intercept at Eg , as in Figure 4.10 Left. The absorption coefficient can also be fit to the Urbach relation α ∝ eE/E0 , where E0 is called the Urbach slope and characterizes the width of exponential band-tailing into the gap and is frequently used as a measure of disorder and impurities in thin films[19]. The Urbach slope is found from a linear fit to a plot of log α vs photon energy, as in Figure 4.10 Right. It should be noted from Figure 4.3 that the substrate itself may have strong absorption features. If the bandgap energy of the film is close to or greater than that of the substrate, the film’s band edge may become convoluted with or hidden by substrate absorption. In cases such as these, as in Ta2 O5 with Eg = 4.2 eV right at the bandgap of the glass substrates used, the analysis in this section should be applied with caution. Figure 4.10: Left: Plot of α2 vs hν with fit to linear region. The intercept is the bandgap energy. Right: Plot of ln α vs hν with fit to linear region. The slope is the inverse of the Urbach disorder parameter. Sample Zn-005, with Eg = 3.28 eV and E0 = 75 meV. 23 Once α has been determined near the band edge, it can be joined to the interference-zone absorption to obtain α(λ) across the whole spectrum. Mathematically, α(λ) = αgap u(λc − λ) + αint u(λ − λc ), (4.14) where u(λ) is the unit step function and λc is the wavelength where the band edge drop meets the interferencezone transmission. 4.8 Absorption in Doped Films The doping processes discussed in Section 2 can have a dramatic effect on the transmission spectrum, as in Figure 4.11 Left, where I have graphed the transmission spectrum of a doped ZnO film in blue and that of an uncoated substrate in red. The increased absorption in the infrared region of the spectrum is due to excitations of donor-level electrons to the conduction band at these energies. A nearly-free-electron model of doped semiconductors gives the following relation between the absorption coefficient α and the room-temperatue carrier density N associated with the doping[21]: N 2 e2 λ , (4.15) 4π 2 m∗ 0 nc3 τ where e is the electron charge, m∗ is the effective mass of conduction band carriers, 0 is the vacuum permittivity, n is the index of refraction and c is the speed of light. The scattering time τ is related to the mobility µ = eτ /m∗ . By fitting the infrared absorption coefficient with a λ2 dependence, one can extract the parameter N/τ (see Figure 4.11 Right). To fully determine the carrier density and mobility, a 4-probe resistivity measurement can be made and the relation 1/ρ = σ = N eµ employed. The complete absorption coefficient is now built from its value in the three regions of the spectrum, the band edge, the interference zone, and the infrared “carrier zone”: α= α(λ) = αgap u(λc − λ) + αint u(λ − λc )u(λd − λ) + αcar u(λ − λd ), (4.16) where αcar is the carrier zone absorption and λd is the critical wavelength where doping effects become noticeable. Figure 4.11: Left: Example transmission spectra for a doped ZnO film (blue) showing infrared absorption and uncoated glass microscope slide (red). Right: By fitting the infrared absorption coefficient (blue) with a λ2 dependence (red), one can determine the ratio of the carrier density N to the scattering time τ . Sample Zn-044 with N/τ = 7.1 × 1035 cm−3 s−1 as determined through Hall effect measurements. 24 5 5.1 Results and Discussion Reconstructed Curves Once s(λ), n(λ), d, and α(λ) have been found according the above analysis, the full transmission curve can be reconstructed using Equation 3.21. This is a useful check to ensure the extracted optical properties are accurate. Figure 5.1 is a typical example. It also shows one of the limitations of this method, which is that in the long wavelength region there are insufficient data (transmission extrema) to construct accurate transmission envelopes. This results in uncertainty in all properties derived from these envelopes. It is sometimes possible to improve the reconstruction by extrapolating n from the visible region using Equation 4.8, or some other dispersion rule. For very thin (d < 200 nm) films with only 2-3 measurable fringes, even this approach is difficult. Figure 5.1: Example of reconstructed transmission curve. The blue curve is raw data and the red curve is the reconstruction. Sample Zn-005. Compare Figure 4.3. 5.2 Refractive Indices Refractive indices for materials are often quoted as a single number which is independent of wavelength. Taking note of Figure 4.7, this may be seen to be a reasonable approximation. The values of n calculated for transparent films usually vary by only 5-10% across the entire measurement region. For calculating the thickness from Equation 3.22, approximating n as a constant results in an error of 5-10%, which for our deposition process is less than the variation in the thickness across the substrate. The same is true for many other properties of the material that depend on n. For reconstructing the transmission curve, however, this approximation falters. The term in the transmission equation that determines the amplitude of the interference fringes (Equation 4.3c) depends on n4 , meaning that even a 5% error in n will yield a 20% error in the fringe amplitude. This effect can be seen in Figure 5.2, where the location of the maxima is accurate but the amplitude is distorted. Allowing n to vary with wavelength allows us to capture the complete transmission curve more accurately. Transmission near the bandgap (where this effect is most pronounced) is of special importance for photovoltaic applications, making the constant n approximation unjustifiable. Physically, the real part n of the refractive index must vary with wavelength to satisfy causality in the universe. The imaginary part κ can be seen to vary with wavelength from the absorption features at the band edge an in the infrared. Because n and κ are the real and imaginary parts of a single complex parameter, they must satisfy the Kramers-Kronig relationships[21], 25 1 n(ω) = 1 + π and κ(ω) = − 1 π Z Z ∞ −∞ ∞ −∞ κ(ω 0 ) dω 0 ω0 − ω n(ω 0 ) − 1 0 dω , ω0 − ω (5.1a) (5.1b) where ω = 2πν = 2πc/λ is the radial frequency. So any variation in κ at any wavelength implies that n will also have to vary with λ. These relationships can be used, in principle, to determine the imaginary part of the refractive index from the real part and vice- versa, but this would require information over a broader spectral range[21]. Figure 5.2: Left: If n is taken as a constant, the reconstructed transmission curve (blue curve) fails to capture the variation in the amplitude of the interference fringes present in the data (red curve). Right: If n is allowed to vary, better agreement is achieved, particularly in the fringe amplitude. Sample WT-148. Table 5.1 summarizes the typical values of n extracted using my method. In total I analyzed the spectra for 4 Ta2 O5 films, 15 ZnO films (about half of which were doped) and 2 In2 O3 films (both of which were doped). For Ta2 O5 and ZnO, a range of values is given to show the variation of n with wavelength. Bulk Ta2 O5 is usually taken to have n ≈ 2.2, with thin films of good transparency showing n ≈ 2.1[3]. The films I studied showed slightly smaller n values. The observed indices for the ZnO films was in good agreement with accepted values[18]. Too few In2 O3 films were studied to get a handle on the statistics, but there was rough agreement with accepted values[22]. Material Ta2 O5 ZnO In2 O3 Measured n 1.96(5)-2.04(4) 1.87(6)-1.95(7) 2.0(5) Literature n 2.09-2.13[3] 1.93[18] 1.95[22] Table 5.1: Real part of the index of refraction n as extracted with Equation 4.6 and literature values. Uncertainties in parentheses. 5.3 Optical Versus Mechanical Thickness Measurements To assess the accuracy of this method, the thicknesses determined through the optical measurement, dopt , were compared to those made using a Dektak stylus profilometer, dmech . The results for the Ta2 O5 films 26 studied are presented in Table 5.2. Given the surface irregularity inherent to our sputtering system (Figure 2.1 Right), the numbers show good agreement, supporting the validity of this method. Film WT-148 WT-150 WT-151 WT-152 dopt (nm) 588(2) 351(3) 353(3) 306(3) dmech (nm) 535 300 400 325 % Difference 9% 14% 12% 6% Table 5.2: Comparison of optical (dopt ) and mechanical (dmech ) thickness measurements for Ta2 O5 films. Uncertainties in optical measurements in parentheses. 5.4 Bandgap Energies For the ZnO and In2 O3 studied, the bandgap energies were found using Equation 4.13. Average values are reported in Table 5.3. The bandgap for ZnO was observed to be slightly smaller than the literature value[18]. Though only two In2 O3 films were analyzed, the bandgap energies were in excellent agreement with accepted values[20]. Average values of the Urbach slope parameter E0 are also included in Table 5.3. This energy characterizes the width of exponential band tailing (“fuzziness”) associated with sample impurities[19]. Note the wider band tail observed in In2 O3 . Material ZnO In2 O3 Measured Eg (eV) 3.27(3) 3.63(2) Literature Eg (eV) 3.35[18] 3.6[20] E0 (meV) 69(8) 570(70) Table 5.3: Measured and literature values of energy bandgap Eg , and Urbach slope E0 for ZnO and In2 O3 films. Uncertainties in parentheses. This analysis was also applied to the Ta2 O5 films. The measured bandgaps agreed well with the accepted 4.2 eV value[3], but the proximity of the film’s band edge to that of the substrates casts doubt on these measurements (see Figure 5.3). Figure 5.3: Close-up of Ta2 O5 transmission curve near band edge (lower curve). The proximity of the Ta2 O5 bandgap to that of the substrate (upper curve) complicates analysis in this region. Sample WT-152. 27 5.5 Carrier Densities A number of complications arise when one tries to calculate carrier densities of doped films using this method and Equation 4.15. My method of extracting the real part of the index of refraction relies on measuring the amplitude of interference fringes, which are wiped out in the infrared by carrier absorption (see Figure 4.11 Left). The index of refraction must then be extrapolated from the interference zone. As n is nearly constant away from the band edge, this is at most a 10% error. I tried several methods of obtaining numerical values for the absorption coefficient in the carrier zone, including solving Equation 4.2 for x. In the end, a simple exponential dependence was used. A more general expression than Equation 3.21 may be needed to obtain accurate information about α. That being said, the values of α used fit well to a λ2 dependence (Figure 4.11 Right) and the reconstructed transmission curves agreed well with the data (Figure 5.4). This qualitative agreement suggests that quantitative results are not unattainable. Figure 5.4: Example of reconstructed transmission curve for a doped film. The blue curve is raw data and the red curve is the reconstruction. Sample Zn-044, with N/τ = 7.1 × 1035 cm−3 s−1 as determined through optical measurements. Compare Figure 4.11. After fitting the α values to Equation 4.15, I extracted the values of N/τ in Table 5.4. I compared these to those obtained earlier by John Scofield from Hall effect measurements. For the ZnO films (m∗ = 0.25me [18]) I studied, the results were somewhat inconsistent. Agreement between the two measurements ranged from being within a factor of 2 to a factor of 50, though almost all were within an order of magnitude. A further complication arises when impurities in the sample are considered. The Hall measurements determine the difference between the positively and negatively charged carrier densities (see Equation 2.3. If both types of carriers are present in significant numbers, as is likely the case with these films, the measurement Film Zn-025 Zn-037 Zn-041 Zn-043 Zn-044 (N/τ )opt (1035 cm−3 s−1 ) 2.5 2.9 3.6 0.5 2.1 (N/τ )Hall (1035 cm−3 s−1 ) 160 18 4.8 3.5 7.1 Table 5.4: Values of N/τ as extracted from optical data ((N/τ )opt ) and from Hall effect measurements ((N/τ )Hall ). 28 will underestimate N and overestimate τ by as much as a factor of 2. This does not explain the discrepancy between the Hall and optical measurements, as the optically obtained values were always smaller. The impurity density contributes to the absorption coefficient additively, but with a different effective mass and scattering time[21]. The effect of impurities on absorption thus depends on the mobility of the carriers associated with them. 6 Conclusion This project has shown the power of Maxwell’s equations in understanding the optical and electrical properties of thin films. Performing a thorough analysis of the transmission spectrum using Equation 3.21 allows us to deduce many of the parameters important to photovoltaic applications. Armed with these techniques, we can more deterministically tailor the performance of a complete cell (Figure 1.1). By systematically determining the effects of the deposition parameters on the optical properties, one could use the same equations to simulate the transmission of a film before it is made. Trade-offs such as conductivity versus infrared absorption could be computationally optimized without the need to deposit dozens of films by trial and error. Linking deposition parameters and optical properties (now readily available using this analysis) is then an important next step for solar cell research at Oberlin. It is likely that improved accuracy could be achieved by incorporating the full complex refractive index in the analysis, especially in the infrared where the absorption is more gradual than near the band edge. Combined with Hall effect measurements, this would enable complete and accurate characterization of the n- and p-type carrier densities. These are critical factors for photovoltaic applications (see Figure 1.2), but more work is needed to understand the role of impurities and their effect on the transmission. Pinning down the discrepancy between the Hall effect and optical measurements is the logical starting point. We have developed this analysis as a tool for quickly determining the properties of the samples we generate to guide our research in an informed, orderly fashion. Acknowledgements I’d like to thank my advisor, John Scofield, for his years of support and guidance, and the rest of the Physics and Astronomy community. References [1] J H Scofield et el. Sputtered molybdenum bilayer back contact for copper indium diselenide-based polycrystalline thin film solar cells. Thin Solid Films, 260:26, 1995. [2] J Gaumer. Fabrication and characterization of zno thin films for cis solar cell window layers. Oberlin College Honors Thesis, 2002. [3] J M Ngaruiya et al. Preparation and characterization of tantalum oxide films produced by reactive dc magnetron sputtering. Physics Status Solidi, 198:99, 2003. [4] T J Gillespie et al. Reactive magnetron sputtering of transparent and conductive zinc oxide films deposited at high rates onto cis/cigs photovoltaic devices. Proceedings of the 26th IEEE Photovoltaic Specialists Conference, 1997. 29 [5] C Kittel. Introduction to solid state physics. Wiley, 2005. [6] K Chu and J P Chang. Material and electrical characterization of carbon-doped tantalum oxide films for embedded dynamic random access memory applications. Journal of Applied Physics, 91:308, 2002. [7] E Krikorian and R J Sneed. Deposition of tantalum, tantalum oxide, and tantalum nitride with controlled electrical characteristics. Journal of Applied Physics, 37:3674, 1966. [8] D A Sager et al. Diffusion of neodymium into sputtered films of tantalum pentoxide. Journal of the American Ceramics Society, 85:2581, 2002. [9] L Maissel et al. Handbook of thin film technology. McGraw-Hill, 1970. [10] B G Streetmnan. Solid state electronic devices. Prentice Hall, 1990. [11] O Heavens. Optical properties of thin solid films. Dover, 1965. [12] M Born and E Wolf. Principles of Optics. Pergamon, 1959. [13] F Manoocheri S Nevas and E Ikonen. Determination of thin-film parameters from high accuracy measurements of spectral regular transmittance. Metrologia, 40:S200, 2003. [14] F Manoocheri A Haapalinna and E Ikonen. High-accuracy measurements of specular spectral reflectance and transmittance. Anal Chim Acta, 380:317, 1999. [15] R Swanepoel. Determination of the thickness and optical constants of amorphous silicon. J Phys E, 16:1214, 1983. [16] J Cisneros. Optical characterization of dielectric and semiconductor thin films by use of transmission data. Applied Optics, 37:5262, 1998. [17] J Taylor. An introduction to error analysis. University Science Books, 1997. [18] U Ozgur et al. A comprehensive review of zno materials and devices. Journal of Applied Physics, 98:041301–1, 2005. [19] A Aqili and A Maqsood. Determination of thickness, refractive index, and thickness irregularity for semiconductor thin films from transmission spectra. Applied Optics, 41:219, 2002. [20] H Jia et al. Efficient field emission from single crystalline indium oxide pyramids. Applied Physics Letters, 82:4146, 2003. [21] M Fox. Optical Properties of Solids. Oxford University Press, 2001. [22] S Laux et al. Room temperature deposition of indium tin oxide thin films with plasma ion-assisted evaporation. Thin Solid Films, 335:1, 1998. 30 Appendix A.nb Clear@slidedata, slide100, t, sD; slidedata = Import@"filename.csv"D; êê Imports raw data on blank slide slide100 = slidedata.881, 0<, 80, .01<<; êê Converts data from 0 - 100 % to 0 - 1 t = Interpolation@Join@Drop@slide100, -300D, Transpose@ Join@8Array@Identity, 300, 200D<, 8Table@Last@slide100@@-300DDD, 8i, 300<D<DDDD; êê Makes Ts a constant below 500 nm s@x_D = 1 ê t@xD + H1 ê t@xD^ 2 - 1L ^ H1 ê 2L; êê Determines s@lD Clear@data, data100, intdataD; data = Import@"êVolumesêMARKASAURUSêOpticsêSpectraêzn037.csv"D; êê Imports raw data on film + slide data100 = data.881, 0<, 80, .01<<; êê Converts data from 0 - 100 % to 0 - 1 intdata = Interpolation@data100, InterpolationOrder Ø 1D; Clear@max, min, TM, TmD; max = 881135, .830656<, 8660, .8829886<, 8473, .8679934<<; min = 88868, .788655<, 8543, .8069522<<; êê These are example extreme data found using a peak - fitting program TM = Interpolation@max, InterpolationOrder Ø 1D; Tm = Interpolation@min, InterpolationOrder Ø 1D; êê Builds the linear interference envelope 1 Appendix A.nb 2 Clear@n, cauchyD; i j1 2 s@xD H-Tm@xD + TM@xDL j n@x_D = . j H1 + s@xD2 L + ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ + j j ÅÅÅÅ 2 Tm@xD TM@xD k z 1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% s@xD H-Tm@xD +%%%%%%%%%%%%%%%% TM@xDL 2% y z $%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -s@xD2 + J ÅÅÅÅ H1 + s@xD2 L + ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ%%%%%%%%% ÅÅÅ N z z z; 2 Tm@xD TM@xD { êê Finds n@lD Plot@8n@xD<, 8x, 200, 2000<, PlotStyle Ø [email protected]<D êê Plots n@lD to determine limits for Cauchy fit cauchy@x_D = Fit@Transpose@8Array@Identity, points, startD, Array@n, points, startD<D, 81, x ^ -2, x ^ -4<, xD êê Finds Cauchy fit Plot@8n@xD, cauchy@xD<, 8x, 200, 2000<, PlotStyle Ø [email protected], [email protected]<D êê Plots both to determine fit success Clear@all, linelist, ell, d, b, hD; all = Reverse@Sort@Join@max, minDDD linelist = 8<; Do@linelist = Append@linelist, Join@8n@all@@iDD@@1DDD ê all@@iDD@@1DD<, 8Hi - 1L ê 2<DD, 8i, Length@allD<D êê Creates the list of points used for graphical thickness measurement ell@x_D = ReplaceAll@h * x - b ê 2, FindFit@linelist, h * x - b ê 2, 8b, h<, xDD êê The parameter b is varied by hand Show@ListPlot@Drop@linelist, -1D, PlotStyle Ø [email protected]<D, Plot@ell@xD, 8x, 0, .005<, PlotStyle Ø [email protected]<DD êê Makes the graph to determine fit accuracy d = ReplaceAll@h ê 2, FindFit@linelist, h * x - b ê 2, h, xDD êê Finds thickness from slope of fit line Appendix A.nb 3 Clear@a, an, acD; 4 n s Hdown+upL 16 n s Hdown+upL ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅÅÅ -$%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -H-1+n2 L3 Hn2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -s4 L+ ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ %%%%%%%%%%%%%%%% ÅÅÅÅÅÅÅÅ ÅÅÅÅ %%%%%%% ÅÅÅÅ down up 2 2%%%%%%%% LogA ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ E H-1+nL3 Hn-s2 L a@dee_, n_, s_, up_, down_D = - ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ; dee 2 4 2 down 2 up êê Finds the absorption coefficient a@lD an@x_D = a@d, n@xD, s@xD, TM@xD, Tm@xDD; ac@x_D = a@d, cauchy@xD, s@xD, TM@xD, Tm@xDD; Clear@T, Tn, TaD; T@n_, s_, dee_, a_, x_D = H16 ‰-a *dee n2 sL í 4 * dee * n p J‰-2 a *dee H-1 + nL3 Hn - s2 L + H1 + nL3 Hn + s2 L + 2 ‰-a* dee H-1 + n2 L Hn2 - s2 L CosA ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ EN; x Tn@x_D = T@n@xD, s@xD, d, an@xD, xD; Tc@x_D = T@n@xD, s@xD, d, ac@xD, xD; êê Builds the reconstructed transmission curves Plot@8intdata@xD, Tn@xD, Tc@xD<, 8x, 200, 2000<, PlotStyle Ø [email protected], [email protected], [email protected]<D êê Plots the reconstruction Appendix A.nb Clear@ao, cut, atotn, atotc, ad, fit, Ncm, cut, dopeD; ao@x_D = -H1 ê dL Log@intdata@xD ê Ht@473DLD; ad@x_D = -H1 ê dL Log@intdata@xD ê Ht@1200DLD; êê Exponential absorption features near band edge and in infrared fit@x_D = ReplaceAll@s * Hx - pL ^ 2 + b, FindFit@Transpose@Join@8Array@Identity, points, startD<, 8Array@ad, points, startD<DD, s * Hx - pL ^ 2 + b, 8s, b, p<, xDD êê Fits infrared absorption to l2 dependence to find carrier density Plot@8fit@xD, ad@xD<, 8x, 500, 2000<, PlotStyle Ø [email protected], [email protected]<, PlotRange Ø 88500, 2000<, 8-.0001, .0005<<D êê Plots infrared absorption and fit to assess fit accuracy cut = 473; dope = 1200; atotn@x_D = ao@xD * UnitStep@cut - xD + an@xD * UnitStep@x - cutD * UnitStep@dope - xD + Hfit@xDL * UnitStep@x - dopeD; atotc@x_D = ao@xD * UnitStep@cut - xD + ac@xD * UnitStep@x - cutD * UnitStep@dope - xD + ad@xD * UnitStep@x - dopeD; êê Builds full absorption coefficient Clear@Ttotn, TtotcD; Ttotn@x_D = T@n@xD, s@xD, d, atotn@xD, xD; Ttotc@x_D = T@n@xD, s@xD, d, atotc@xD, xD; êê Builds full reconstructions Plot@8intdata@xD, Ttotn@xD, Ttotc@xD<, 8x, 300, 2000<, PlotStyle Ø [email protected], [email protected], [email protected]<D êê Plots reconstructions 4 Appendix A.nb Clear@h, c, sub, e, low, high, asq, afitD h = 4.14 * 10 ^ -15; c = 3 * 10 ^ 17; sub = h * c ê e; e@x_D = h * c ê x Plot@8ao@subD^ 2<, 8e, 3, 4.5<, PlotStyle Ø [email protected], [email protected]<, PlotRange Ø 883, 4.5<, 80, .0004<<D; asq@x_D = ao@xD ^ 2; afit@e_D = ReplaceAll@m He - aL, FindFit@Transpose@ Join@8Array@e, points, startD<, 8Array@asq, points, startD<DD, m He - aL, 8m, a<, eDD Plot@8ao@subD^ 2, afit@eD<, 8e, 3, 4.5<, PlotStyle Ø [email protected], [email protected]<D; êê This linear fitting routine finds the bandgap energy Clear@high, low, alog, aurbD Plot@8Log@ao@subDD<, 8e, 3, 4<, PlotStyle Ø [email protected], [email protected]<D alog@x_D = Log@ao@xDD; aurb@e_D = ReplaceAll@e ê u + b, FindFit@Transpose@ Join@8Array@e, points, startD<, 8Array@alog, points, startD<DD, e ê u + b, 8u, b<, eDD Plot@8Log@ao@subDD, aurb@eD<, 8e, 1, 6.2<, PlotStyle Ø [email protected], [email protected]<, PlotRange Ø 882.8, 4.7<, 8-8, -4<<D êê This linear fitting routine finds the Urbach disorder parameter 5