RF magnetron sputtering deposition of multilayers optical filters for ultra-broadband applications with a large number of thin layers

This paper presents advancements in manufacturing ultra-broadband optical multilayer filters with over 100 thin layers using improved RF magnetron sputtering techniques, specifically deposition duration control and in situ optical reflectance monitoring, which were optimized through cross-studies of Nb2O5, TiO2, and SiO2 refractive indexes.

The Gist

Imagine you are trying to build a super-precise sandwich, but instead of bread and ham, you are stacking hundreds of microscopic layers of glass and metal. The goal? To create a special "optical filter" that acts like a bouncer at a club, letting only specific colors of light in while blocking others.

This paper is about a team of scientists who figured out how to build these sandwiches with extreme precision, even when the stack gets incredibly tall (over 100 layers).

Here is the breakdown of their work using simple analogies:

1. The Goal: The "Perfect Light Filter"

Think of light as a rainbow of colors. Sometimes, you want a mirror that reflects all colors (like a shiny car hood), and sometimes you want a filter that blocks just the red ones but lets the blue ones through.

• The Challenge: To make these filters work perfectly, you need to stack layers of "High" refractive index material (let's call it Heavy Glass) and "Low" refractive index material (Light Glass).
• The Problem: If you stack 10 layers, a tiny mistake in one layer is okay. But if you try to stack 100 layers, a tiny mistake in the first layer gets magnified. By the time you reach the 100th layer, the whole sandwich is ruined. The scientists call this the "Avalanche of Deviation." It's like trying to build a tower of Jenga blocks where you accidentally nudge the bottom one; the whole thing eventually topples.

2. The Tools: Two Different "Spray Painters"

To build these layers, the team used two different machines (RF magnetron sputtering). Imagine these as two different chefs using two different spray guns to paint the layers.

• Chef A (The Timer): This chef sprays the material for a set amount of time. They guess how long to spray based on a stopwatch.
• Chef B (The Eye): This chef has a super-visor (an optical monitoring system) that watches the light bouncing off the glass while it's being sprayed. As soon as the layer reaches the perfect thickness, the machine says, "Stop!" and cuts the spray.

3. The Ingredients: The "Big Three"

They tested three main ingredients to see which made the best "sandwich":

• SiO₂ (Silica): The "Light Glass" (Low index).
• Nb₂O₅ (Niobium Oxide): A "Heavy Glass."
• TiO₂ (Titanium Dioxide): An even "Heavier Glass."

The Discovery: They found that TiO₂ and SiO₂ make the best pair because they have the biggest difference in "heaviness" (refractive index). It's like trying to make a sound echo in a canyon; the bigger the difference between the walls, the better the echo. However, TiO₂ is a bit "greasy" (absorbs a little light) in the blue/violet range, while Nb₂O₅ is cleaner but slightly less "heavy."

4. The Breakthrough: Calibrating the Chefs

The biggest hurdle was that the two chefs (machines) didn't agree on how thick the layers were.

• The team used a high-tech microscope (Ellipsometry) to measure the exact thickness and "heaviness" of single layers made by both chefs.
• They realized that even though the machines were different, they could make the layers identical if they adjusted the recipe (power settings and gas mix).
• They also measured the "roughness" of the surface. Imagine looking at a smooth mirror vs. a rough piece of sandpaper. If the layers are too rough, light scatters like a ball hitting a bumpy floor, and the filter fails. They found their layers were smooth enough (mostly) to work, though some machines produced slightly rougher surfaces than others.

5. The Result: Building the 36-Layer Tower

Once they calibrated their tools, they built a 36-layer "Bragg Reflector" (a fancy mirror that reflects a huge range of colors).

• The Test: They compared the real mirror to a computer simulation.
• The Outcome: The mirror worked! It reflected light perfectly from 450nm (blue) to 1200nm (infrared).
• The Catch: Even with perfect tools, the layers weren't exactly the thickness the computer predicted. Some were a tiny bit too thick, some too thin. The "avalanche" of errors started to show up in the thinner layers, causing a small drop in performance (about 2.7% less reflection than perfect).

6. The Future: The 100-Layer Dream

The paper concludes that they have successfully mastered the art of making these complex stacks.

• Why it matters: If you can make a 36-layer stack with high precision, you can now aim for 100+ layers.
• The Application: This opens the door to "Ultra-Broadband" filters. Imagine a camera lens that can see everything from deep ultraviolet to far-infrared without needing to swap lenses, or a communication system that can send data using a massive spectrum of light without interference.

In a Nutshell

The scientists took two different ways of building optical filters, measured them with extreme care, fixed the differences, and proved they can build a 36-layer "light sandwich" that works almost perfectly. This success is the first step toward building massive, 100-layer structures that will revolutionize how we use light in technology.

Technical Summary

1. Problem Statement

The fabrication of complex optical multilayer stacks (containing more than 100 layers) for ultra-broadband applications faces a critical challenge known as the "avalanche of deviation." This phenomenon refers to the cumulative accumulation of errors in layer thickness ($e$) and optical constants ($n$ and $k$) as the number of layers increases. Even minor deviations in individual layers can compound, leading to significant discrepancies between the designed and actual optical performance of the final filter. The authors aim to quantify and reduce these errors to enable the reliable manufacturing of high-complexity optical systems.

2. Methodology

The research employed a comparative approach using two distinct RF magnetron sputtering systems to deposit thin films of three dielectric materials: Nb₂O₅, TiO₂, and SiO₂.

• Deposition Systems:
  1. AJA International System: Utilized deposition duration control. Deposition rates were pre-evaluated, and layer thickness was controlled by timing, verified via ex-situ spectroscopic ellipsometry.
  2. Elettrorava S.p.A. System: Utilized in-situ Optical Monitoring System (OMS). This system monitored absolute reflectance in real-time during deposition at specific wavelengths (550–1650 nm) to trigger cut-off conditions at turning points.
• Materials & Substrates: Films were grown on p-type [001] 2'' Silicon substrates in an O₂/Ar plasma environment.
• Optical Characterization:
  • Ellipsometry: Variable-angle spectroscopic ellipsometry (Horiba UVISEL) was used to determine the complex refractive index ($n$ and $k$) of monolayers.
  • Spectrophotometry: Specular reflectance spectra were measured (Perkin Elmer Lambda 1050) from 250 nm to 2500 nm.
  • Modeling: Data was fitted using the New Amorphous (NA) model for SiO₂ and TiO₂, and the Tauc-Lorentz (TL) model for Nb₂O₅.
  • Surface Analysis: Atomic Force Microscopy (AFM) was used to measure surface roughness (RMS) and calculate light scattering losses.
• Multilayer Fabrication: Two Broad Band Bragg Reflectors (BBBR) consisting of 36 layers were fabricated: one using Nb₂O₅/SiO₂ and the other using TiO₂/SiO₂.

3. Key Contributions

• Cross-Validation of Deposition Techniques: The study successfully demonstrated that two different sputtering approaches (time-controlled vs. in-situ optical monitoring) yield highly consistent optical properties for Nb₂O₅, TiO₂, and SiO₂ when normalized for power density and gas ratios.
• Advanced Material Characterization: The authors provided precise dispersion models ($n(\lambda)$ and $k(\lambda)$) for the three materials across the UV-NIR spectrum, utilizing both NA and TL models with Monte Carlo uncertainty analysis.
• Quantification of Optical Losses: The study quantified the contribution of surface roughness to optical losses, calculating light scattering ratios ranging from 1.75% to 4.35% depending on the material and deposition method.
• Error Analysis in Multilayers: The paper analyzed the thickness deviations in a 36-layer stack, identifying that the most significant deviations occurred in thinner layers (<70 nm), providing a roadmap for process optimization.

4. Key Results

• Monolayer Consistency:
  • Nb₂O₅: Both systems produced nearly identical refractive index and extinction coefficient curves.
  • TiO₂ vs. Nb₂O₅: TiO₂ exhibited a slightly higher refractive index than Nb₂O₅ in the VIS-NIR range (2.285–2.712 vs. 2.239–2.583), offering better contrast with SiO₂. However, TiO₂ absorbs at ~420 nm (2.98 eV), whereas Nb₂O₅ absorbs at ~380 nm (3.19 eV).
  • SiO₂: Both systems produced SiO₂ with $n \approx 1.47–1.54$ and negligible extinction ($k \approx 0$). The Elettrorava system showed a slight effective medium (EM) layer (~2.4 nm) due to roughness, which was absent in the AJA system.
• Multilayer Performance (36-Layer BBBR):
  • Nb₂O₅/SiO₂ Stack: Achieved strong reflection from 450 nm to 1200 nm. The experimental reflectance showed a slight decrease (2.7% ± 0.7) compared to the theoretical Macleod simulation, with a $\chi^2$ deviation of 12.45 ± 2.25.
  • Thickness Deviation: The absolute thickness deviation between theoretical and simulated layers ranged from 0.4% to 20%, with the highest errors in the thinnest layers.
  • Bandwidth: The Nb₂O₅/SiO₂ filter achieved a bandwidth ($\Delta\lambda/\lambda$) of 1.18 ± 0.02 at a central wavelength of 900 nm.
• Scattering Losses: AFM analysis revealed RMS roughness values of 1.4 nm (TiO₂), 3.4 nm (Nb₂O₅), and 0.74 nm (SiO₂), correlating directly to the calculated scattering losses.

5. Significance

This work represents a significant step forward in the manufacturing of ultra-broadband optical filters with high layer counts. By rigorously characterizing material optical constants and comparing two complementary deposition strategies, the authors have established a reliable workflow to mitigate the "avalanche of deviation."

The successful fabrication and characterization of 36-layer stacks with high reflectance performance validate the feasibility of scaling up to 100+ layer systems. This capability is crucial for next-generation applications in scientific instrumentation, telecommunications, and advanced optical systems requiring complex, high-performance spectral filtering. The study provides a foundational framework for reducing cumulative errors in thin-film deposition, moving the industry closer to the theoretical limits of multilayer optical design.