Abstract
We use a state-of-the-art optimization algorithm combined with a careful methodology to find optimal anti-reflective coatings. Our results show that ultra thin structures (less than $300 \,nm$ thick) outperform much thicker gradual patterns as well as traditional interferential anti-reflective coatings. These optimal designs actually combine a gradual increase of the refractive index with patterns meant to leverage interferential effects. Contrarily to gradual patterns, they do not require extremely low refractive index materials, so that they can actually be fabricated. Using a cheap and easily deployable vapor deposition technique, we fabricated a 4-layer anti-reflective coating, which proved very efficient over the whole visible spectrum despite a total thickness of only 200 nm.
Introduction: The Challenge of Anti-Reflective Design
Anti-reflective coatings (ARCs) are critical components in optical systems, ranging from photovoltaic cells to high-end camera lenses. Historically, two distinct physical paradigms have dominated ARC design: the graded-index approach and the interferential approach.
The graded-index approach relies on a smooth transition of the refractive index from air to the substrate to minimize reflections across broad bandwidths and angles. However, this typically requires coatings several microns thick and materials with refractive indices near 1.0, which are physically fragile and difficult to manufacture. Conversely, the interferential approach uses thin layers (e.g., quarter-wave layers) to induce destructive interference. While thin, these designs are often narrow-band and sensitive to the angle of incidence. This research addresses the "missing middle": creating ultra-thin coatings (under 300 nm) that achieve the high performance of thick graded-index structures.
Methodology: Global Optimization via Differential Evolution
The researchers framed the ARC design as a global optimization problem. They divided a 200 nm thick design space into $N$ sub-layers or "pixels" of equal thickness. To navigate the complex, "rugged" solution landscape, they employed Differential Evolution (DE), specifically the DE2 variant. DE is a population-based stochastic optimization algorithm superior to gradient descent for finding global minima in non-convex spaces.
The performance of these pixelated structures was simulated using the S-matrix method, calculating reflectivity across the visible and near-infrared spectrum (400–900 nm) for various angles of incidence (0° to 80°). To validate the theoretical findings, a 4-layer coating was fabricated using reactive magnetron sputtering of silicon oxynitride ($Si_xO_yN_z$), a standard industrial process.
Key Insights: The Hybrid "Chopped" Gradient
A central discovery of this work is that the optimal profile for an ultra-thin ARC is not a smooth curve. As the number of pixels increases, the optimization converges on a "chopped" or discontinuous gradient. This profile represents a hybrid physical strategy: it utilizes a general gradual increase in refractive index for broadband performance while incorporating sharp discontinuities to leverage interferential effects.
This hybrid approach is particularly fabrication-friendly. Unlike theoretical graded-index models that require a starting refractive index as low as 1.05, DE-optimized designs remain effective starting at an index of 1.2, which is readily achievable with common materials. Furthermore, the designs are robust; a 10% manufacturing error in the refractive index only marginally increases average reflectivity to approximately 4.1%.
Results: Performance and Experimental Validation
The DE-optimized coatings demonstrate significant performance gains over traditional profiles. A 200 nm thick optimized coating reduced the average reflectivity of silicon from 41% to just 3.3%.
As shown in the figure above, the DE2-optimized coating at 200 nm thickness matches the performance of 1,000 nm thick Gaussian or Quintic coatings, representing a 5x reduction in thickness for the same performance.
Experimental results confirmed these simulations. A fabricated 4-layer sample with a total thickness of 200 nm achieved reflectivity between 0.56% and 7.4% across the 400–900 nm range.
Conclusion and Future Directions
This study demonstrates that combining global optimization with hybrid physical principles allows for the design of ultra-thin, high-performance anti-reflective coatings. By moving away from smooth gradients toward "chopped" profiles, researchers can achieve significant thickness reductions while maintaining ease of fabrication.
Future work could involve incorporating material dispersion and optical losses directly into the optimization loop to further refine performance. Additionally, extending this methodology to other spectral ranges, such as the ultraviolet or mid-infrared, could open new applications for these ultra-thin structures.
Summary updated