Fluidic Shaping over arbitrary domains: theory and high order finite-elements solver

This paper establishes the theoretical foundation for Fluidic Shaping over arbitrary domains and presents a novel high-order (quintic) finite element solver capable of accurately predicting the topography and curvature of complex optical liquid surfaces, overcoming the limitations of previous analytical and numerical methods restricted to linearized equations or simple geometries.

The Gist

The Big Idea: Sculpting with "Floating" Liquid

Imagine you want to make a perfect camera lens or a pair of glasses. Traditionally, you have to take a block of glass, grind it down with sandpaper, polish it for days, and hope you didn't make a mistake. It's slow, expensive, and creates waste.

The authors of this paper are working on a method called Fluidic Shaping. Think of it like this: instead of grinding glass, you take a blob of liquid polymer (like a fancy, clear glue) and drop it into a tank of water.

But here's the trick: the liquid polymer is mixed so perfectly with the water that it has neutral buoyancy. It doesn't sink, and it doesn't float up. It just hovers, weightless, like a jellyfish in the deep ocean.

Because it has no weight pulling it down, the only thing shaping the liquid is surface tension (the "skin" on the water) and the shape of the container holding it. If you pin the edges of the liquid to a specific frame, the liquid naturally settles into a mathematically perfect curve. This curve can be a lens, a mirror, or any other optical shape.

The Problem: The "Rough Sketch" vs. The "Masterpiece"

The authors had already figured out how to do this for simple shapes, like a perfect circle (like a round cookie cutter). They could predict the shape of the liquid using math.

However, real-world glasses aren't always round. They are often oval, or have weird curves to fix specific vision problems (astigmatism). When you try to use a non-round frame (an "arbitrary domain"), the math gets incredibly messy.

• The Old Way: Previous computer programs could only solve this for simple, round shapes, or they used "low-resolution" math. It was like trying to draw a detailed portrait using only a few large, blocky pixels. The result was close, but not good enough for high-precision optics.
• The Issue: For a lens to work perfectly, the computer needs to know not just the shape of the liquid, but exactly how curved it is at every single point. If the math is slightly off, the lens will blur your vision.

The Solution: The "High-Definition" Solver

The authors built a new computer program (a "solver") to handle these complex, weird shapes. Here is how they did it, using an analogy:

1. The "Deformed" Tiles (High-Order Elements)
Imagine you are tiling a floor.

• Low-Order Method: You use square tiles. If the wall you are tiling against is curved, you have to chop the tiles into jagged pieces to fit. There are gaps and rough edges. This creates errors.
• The Authors' Method: They invented special "deformed" tiles. These tiles are flexible; they can stretch and bend to match the curve of the wall perfectly, leaving no gaps.
• Why it matters: In their math, these "tiles" are actually 5th-degree polynomials (very complex curves). This allows the computer to describe the liquid's surface with extreme smoothness, capturing the tiny details needed for a perfect lens.

2. The "Smart" Boundary
When the liquid hits the edge of the frame, the math has to be very careful. If the frame has a sharp corner (like a hexagon), the liquid behaves differently than if the frame is a smooth circle.
The authors created a special "translator" for the computer. It tells the math: "Hey, at this sharp corner, the rules change. Don't try to force a smooth curve here; adjust the calculation to fit the sharp point." This ensures the computer doesn't get confused by weird shapes.

Why This is a Big Deal

The paper proves that their new computer code is incredibly accurate. They tested it in three ways:

1. The "Fake" Test: They told the computer, "Pretend the answer is a perfect sphere," and checked if the code could find it. It did, with errors smaller than the width of a single atom.
2. The "Round" Test: They compared their code to old math formulas for round lenses. They found that the old formulas were actually 500 nanometers off. That sounds tiny, but for a precision lens, that's huge—it's like the difference between a blurry photo and a 4K photo. Their new code fixed this.
3. The "Real World" Test: They simulated making glasses for people with astigmatism (which require oval, not round, lenses). They showed that their code could predict exactly how the liquid would shape itself, even if the manufacturing frame had tiny defects or if the liquid volume was slightly off.

The Bottom Line

This paper is about building a super-accurate digital blueprint for making lenses out of floating liquid.

Before this, we could only make simple, round lenses with this method. Now, thanks to this new "high-definition" math, we can design and predict the behavior of complex, custom-shaped lenses (like those needed for modern glasses or advanced cameras) before we even pour the liquid.

It turns the art of "guessing and grinding" into the science of "predicting and pouring," potentially making high-quality custom optics faster, cheaper, and easier to make.

Technical Summary

1. Problem Statement

Fluidic Shaping is a fabrication method for optical components where a volume of liquid polymer is submerged in an immiscible immersion liquid of equal density (neutral buoyancy). The liquid is pinned to a geometric boundary (a frame or dish), and surface tension combined with gravity shapes the liquid interface into a desired optical surface.

While previous work provided analytical solutions for linearized equations on circular or elliptical domains, these approaches are insufficient for modern optical needs because:

• Nonlinearity: High-precision optics require solving the full nonlinear partial differential equation (PDE), as linearization introduces significant errors (hundreds of nanometers) when boundary variations are large.
• Arbitrary Domains: Complex optical prescriptions (e.g., ophthalmic lenses with astigmatism) require non-circular, arbitrary boundary footprints, which analytical methods cannot handle.
• Curvature Accuracy: Optical power depends on the second derivatives (curvature) of the surface. Standard low-order numerical solvers often fail to provide the necessary accuracy for these derivatives.
• Geometry Mismatch: Standard finite element methods (FEM) approximate curved boundaries with straight edges, introducing geometric errors that degrade the convergence rate of high-order solutions.

The authors aim to develop a high-order numerical solver capable of solving the nonlinear Fluidic Shaping problem over arbitrary domains with the precision required for optical applications (accuracy $\approx \lambda/10 \approx 50$ nm).

2. Methodology

Mathematical Formulation

The problem is formulated as a minimization of the total energy functional $\Pi[u; P]$, consisting of surface tension energy and gravitational potential energy, subject to a global volume constraint.

• Governing Equation: A highly nonlinear PDE with Dirichlet boundary conditions ($u = \bar{u}$ on the boundary $\Gamma$) and an integral volume constraint.
• Lagrange Multiplier: A pressure term $P$ is introduced to enforce the volume constraint.
• Generalization: The framework is extended to handle multiple interfaces (e.g., meniscus lenses) and varying Bond numbers ($Bo$) across the domain.

Numerical Solver: Reduced Quintic Finite Elements

To achieve the required accuracy for surface curvature (second derivatives), the authors employ reduced quintic (5th-order) triangular finite elements.

• Degrees of Freedom (DOF): Each node possesses 6 DOFs: the function value ($u$), first derivatives ($u_x, u_y$), and second derivatives ($u_{xx}, u_{yy}, u_{xy}$). This ensures $C^1$ continuity, essential for accurate curvature calculation.
• Shape Functions: The authors utilize Bell's reduced quintic elements, which span the approximation space with high-order accuracy and continuity.

Handling Arbitrary and Curved Boundaries

A key innovation is the treatment of the domain boundary to eliminate geometric mismatch errors:

1. Boundary Node Transformation: For nodes on the boundary, the nodal DOFs are transformed from the global coordinate system to the local boundary parameterization ($\bar{u}, \bar{u}_t, \bar{u}_{tt}$) to strictly enforce the Dirichlet condition. Separate transformations are derived for smooth boundaries and sharp points (corners).
2. Deformed Elements: Instead of approximating a curved boundary with straight element edges, the authors use a transfinite mapping to deform the triangular elements so their edges match the exact curved boundary.
   • This transformation modifies the Jacobian of the integration but preserves the nodal DOFs.
   • This approach prevents the "geometry mismatch error" that typically limits the convergence rate of high-order methods to $O(h^2)$.

Implementation

• Galerkin-Ritz Formulation: The functional is discretized, leading to a nonlinear system of equations solved via Newton iterations.
• Quadrature: High-order Gauss-Legendre quadrature is used for numerical integration over the parent element.
• Code Availability: The solver is implemented in Python and made available via GitHub.

3. Key Contributions

1. Theoretical Foundation: Established a general mathematical description for Fluidic Shaping over arbitrary domains, including multi-interface configurations and variable Bond numbers.
2. High-Order Solver Development: Developed a dedicated solver using reduced quintic finite elements capable of resolving both the surface topography and its second derivatives (curvature) with high precision.
3. Curved Boundary Treatment: Demonstrated that deforming finite elements to match the physical boundary is critical. This eliminates geometric mismatch errors, restoring the theoretical convergence rate of the high-order method.
4. Validation against Analytical Solutions: Proved that linearized analytical solutions are insufficient for precision optics when boundary variations are large, whereas the nonlinear numerical solver captures the true physics.

4. Results

Accuracy and Convergence

• Convergence Rates: The solver was tested against a manufactured spherical solution.
  • Deformed Elements: Achieved convergence rates of $O(h^{4.2})$ for $L^2$ norm, $O(h^{3.0})$ for $H^1$, and $O(h^{2.0})$ for $H^2$ (curvature).
  • Non-Deformed/Linear Elements: Showed significantly lower convergence rates (e.g., $O(h^{0.75})$ for $H^2$ with non-deformed elements), confirming that geometry mismatch destroys high-order accuracy.
• Error Magnitude: For a 3.5 cm lens with 550 $\mu$m boundary variation, the linear analytical solution deviated from the numerical solution by ~500 nm, which is unacceptable for precision optics. The numerical solver achieved errors on the order of $10^{-11}$ m.

Application to Optical Design

• Ophthalmic Lenses: The solver successfully modeled lenses on eyewear-shaped (non-circular) domains, calculating geometric spherical and cylindrical powers. It performed sensitivity analysis on manufacturing errors (boundary height, liquid density, volume injection), showing how these affect the final optical power.
• Micro-lens Arrays: The code was used to analyze the trade-off between fill factor and optical quality.
  • Circular Arrays: Perfectly spherical but low fill factor (gaps between lenses).
  • Hexagonal Arrays: High fill factor but non-spherical topographies due to flat boundary constraints. The solver quantified the resulting optical aberrations.

5. Significance

• Enabling Complex Optical Fabrication: This work removes the geometric limitations of Fluidic Shaping, allowing for the design and prediction of complex, freeform optical components (e.g., custom eyewear, non-rotationally symmetric lenses) that were previously impossible to model accurately.
• Precision Manufacturing: By providing a tool to predict optical properties (power, astigmatism) with nanometer-scale accuracy, the solver enables the definition of manufacturing tolerances and the optimization of process parameters (volume, density, frame shape).
• Inverse Design Potential: The solver lays the groundwork for inverse design algorithms, where one can determine the necessary boundary conditions and liquid parameters to achieve a specific target optical surface.
• Broader Applicability: The methodology of using deformed reduced quintic elements is applicable to other mechanical problems involving high-order differential operators, such as thin liquid film evolution and thin plate deformation.

In conclusion, the paper presents a robust, high-fidelity numerical framework that bridges the gap between the theoretical potential of Fluidic Shaping and its practical application in high-precision optical manufacturing.