A team of researchers at the University of Melbourne has developed a low-cost prototype of a liquid-based geometric waveguide combiner for augmented reality (AR) applications. The device replaces complex fabrication steps used in conventional optical combiners by integrating silicone oil and PolyJet 3D printing in a simplified production method. The work was published in a journal by Springer Nature.
Geometric waveguides are key components in AR near-eye displays, allowing virtual content to be optically superimposed over the real world. Traditional fabrication methods involve dicing, layer bonding, and polishing of stacked glass substrates with reflective coatings, all of which require high precision and significant labor. The liquid waveguide developed by the Melbourne team avoids these steps entirely by sealing a silicone oil-filled 3D printed frame between cover glasses, with dielectric reflectors inserted into pre-defined slots.
Dimensions of the prototype measure 38 mm in length, 28.5 mm in width, and 3.2 mm in thickness. A triangular prism with a 50° slant angle is used to couple light into the waveguide, while three parallel dielectric reflectors, spaced 2.024 mm apart and angled at 25° relative to the base, redirect light toward the viewer. These structural features were optimized using COMSOL Multiphysics with the finite element ray optics module. The optical simulation confirmed that total internal reflection maintained light transmission efficiency, with a checkerboard image projected via a collimating lens and reconstructed on a simulated retina.
Stratasys, a manufacturer of industrial 3D printers, provided the PolyJet platform used in fabrication. The researchers used its J826 system with transparent RGD810 resin and water-soluble SUP707 support material. A thin film of fluorinated ethylene propylene or flat glass was applied to the print bed to ensure a smooth finish. Slots for reflector insertion were printed at widths ranging from 165 µm to 200 µm, with 180 µm determined to be optimal for alignment. The printed reflectors were 175 µm thick. Venting holes of 0.5 mm diameter were incorporated to assist silicone oil injection and air displacement during assembly.
Dielectric bandpass filters were inserted using precision tweezers under a microscope. Parallelism between reflectors was visually inspected using ceiling light reflections, and distance measurements were performed at multiple points using a microscope stage with 1 µm resolution. Although this process is time-consuming, the researchers propose omitting the inspection at early stages and replacing it with visual inspection techniques. More accurate alignment methods, such as laser autocollimators or Fizeau interferometers, were noted as areas for future improvement.
The frame was sealed with three 0.1 mm-thick cover glass pieces, cleaned by plasma and bonded with UV epoxy under controlled pressure. After sealing, silicone oil with a viscosity of 20 cSt was injected into the waveguide at a flow rate of 0.2 ml/min using a 27-gauge needle and syringe pump. UV resin was applied to seal the vent hole. The oil, chosen for its thermal and chemical stability, has a refractive index of 1.41 and exhibits only a 0.28° change in critical angle over a 20 °C temperature range.
Optical characterization was carried out using a Cytoviva hyperspectral microscope. Reflector transmittance ratios showed a 3% drop across the 400–700 nm visible spectrum at 0° incidence, and a 2% reduction on average at 25°. Surface roughness was measured using a Bruker Dimension Icon atomic force microscope, with root mean square roughness (Rq) values of 1.4 nm for reflectors and 1.3 nm for cover glass. Modulation transfer function (MTF) performance was assessed at 15 cycles per degree, showing 70% retention in see-through mode and 24% in virtual image mode.
Horizontal and vertical fields of view were measured at 19.52° and 12.56°, respectively, by projecting onto a target board placed 25 cm from the eye pupil plane. Due to the use of only three reflectors, the eyebox was limited to approximately 1 mm horizontally and 2 mm vertically. The researchers note that increasing the number of reflectors could expand the eyebox and FOV but would result in greater optical losses and higher power consumption.
Reflection efficiency measurements were performed indirectly in a darkroom using a monochrome CMOS sensor. The team compared the projector’s original light output with the redirected light from the waveguide. Recorded efficiencies were 4.48% at 460 nm, 4.49% at 515 nm, and 4.56% at 625 nm.
Manual assembly posed challenges in precisely cutting and inserting reflectors without damage. Although labor-intensive, the current assembly process allows for fast prototyping iterations. The full process includes approximately 3 hours of 3D printing, 2 hours of post-processing, and 4 hours of manual assembly. Total material cost was calculated at $18, including $12 for reflectors, $3.30 for equipment use, $1.10 for resin, $0.70 for silicone oil, and $0.60 for cover glasses.
To address thermal expansion of the liquid medium—estimated at 900 ppm/°C—the design includes optional air pockets to prevent pressure buildup. These bubbles rise to the top and can be hidden in the upper sealed region. More expensive high-index transparent liquids such as microscope immersion oils were discussed as alternatives but deemed cost-prohibitive for prototyping.
Waveguide performance may also be limited by the choice of projection system. The team used a standard LCoS projector, which may not align optimally with the prism’s pupil location, potentially reducing the effective eyebox. Custom-designed projectors were proposed for future development.
FiconTEC, a German photonics automation company, collaborated with the team to develop a custom waveguide assembly system. The apparatus integrates a gantry system, pressure-sensing reflectors pickers, a UV glue preparation platform, and high-resolution imaging systems for alignment. Its current evaluation aims to support scalable production of waveguides based on the team’s patented method.
The authors selected glass as the chamber sealing material for simplicity, noting a stable transmittance ratio of 93%. They recommend sapphire as a potential future replacement due to its hardness and scratch resistance. Additional improvements could include anti-reflective coatings such as MgF₂ or TiO₂/SiO₂ to reduce reflection losses at wider viewing angles.
Researchers involved in the project are affiliated with the University of Melbourne’s Department of Electrical and Electronic Engineering, the Neural Dynamics Laboratory in the Department of Medicine, and KDH Advanced Research Pty Ltd.
Ready to discover who won the 2024 3D Printing Industry Awards?
Subscribe to the 3D Printing Industry newsletter to stay updated with the latest news and insights.
Featured image shows optical performance of the fabricated waveguide. Image via Springer Nature.
