The market for augmented reality (AR) portable and wearable devices is rapidly growing. Among various hardware implementation forms, head-mounted display (HMD) or near-eye display (NED) with see-through glasses offers the most effective and immersive AR experience. Optical waveguide, due to its thin and light nature, is considered to be an unparalleled choice for consumer-grade augmented reality (AR) glasses, but is still prohibitive due to its high price and technical barrier. As the mainstream AR wearable devices such as Hololens II and Magic Leap One have adopted waveguide solution and demonstrated its mass production capability, as well as the recent disclosure of financing news for AR optical module manufactures DigiLens, NedAR, and LingXi, waveguide has become the hot topic in AR glass industries. The optical system for NED is usually comprised of a microdisplay and imaging optics.

How does optical waveguide work in AR NED system? What is the relationship between the so-called “array waveguide”, “geometric waveguide”, “diffractive waveguide”, “holographic waveguide”, and “volumetric waveguide”? How was waveguide developed on the way of revolutionizing the AR glass industry?

1. Optical Waveguide — Born on Demand

The optical system is usually comprised of a microdisplay and imaging optics, for both VR and AR near-eye displays (NED). A microdisplay provides the image either actively like in micro OLED or the trendy micro LED panel, or indirectly by illuminating light on liquid-crystal based display (including transmissive LCD, and reflective LCOS), and digital micromirror device (DMD) as well as laser beam scanner (LBS) both enabled by microelectromechanical system (MEMS). Similar to VR, the display pixels are imaged to a certain distance and formed a virtual image to project to human eyes. Different from VR, AR NED needs the “see-through” functionality for the eye to be able to view the real world at the same time. The imaging system cannot block the front view, which therefore requires one or several additional optical elements to form an “optical combiner”. The optical combiner reflects virtual image while transmitting external light to human eye, overlaying the virtual content on top of the real scene, for them to complement and “augment” each other.

Figure 1. (a) Illustration of optical system in VR near-eye display; (b) Illustration of optical system in AR near-eye display.

Among all the current optical solutions though, we think optical waveguide has the best potential to enable the consumer-grade AR glasses, based on its optical performance, appearance, and mass production capability.

Figure 2. Categories of AR optical solutions, their representing products, and characteristics.


2. Pros and Cons of Waveguide Technology

The optical waveguide technology was introduced most recently as a unique type of optical combiner as it generally carries no optical power. It is not a completely new concept though, which works in the same way as optical fibers for communication network. The only difference is the later transports infrared light rather than visible light in our case. In order for the light to be bounced back and forth inside the waveguide like a swimming snake, “total internal reflection (TIR)” is the key. Two conditions need to be met for TIR to take place: (1) high refractive index material in waveguide (n1> n2); (2) input angle of light is larger than the critical angle θc.

After the optical engine generates a virtual image, the waveguide couples in the image, transports it inside the glass substrate through TIR with almost zero leakage, then couples the image out when reaching the position of viewer’s eye. Throughout this process, the waveguide typically does not influence the image itself, thus it is an optical combiner independent of the imaging system.

Figure 3. Illustration of the total internal reflection (TIR) effect in optical waveguide (light guide).

The biggest advantage of using waveguide as an optical combiner is the optimization of space in the design of AR glasses. By having the microdisplay and imaging optics out of the way (either on the top of forehead or by the side) not only minimizes vision blocking, but also optimizes weight balancing of the device and improves ergonomics. The pros and cons of the waveguide configuration are listed below and will be explained throughout the context of this article.


  • Big eyebox and improved mechanical tolerance to fit more population — 1D and 2D exit pupil expansion.
  • Vision clearance and weight balancing — waveguide transporting virtual image to eye.
  • Eyewear appearance, close to consumer product — flat and thin eyepiece, good external light transmission.
  • Friendly for design iteration and mass production — flat glass substrate with customized contour, nanofabrication.
  • Multi-layer stackable — creating virtual images at different depths, going 3D.


  • Relatively low optical efficiency — low in/out coupling efficiency and sacrifice for big eyebox.
  • Geometric waveguide: complicated manufacturing process with possible low yield.
  • Diffractive waveguide: color dispersion from diffraction causes rainbow and haze effects influencing image quality.
  • Diffractive waveguide: high design barrier.

Figure 4. Illustration of the waveguide-based AR glass configuration. Image source:


3. Waveguide Categories and Comparison

As mentioned above, the major part of a waveguide is the transparent thin glass substrate (thickness usually varies from sub-nanometer to a few nanometers), carrying light bouncing between the top and bottom surfaces with very little leakage thanks to TIR. If you do a calculation for the range of input angles allowed to do TIR inside a waveguide, you’ll find the field-of-view (FOV) is limited by the refractive index of the glass. Thus to achieve higher FOV, glass manufacturers like Corning and Schott are developing high-index glass substrates on the wafer scale, especially for this market.

What differentiates the waveguide techniques from one another lies in the structures used to couple light in and out of the waveguide, connected by the TIR transportation of light in between. The optical waveguide can be generally categorized into geometric type and diffractive type. The geometric waveguide is the so-called “array waveguide”, which expands the eyebox in one dimension through an array of transflective mirrors. The pioneering optical company for geometric waveguide is Lumus. So far there is no mature AR glass product available in the market yet in large quantity. The diffractive waveguide covers the surface relief grating (SRG) structure and volumetric holographic grating (VHG) structure. Hololens and Magic Leap both use SRG structure, while Digilens pioneers in VHG as well as Akonia which was adopted by Apple last year. The VHG technology is relatively less mature, which currently offers limited FOV but could potentially provide better color performance. Due to the limit of the article length, we will introduce geometric waveguide first and save the diffractive waveguide to next time.

(1) Geometric Waveguide

The geometric waveguide was firstly introduced almost two decades ago and pioneered by an Israeli-based company Lumus. As shown in Figure 5 (a), light from the optical engine is coupled into the waveguide through a reflective mirror or prism structure. After several TIR bounces inside the glass substrate when reaching the location right in front of the viewer’s eye, light encounters an array of transflective surfaces to release the image. A transflective (transmission + reflection) surface is embedded inside the waveguide substrate at a certain angle to reflect part of the light to our eye, and transmit the rest of the light through for further propagation. It also transmits light from the real world so as to serve as an optical combiner. Then the transmitted light encounters another transflective surface and repeats the same transmission and reflection process.

In a conventional optical imaging system, the light only has one way out through the so-called “exit pupil”. Here the transflective surface repeats itself several times providing the same image output, thus expanding the exit pupil in the horizontal direction. This design is called “1D exit pupil expansion (EPE)”. You may wonder, would multiple exit pupils cause double image or shadowy image in our eyes? Don’t worry, the exit pupil is only the “Fourier plane” of the virtual image, and the human eye will convert the angular information from this plane to spatial information through its only lens. The image is then formed on the “image plane” — our retina, with all light rays at the same angle (even they are from different exit pupils) merged onto the same pixel, thus creating only one image. It may be a little too abstract to understand, but this is the essence of EPE. For instance, if the input light beam to the waveguide is 4 mm in diameter, without an EPE structure, the output pupil will remain as 4 mm since the waveguide does no modification to the light other than transport it. That means your eye can only see the virtual image clearly with your pupil center moving within this 4 mm range. By implementing the EPE structure, the exit pupil can be expanded to above 10 mm, resulting in a larger motion box of your eye.

The eyebox is very important for the AR glasses to be able to accommodate users with different inter-pupil distances which could range from 51 mm to 77 mm depending on age, gender, etc.

This technique solves a lot of problems in the product design of AR glasses, such as mechanical tolerance, product SKU (e.g. different specifications for male and female), ergonomic and user interface design etc. Thus optical waveguide with EPE has pushed AR glasses a big step forward to consumer-level products. However, there is no free lunch. The expansion of eyebox is at the cost of less light output at each spot after being averaged out. This is the major reason that waveguide has a lower optical efficiency compared with conventional approaches.

The geometric waveguide utilizes conventional geometric optical design process, simulation tools, and manufacturing process, without involving any fancy subwavelength structures. Because the geometric optical structures pose no bias on color, the resulting image can be of very high quality. However, there seem to have challenges in the manufacturing process. One challenge is the coating of the transflective mirror. Because there is less and less light left during the propagation inside waveguide, the reflection/transmission ratio required will be different for each of the mirrors to guarantee a uniform light output within the whole eyebox. And since the light is polarized due to the nature of LCOS which is often used as the microdisplay for geometric waveguide system, each mirror can have more than ten layers of thin film coatings on the surface.

In addition, after the coating process of each mirror, they need to be stacked and glued together, then cut at a precise angle. The precision of gluing and cutting will also influence the parallelism of the glass plates thus the image quality. Although each step is conventional optical manufacturing with possible high yield, achieving a reasonable total yield with all the tedious steps and pieces combined together is rather challenging. The imprecision of any process step might cause imperfections in the final virtual image such as black lines, nonuniformity, ghost image, etc. In addition, although the manufacturing advancement has minimized the visibility of the mirror array when microdisplay is turned off, we can still see them as “stripes” on the eyepiece, which blocks normal vision and also affect the appearance of the AR glasses.

Figure 5. Categories of waveguide technologies: (a) geometric waveguide with reflective mirror array, (b) diffractive waveguide with surface relief gratings, © diffractive waveguide with volumetric holographic gratings. Images modified from source:

We will conclude the geometric waveguide here and continue with diffractive waveguide in Part II of this article. During AWE USA 2019 (Augmented Wold Expo) , Rokid just released its new generation of AR glasses -Rokid Vision(, which adopts diffractive waveguide technology. Does diffractive waveguide outperform geometric type? How do different diffractive gratings compare with each other? Which is the ultimate solution for consumer-level AR glasses? Please stay tuned for our next sharing.


The Author: Kun (Linda) Li

The Author:  Kun (Linda) Li got her PhD degree from Electrical Engineering at University of California Berkeley, and currently works at Rokid as a senior research scientist and project manager. Her research areas include optical imaging system, optical metrology, optoelectronic devices, semiconductor laser, nanotechnology, etc.