Content uploaded by Shin-Tson Wu
Author content
All content in this area was uploaded by Shin-Tson Wu on Nov 17, 2023
Content may be subject to copyright.
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 783
AR/VR light engines: perspectives
and challenges
En-Lin Hsiang,1,†Zhiyong Yang,1,†Qian Yang,1,†
Po-Cheng Lai,2Chih-Lung Lin,2AND Shin-Tson Wu1,∗
1College of Optics and Photonics, University of Central Florida, Orlando, Florida 32816, USA
2Department of Electrical Engineering, National Cheng Kung University, Tainan 701-01, Taiwan
*Corresponding author: swu@creol.ucf.edu
†These authors contributed equally to this work.
Received June 16, 2022; revised September 9, 2022; accepted September 9, 2022;
published 9 November 2022
Augmented reality (AR) and virtual reality (VR) have the potential to revolutionize the
interface between our physical and digital worlds. Recent advances in digital processing,
data transmission, optics, and display technologies offer new opportunities for ubiqui-
tous AR/VR applications. The foundation of this revolution is based on AR/VR display
systems with high image fidelity, compact formfactor, and high optical efficiency. In this
review paper, we start by analyzing the human vision system and the architectures of
AR/VR display systems and then manifest the main requirements for the light engines.
Next, the working principles of six display light engines, namely transmissive liquid
crystal display, reflective liquid-crystal-on-silicon microdisplay, digital light process-
ing microdisplay, micro light-emitting-diode microdisplay, organic light-emitting-diode
microdisplay, and laser beam scanning displays, are introduced. According to the charac-
teristics of these light engines, the perspectives and challenges of each display technology
are analyzed through five performance metrics, namely resolution density, response time,
efficiency/brightness/lifetime, dynamic range, and compactness. Finally, potential solu-
tions to overcoming these challenges are discussed. ©2022 Optica Publishing Group
https://doi.org/10.1364/AOP.468066
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785
2. Human-Centric AR/VR Displays . . . . . . . . . . . . . . . . . . . . . . . 786
2.1. Comfort and Immersion . . . . . . . . . . . . . . . . . . . . . . . . 786
2.2. Architecture of Human Eye . . . . . . . . . . . . . . . . . . . . . . 786
2.3. Eyebox and FoV . . . . . . . . . . . . . . . . . . . . . . . . . . . . 788
2.4. Eye Safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 790
3. AR/VR Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 790
3.1. VR Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . 790
3.2. AR Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . 791
3.2a. Free-Space Combiner . . . . . . . . . . . . . . . . . . . . . . 791
3.2b. Freeform Prism and Waveguide Combiners . . . . . . . . . . 791
784 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
4. Light Engines for AR/VR Displays . . . . . . . . . . . . . . . . . . . . . . 792
4.1. Light Modulation Display . . . . . . . . . . . . . . . . . . . . . . . 792
4.1a. LCD...............................793
4.1b. LCOS..............................794
4.1c. DLP...............................795
4.2. Self-Emissive Displays . . . . . . . . . . . . . . . . . . . . . . . . . 796
4.2a. OLED..............................796
4.2b. µLED ..............................798
4.3. Light Scanning Displays . . . . . . . . . . . . . . . . . . . . . . . . 798
5. Display Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 799
5.1. Resolution Density . . . . . . . . . . . . . . . . . . . . . . . . . . . 799
5.1a. LCD...............................799
5.1b. OLED..............................802
5.1c. µLED ..............................804
5.1d. LCOS..............................807
5.1e. DLP...............................810
5.1f. LBS...............................810
5.2. Fast Response Time . . . . . . . . . . . . . . . . . . . . . . . . . . 811
5.2a. LCD...............................812
5.2b. LCOS..............................813
5.2c. DLP...............................814
5.2d. LBS...............................816
5.2e. µLED and OLED . . . . . . . . . . . . . . . . . . . . . . . . 816
5.3. Efficiency................................818
5.3a. LCD...............................820
5.3b. OLED..............................822
5.3c. µLED ..............................826
5.3d. LCOS..............................827
5.3e. DMD ..............................830
5.3f. LBS...............................831
5.4. High Dynamic Range . . . . . . . . . . . . . . . . . . . . . . . . . 832
5.4a. Dual Modulation Display . . . . . . . . . . . . . . . . . . . . 832
5.4b. HDR Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . 833
5.4c. HDR LCOS . . . . . . . . . . . . . . . . . . . . . . . . . . . 834
5.4d. HDR DLP . . . . . . . . . . . . . . . . . . . . . . . . . . . . 835
5.4e. HDR LBS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 837
5.4f. HDR OLED and µLED.....................838
5.5. Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 838
6. Discussion and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 842
Funding ......................................844
Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 844
Disclosures ....................................844
Data availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 844
References.....................................844
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 785
AR/VR light engines: perspectives
and challenges
En-Lin Hsiang, Zhiyong Yang, Qian Yang, Po-Cheng
Lai, Chih-Lung Lin, AND Shin-Tson Wu
1. INTRODUCTION
From televisions and monitors to smartphones and tablets, flat-panel displays have per-
meated our daily lives. These devices act as an interface between humans and machines
to display digital information. From the oldest cathode-ray tube (CRT) displays, liquid
crystal displays (LCDs) [1], organic light-emitting-diode (OLED) displays [2], to the
latest micro light-emitting-diode (µLED) displays [3], the prosperous development in
flat panel displays is primarily focused on high resolution, high dynamic range (HDR),
vivid colors, wide viewing angle, fast motion picture response time (MPRT), thin pro-
file, lower power consumption, and low cost [4–6]. Despite the tremendous progresses
made in flat panel displays, there is still an ultimate desire for interactive displays to
exhibit vivid three-dimensional (3D) visual experiences [7,8]. To achieve this goal,
different types of augmented reality (AR) and virtual reality (VR) headsets have been
proposed since the 1990s. However, the first wave of AR/VR headsets did not last too
long due to insufficient support for high-speed communication (5G), small-sized dis-
plays, computing platforms, and battery capacity. Nowadays, all the above-mentioned
bottlenecks have been gradually overcome, so another wave of AR/VR displays is once
again receiving enthusiastic attention. Due to the broad interests and rapid development
of AR/VR headsets, several articles reviewing recent advances from the perspectives of
microdisplays, optical system integration including liquid crystal planar optics, holo-
graphic optics, freeform optics, meta-structure, and potential applications have been
published [9–12]. Some of these reports emphasize how to generate a comfortable 3D
visual experience by multiplane displays, integral imaging displays, and holographic
displays, but very few analyze advanced light engines for AR/VR headsets. To remedy
this deficiency, in this review paper, we focus on advanced light engines for AR/VR
applications. It is worth noting that, depending on the applications and functionalities,
the display characteristics of the optical engines in AR/VR devices are quite different
from those used in direct-view flat panels. For example, in direct-view displays, such
as TVs and desktop computers, Lambertian radiation pattern is favored to enable a
wide viewing angle for multiple viewers, but in the near-eye projection systems, wide
emission cones can cause image smudges and lower optical efficiency. In addition,
the magnified images in AR/VR projection systems demand a much higher resolution
density than traditional direct-view displays.
In the following sections, we first review the basic structure of the human visual system
(HVS) and the architecture of AR/VR displays, and then manifest the requirements of
advanced light engines to support AR/VR displays with a vivid 3D visual experience.
After that, we dive into the working principles of six types of light engines: two of
them are self-emissive types, OLED microdisplays and µLED microdisplays; three
of them are light modulation types, transmissive LCDs, reflective liquid-crystal-on-
silicon (LCOS) microdisplays, and digital light processing (DLP) microdisplays; and
the last one is a scanning type, laser beam scanning (LBS) display. Next, five display
performance metrics, resolution density, response time, efficiency/brightness/lifetime,
HDR, and compactness, are proposed to analyze the advantages and challenges of each
786 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
light engine technology. During the analysis, advanced display configurations for each
light engine to enhance its performance are also discussed.
2. HUMAN-CENTRIC AR/VR DISPLAYS
2.1. Comfort and Immersion
To achieve a spectacular viewing experience, AR/VR devices should offer both comfort
and immersion [13,14]. Although there are some AR/VR applications realized with
tabletop devices such as transparent displays [15], in this paper, we focus on the near-
eye AR/VR devices. Comfort determines whether a user can wear the headset for
an all-day use, whereas immersion requires that the display turns the virtuality into
reality. The comfort comes with wearability [16], visual experience [17], and social
interaction [18]. The immersion relates to all kinds of human senses, especially in
aural, visual, and haptic senses [19]. The optics mainly determines the visual comfort
and immersion [20], but it also plays an essential role for wearable and social comfort.
For example, an AR/VR headset must have a small formfactor and be lightweight to
widen acceptance by consumers [21]. The weight of a headset should be as evenly
distributed as possible and, therefore, the center of gravity should be close to the
head, enabling comfortable wearing for long-time use. As optics occupies a larger
volume in a near-eye headset, both light engines and optical combiners should be
lightweight and compact. Therefore, the first requirement for an advanced light engine
is compactness. For an optical see-through AR device, it is expected that the user’s
eyes can be clearly seen by others, allowing for true eye contact for social interaction.
Visual comfort and immersion are not easily measurable objective metrics. They
are subjective experiences and vary from person to person [22]. Thus, the design of
AR/VR devices should be a human-centric task. The HVS has its unique capabilities
and limitations. A detailed understanding of the HVS helps designers to make sensible
trade-offs in optical specifications or even reduce the system complexity by taking
advantage of the HVS.
2.2. Architecture of Human Eye
The HVS consists of two eyes and the interpupillary distance (IPD) is the distance
between the center of two eyes (Fig. 1(a1)), usually expressed in millimeters. This
value (49–76 mm) varies by individual [23], depending on age, gender, and ethnicity.
An eye is essentially an imaging system including multiple refractive surfaces and an
adjustable iris [24]. A simplified eye model is illustrated in Fig. 1(a2). The incident light
passes through cornea and aqueous humor and then enters the pupil, a round opening
in the center of the iris. The iris adjusts the effective Fnumber of the imaging system
by changing the size of the pupil, limiting the light throughput. As light continues, it
passes through the lens. The lens is attached to muscles which can change the shape
and optical power of the lens by contracting or relaxing. This feature enables humans
to accommodate for an in-focus image at different depths. The last refraction occurs
on the interfaces between the lens and vitreous and finally an image is formed on
the retina. From there, photoreceptors including cone cells and rod cells convert light
intensity into electrical impulses, which are unevenly distributed on the retina. Cone
cells concentrate most in the central region of the retina called the macula, which spans
about 5 mm. Visual acuity refers to the ability for a human eye to resolve small features.
The fovea, in the center of macula, has the maximum visual acuity due to its highest
photoreceptor density, covering only 2–3°[13]. For a 20/20 vision, an eye should be
able to resolve detail as small as 1 arcmin, or, in other words, the angular resolution
is 60 pixels per degree (PPD) in the fovea. Outside the fovea, visual acuity declines
rapidly in the rest region of macula. Rod cells can be found away from the macula,
responsible for scotopic vision with low-resolution perception. In AR/VR systems,
both the imaging quality of optics and the resolution density of light engines affect the
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 787
Figure 1
(a1) Illustration of interpupillary distance (IPD) in humans. (a2) Schematic of a sim-
plified human eye model. (b) Effect of excellent (red) and poor (blue) MTF to display
quality. (c) Displayed image with screen door effect. (d) Vergence-accommodation
tolerances at a near (0.5 m) and far (2.5 m) distance. (e) Displayed image with motion
blur.
final image quality on the retina. The modulation transfer function (MTF) [25] is an
indicator of how well an imaging system can reproduce fine details and sharp edges
as shown in Fig. 1(b). MTF shows the contrast performance as a function of angular
or spatial frequency, and it decides the spatial frequencies which the display and
following optics can deliver to an eye. The perceived PPD may be lower than expected
if the optics between the display panel and eyes shows a poor MTF. Failing to satisfy
such a high PPD leads to the screen door effect [26], where a mesh pattern is overlaid
over the image, like seeing the world through a screen door (Fig. 1(c)). The “screen
door” is essentially the pixel structure of a display panel as the fill factor is not 100%
and only part of pixel is emitting, transmitting, or reflecting light. Here, we present
the second requirement for an advanced display light engine: high resolution density.
The image perceived by human eyes is not a flat 2D image but with 3D sense.
Depth cues exist in the HVS [27]. Physiologically speaking, the depth cues include
accommodation, convergence, and motion parallax [28]. On the one hand, the lens in
an eye can dynamically accommodate its shape as well as optical power to form clear
images at different depths. On the other hand, with binocular vision, each eyeball will
rotate to converge its line of sight on the focused object. Thus, this convergence angle
provides depth information to the visual cortex. These two depth cues, vergence and
accommodation, are closely linked within the HVS and are intrinsically matched with
each other in reality. The ability for the human eye to perceive depth variation is called
stereo acuity. The HVS is not sensitive to the depth change caused by absolute distance
difference, but by diopter difference [29,30]. This feature indicates that the HVS is
more sensitive to the depth variation at a close range, usually referred to as “one-arm’s
788 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
length,” which is about 30–50 cm. For objects at further range, stereo acuity is not
sensitive. When looking at a fixed focus stereo display, a user is forced to accommodate
to a single distance to obtain a clear image, but stereoscopic disparity tells eyes to
make a vergence for objects at different depths, introducing vergence–accommodation
conflict (VAC) [31,32]. The VAC tolerance range for a near and far distance is illustrated
in Fig. 1(d). If the image plane is set at 2.5 m away from a user, then the acceptance
region covers about 5 m, from 1.5 m to 6.7 m, if the VAC tolerance limit is set to
0.25 diopters. For a higher VAC tolerance limit, say 0.4 diopters, then for 1.25 m
to infinity, there is no VAC issue. However, for a closer imaging plane at 0.5 m, the
acceptance region will be much narrower due to drastic diopter variation. Integral
imaging displays [33,34] and holographic displays [14,35] are proposed to reproduce
the light field by ray approximation or a complete wavefront. Maxwellian displays
[21,36] avoid VAC by creating a large depth of focus image and may introduce natural
blur through image rendering. In light engines with a high frame rate, accompanying
an active combiner, more focal planes [37] can be created statically or dynamically
at near depth to mitigate the VAC issue. In addition, motion artifacts [38] in nature
usually happens when an observed object or an observer’s eye is moving too fast,
resulting in an inability to resolve details, as depicted in Fig. 1(e). A nature-looking
movement requires some degree of motion blur, but if the response time [39] of the
hold-type display is too high [40], extra undesired motion artifacts may be caught
by the users. Motion artifacts may also occur when the head moves too fast, and the
displayed content is not updated synchronously. This is usually described as motion-
to-photon latency [41,42]. Fast sensors and a better video-processing pipeline could
also diminish this nausea. Based on this discussion, we present the third requirement
for an advanced display light engine: fast response time.
The HVS has a huge dynamic range, from the dim starlight at 10−6nits to bright
sunlight at 108nits, as shown in Fig. 2(a). In dark light environments, or at scotopic
light levels, it is mainly the rod cells that are responsive for luminance ranging from
10−6to 10−2nits, whereas cone cells are active for photopic light levels (10–108nits).
Between these two ranges (mesopic range), both rod and cones are involved in the
sensing. The difference in scale between the darkest and brightest objects a human eye
can perceive spans 14 orders of magnitude. At one time, the HVS can only perceive
a subset of such range, say about five orders of magnitude. The traditional standard
dynamic range (SDR) displays cover 3 orders, but HDR [43,44] displays aim to match
the steady dynamic range of the human eye, allowing objects to be represented closer
to their nature. Human eyes are more sensitive to the illuminance variation at low
light levels, so high contrast ratio, lower dark level, and more bits at low gray levels
are preferable in HDR displays. For AR applications, the highest illuminance of a
HDR display is vital for outdoor scenarios. The ambient contrast ratio (ACR) [45]
should achieve at least (3:1) for an acceptable readability. Considering the ambient
illuminance is 3000 nits in a sunny day, the display needs to deliver at least 10,000 nits,
ignoring the optical loss. If methods such as exit pupil expansion (EPE) are applied
to enlarge the eyebox, an AR device will demand a much higher brightness from the
display panel. Due to limited battery capacity, the optical efficiency of the overall
system should be high. Given that ambient light is fully blocked in VR headsets,
150–200 nits of brightness received by human eye is acceptable after considering the
optical losses. Here, we present the fourth and fifth requirements for an advanced
display light engine: HDR, high efficiency and peak brightness, and long lifetime.
2.3. Eyebox and FoV
In HVS, an entire field of view (FoV) spans more than 200°horizontally (H) and
about 130°vertically (V), whereas the binocular overlap is about 120°H, as shown
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 789
Figure 2
(a) Dynamic range of the human visual system. (b1) Horizontal human vision FoV.
(b2) Vertical human vision FoV. (c1) Schematics of FoV calculations in Fresnel and
pancake VR systems at a given display panel width, eye relief, and eyebox, assuming
the optical power is provided by a singlet lens. (c2) Relation between FoV and panel
width in Fresnel- and pancake-lens-based VR systems.
in Fig. 2(b1). When eyes are at a relaxing state, steady gaze is possible for a FoV of
±20°H and +15°/–20°V, without producing any eye strain. Compared with horizontal
FoV, the vertical FoV is asymmetric because the relaxed line of sight is 15°below the
horizontal line of sight, as depicted in Fig. 2(b2). Figures 2(b1) and 2(b2) also show
capabilities of each part of FoV in the HVS. The eyebox [46] is a physical 3D region
where the whole FoV can be viewed without vignetting. In an AR/VR device, this
3D volume is closely related to the exit pupil size of the optical combiner. Although
the eyebox is defined in terms of volume, it is usually expressed as a few millimeters
horizontally. A larger eyebox would allow an AR/VR device to be set up faster and
better account for variations in human IPD and positioning of the head. Eye relief [47]
refers to the distance from the last optical surface to the best viewing spot. A shorter
eye relief generally increases the perceived eyebox size but may prohibit users from
wearing glasses if needed. In addition, due to the conservation of Lagrange invariant
and étendue in an imaging system, a trade-off exists between the eyebox and FoV. In
AR devices, pupil replication [36,48] and pupil steering [49] techniques are proposed
to mitigate the limited eyebox. Moreover, in VR devices, a sufficiently large panel size
is required to support the required FoV with acceptable eyebox. Assuming the eyebox
size is 8 mm and eye relief is 15 mm (Fig. 2(c1)), Fig. 2(c2) illustrates the relation
790 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
between FoV and panel size in a Fresnel lens or folded-optics-based (often called a
pancake lens; see Sec. 3.1) VR system, under the simplification that optical power is
provided by an ideal singlet lens. In Fig. 2(c2), a larger panel size generally leads to a
wider FoV. Here, we label the panels with diagonal dimensions of 1, 2, and 3 inches
for reference. Furthermore, at a given panel width of 40 mm (vertical dashed lines),
the VR system (pancake architecture) with a larger optical power exhibits a wider
FoV (100°) than the VR system (Fresnel architecture) with a lower optical power
(FoV =65°).
2.4. Eye Safety
Cone cells on the retina have different spectral sensitivities and are generally labeled by
their peak wavelengths as short (S), medium (M), and long (L) cone types. According
to the trichromatic theory, the same color can be perceived by a human eye even if
the input light spectrum is different, as long as the tristimulus values remain the same.
This gives some freedom on the choice of the light sources in AR/VR displays to
reproduce the desired colors. However, the designer must take good care of spectrum
brightness perceived by human eyes, following the eye safety regulation. For example,
at the same brightness, deep blue is more harmful to eyes than light blue. Nowadays,
mainstream light sources are LED (both inorganic and organic) and lasers. These
high-brightness emissive light engines could be harmful to the user’s eyes for two
possible reasons: (1) they may contain some ultraviolet (UV) or deep blue component,
such as deep-blue-pumped quantum dot (QD) color conversion layer; and (2) the light
source could be very strong to keep a reasonably high ACR for outdoor applications.
3. AR/VR ARCHITECTURES
3.1. VR Architecture
Most current VR devices adopt a simple optical architecture combining a light engine
and a collimation lens with positive optical power, as Fig. 3(a) depicts. Usually, some
optical power is provided by a Fresnel lens and therefore the collimating lens is a
hybrid Fresnel refractive lens. In addition, a hybrid diffractive and refractive lens can
also be applied to reduce lateral chromatic aberration [50]. To further reduce form
factor, the optical path can be folded by introducing polarization sensitive optical
elements, such as a reflective polarizer (RP) or a cholesteric liquid crystal (CLC) lens.
This optical design is usually referred to as “pancake” optics [51]. A typical pancake
optics design is plotted in Fig. 3(b).
Figure 3
(a) Sketch of the optical architecture of a simplest VR system including a light engine
and a hybrid Fresnel refractive lens. (b) Sketch of a pancake VR system consisting of
a light engine, a half mirror (HM), a refractive lens, a QWP, and a reflective polarizer
(RP). (c) Sketch of an ultra-slim pancake VR system consisting of a holographic lens
and a CLC lens.
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 791
The system is composed of five parts: a light engine, a half mirror (HM), a refrac-
tive/Fresnel/hybrid lens, a quarter-wave plate (QWP), and a RP. Assuming the input
light polarization is right-handed circular polarization (RCP), after passing through
the QWP and RP, the polarization state remains unchanged. Upon reflection on the
HM, the reflected light is switched to left-handed circular polarization (LCP) and
passes through the QWP and RP. The optical path between the HM and RP is tripled,
but the maximum efficiency of the optical combiner is only 25% in theory. In an even
more compact pancake VR design (Fig. 3(c)), the HM and lens could be replaced by
a volume holographic lens [52] and the QWP and RP can be replaced by a CLC lens
[53]. Although all kinds of light engines can work in the VR systems, active-matrix
LCDs and OLED displays are currently the mainstream choices.
3.2. AR Architecture
AR architectures are often categorized by their combiners. Based on where the incident
light propagates, there are three types of optical combiners: free-space combiners,
freeform prism combiners, and waveguide combiners.
3.2a. Free-Space Combiner
The simplest free-space combiner is a 50/50 beam splitter. The output light from an
optical engine is reflected to the eye while the ambient light from the real world passes
through the beam splitter and is combined with the display light, as shown in Fig. 4(a).
Some optical power can be added to the partial reflector by making it a curved surface
[54] (Fig. 4(b)), enabling a large FoV, but this design suffers from image distortion
because all the optical power is provided by a single surface and the form factor is large.
The birdbath optics [55] in Fig. 4(c) folds the optical path for a smaller form factor and
introduces additional optical elements for aberration correction. These architectures
serve as traditional imaging systems, imaging from a “real” plane of a light engine to
a “virtual” plane. Then this “virtual” plane is imaged by an eye to the retina. Thus,
display panels such as OLED, µLED, DLP, and LCOS are preferred. In a Maxwellian
display [21,36,48,49,56], the combiner reflects the image from the light engine to the
eye pupil, as shown in Fig. 4(d). As the focusing spot is much smaller than the eye pupil,
a clear image with infinite depth of focus can be formed on the retina, no matter what
the optical power of the eye is. However, the eyebox is limited in Maxwellian displays,
thus pupil steering or pupil duplication is often required to enlarge the eyebox. A LBS
display [57] is a natural point source and, thus, can be applied directly to Maxwellian
displays. A spatial light modulating display such as LCOS or DLP [35] can also be
used in Maxwellian displays if illuminated by a collimated light, as Fig. 4(e) shows.
Although self-emissive display panels, such as µLED and µOLED, relayed by a 4-f
system with an aperture stop can also be employed in Maxwellian displays, their light
throughput is relatively low.
3.2b. Freeform Prism and Waveguide Combiners
In both freeform prism combiners [58] and waveguide combiners [45], the imaging
light propagates in either prism or waveguide by total internal reflection (TIR), as
depicted in Figs. 4(f) and 4(g), respectively. The optical path is folded in a prism com-
biner and each surface is carefully designed for achieving an excellent image quality.
The most obvious feature in a waveguide combiner is the EPE process, which breaks
the étendue limit and effectively increases the eyebox size. The upper limit of the FoV
is determined by the waveguide refractive index, which is about 70°for a n=2 glass.
Usually, for a display panel, the output image is first Fourier transformed to far field,
converting image information from the spatial to the angular domain. However, the
LBS display does not require such a conversion. Then, display light is coupled into a
792 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Figure 4
(a) A 50/50 beam splitter as a combiner. (b) Single reflective curved surface combiner.
(c) Birdbath design combiner with a folded optical path. (d) Maxwellian display based
on a LBS. (e) Maxwellian display based on a SLM. (f) Freeform prism combiner.
(g) Waveguide combiner consisting of two reflective-type diffractive gratings. HOE,
holographic optical element; SLM, spatial light modulator.
waveguide, propagates through the TIR process, and finally is outcoupled into human
eyes. The in- and outcoupler can be a prism, a diffractive grating, or partial reflective
mirrors.
4. LIGHT ENGINES FOR AR/VR DISPLAYS
4.1. Light Modulation Display
Here, a light modulating display refers to a display that utilizes mechanisms to mod-
ulate the polarization, propagation direction, or phase of the imaging light. The
modulation of polarization or propagation direction can be converted into ampli-
tude modulation, and the phase modulation can be used for holographic displays.
Presently, commercially available light modulating displays mainly include LCDs,
LCOS microdisplays, and DLP microdisplays. Recently, transmissive thin-film tran-
sistor (TFT) LCDs with resolution density of 2016 pixels per inch (PPI) have been
demonstrated by Innolux (in Touch Taiwan 2022) for VR applications. On the other
hand, the reflective LCOS and digital micromirror device (DMD) driven by silicon
backplane can achieve a higher resolution density (>6000 PPI) and smaller form factor
for AR applications.
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 793
Figure 5
Schematics of (a) LCD and (b) LCOS: (c1) FFS mode, (c2) VA mode, (c3) MTN
mode, and (c4) HG mode.
For a light modulation display, the light source also plays a significant role affecting
the final performances of light engines. LEDs and lasers are two widely used light
sources. For LCDs, the LEDs used in the backlight module can be further categorized
into conventional large-size LEDs and mini-LEDs. Mini-LEDs can provide a much
higher dynamic range than conventional LEDs through local dimming effect, but the
halo effect should be minimized. The lasers used in backlight modules can widen the
color gamut and theoretically save about 70% of the optical losses from color filters
and polarizers. However, in comparison with LED backlights, laser sources generally
require a larger space to homogenize the brightness distribution. In addition, during the
homogenization process, the polarization property of the lasers may be deteriorated.
As a result, the optical efficiency improvement is often lower than the theoretical value.
Furthermore, from the perspective of light source efficiency, the wall plug efficiency
of the lasers is lower than that of LEDs. For the light modulated AR displays (LCOS
and DMD), due to the smaller panel size, using a laser as light engine may not suffer
from the severe homogenization issues that LCDs do. Moreover, compared with LEDs,
almost all the light from a laser with a small etendue can impinge on the panel, resulting
in a much smaller optical loss. However, the speckle issue needs to be addressed by
methods such as electromechanical de-speckling and AC driving.
4.1a. LCD
Liquid-crystal-based displays consist of transmissive LCDs (Fig. 5(a)) and LCOS
microdisplay (Fig. 5(b)). An active-matrix LCD is composed of a backlight module, a
TFT array, a LC layer, color filters, and crossed polarizers, as schematized in Fig. 5(a).
The backlight module provides a uniform luminance distribution, and the pixelated
LC cell, modulated by the TFT array, determines the transmittance of each subpixel
for gray-scale images [59]. In LCD panels, two commonly used operation modes have
been developed, depending on the molecular alignment and the electric field direction:
794 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
(1) fringe-field switching (FFS) mode (Fig. 5(c1)) [60] and (2) vertical alignment (VA)
mode (Fig. 5(c2)) [61]. Each mode has its own pros and cons and has been discussed
extensively in several review papers [6,62,63]. Thus, we only briefly introduce their
device configurations and operation principles, and in the following section, we focus
on the recently developed LC modes, which meet the requirements of AR/VR display
applications, such as fast response time, high transmittance, and acceptable viewing
angle. For the FFS mode, the LC directors are homogenously aligned, and the lateral
electric field reorients the LC directors mostly in-plane. According to the in-plane
switching, the FFS LCDs feature a small angular color shift and contrast roll-off
under oblique viewing angles. Moreover, the built-in storage capacitor formed in
FFS mode is more suitable for high PPI display (VR display) because of its larger
aperture ratio. However, the slow response time caused by the small twist elastic
constant of traditional FFS mode should be overcome to suppress the motion image
blurs for VR applications. For VA mode, the LC directors are initially aligned in
vertical direction, so the incident linearly polarized light is not modulated when
traversing through the LC cell and is blocked by the crossed analyzer. As the voltage
exceeds a threshold, the negative dielectric anisotropy LC molecules bend against
the longitudinal electric field, allowing the incident light to pass through the crossed
polarizers. The intrinsic high contrast ratio of VA mode is preferred for providing
HDR images. However, VA mode is plagued by a limited viewing angle, so the
multi-domain structures, such as 4, 8, and even 12 domains, have been proposed
to expand its viewing angle [64]. However, for VR applications, the viewing angle
requirement is much more relaxed than the traditional direct-view displays, such as
TVs. Therefore, for VR applications, the multi-domain structure of VA LCD has
more degrees of freedom to balance the viewing angle, transmittance, and response
time. It is worth mentioning that, due to the scattering from the TFT array, the LC
layer, and the color filter, the contrast ratio of LCDs is limited by around 2000:1
and 5000:1 for the FFS mode and VA mode, respectively. Therefore, the grayish
dark state of the LCDs may severely disrupt the immersive experience of the users.
Fortunately, mini-LED backlight can help the LCD achieve a much higher dynamic
range via local dimming. As a result, the mini-LED backlit LCD is a good candidate for
advanced VR headsets due to its higher contrast ratio, higher bit depth, and lower power
consumption.
4.1b. LCOS
Compared with transmissive TFT-LCDs, LCOS adopts silicon backplane to achieve
a higher resolution density and a reflective operation mode to increase the fill factor
and frame rate. The first attempt to drive LCs using a silicon backplane was made
by Ernstoff et al. in the early 1970s [65]. Dynamic scattering electro-optic effect was
initially used but not appropriate for phase modulation due to the charge leakage. Later
the LCOS devices based on the field effect of nematic LCs and ferroelectric LCs were
reported [66], respectively. A typical LCOS device consists of a CMOS backplane,
metal pixel electrodes, a LC layer, an indium tin oxide (ITO) common electrode, and
a cover glass, as shown in Fig. 5(b). A set of red, green, and blue (RGB) LEDs is
employed as a light source to illuminate the LC panel uniformly and the full color
image is generated via field sequential colors (FSCs). To obtain coaxial RGB beams
from the separated RGB light sources (LEDs or lasers), dichroic mirrors are usu-
ally employed. When the incident light travels through the LC layer, the intersection
between the transverse plane and the refractive index ellipsoid yields two eigenwaves
with two eigenpolarizations and normal indices. One eigenwave experiences an angle-
dependent refractive index (ne), and the other experiences a constant refractive index
(no). Owing to the birefringence (ne−no), effect, the incident polarization is decom-
posed into two eigenpolarizations, and the accumulated phase retardation between the
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 795
two eigenpolarizations can modulate the incident polarization. Polarization modula-
tion can be converted into amplitude modulation through a polarizing beam splitter
(PBS) or a circular polarizer. However, if the incident polarization is parallel to the
eigenpolarization with a variable refractive index, the incident polarization remains
unchanged, and the pixelated phase change depending on an applied voltage can be
obtained. In this way, pure phase modulation is achieved.
To date, three major LC modes have been widely adopted in LCOS devices: VA
mode [61] (Fig. 5(c2)) and mixed-mode twisted nematic (MTN) [67] (Fig. 5(c3)) are
mainly used in amplitude modulation, and homogenous (HG) alignment (Fig. 5(c4))
is dominantly employed in phase modulation. For the VA mode, without any applied
voltage (V =0), an excellent dark state and high contrast ratio can be obtained, which
is a major advantage of the VA mode. In the MTN mode, the LCs are twisted from
the bottom substrate to the top substrate with 90°twist angle. The angle between the
polarizer’s optic axis and the front LC directors is defined as β, which is set at 20°
to maximize the reflectance of the 90°-MTN cell. MTN exhibits a faster response
time, wider viewing cone, and weaker fringing field effect (FFE) than VA, but its
contrast ratio (∼1000:1) is lower. Homogeneous alignment is not suitable for amplitude
modulation due to its strong wavelength dependency and narrow viewing angle, but it
is widely used in phase modulation because of its fast response time.
4.1c. DLP
A DMD is a microelectromechanical system (MEMS) device [68,69] developed by
Texas Instruments in 1987. The DMD originated from the deformable mirror device
in 1977, and then was commercialized in DLP in 1996. At the heart of the DMD
technology are a dense array of micromirrors that can be electrostatically rotated to
an on or off state. Each DMD pixel consists of the micromirror, yoke and hinge,
metal standing, and dual CMOS memory. Figure 6(a) illustrates that the micromirror
is attached by a via to a yoke with a torsion hinge [70]. The two electrodes steer the
micromirror to the two operational positions (Fig. 6(b) [71]) by electrostatic forces.
One electrode addresses directly the micromirror, and another controls the yoke. Dual
CMOS memory is used to store the binary state of the micromirror, and a mirror
clocking pulse will transfer the stored memory state to the mechanical position of
the micromirror. Due to the binary module, gray levels are usually generated by
pulse width modulation (PWM). To create color images, FSC is adopted. In contrast
to LCOS devices, high-speed response and polarization insensitivity are two main
characteristics for DMDs.
In the past two or three decades, the DMD technology has advanced in various
aspects including the mirror tilt angle, mirror pitch, optical/system efficiency, and
contrast ratio. The tilt angle has been continuously improved from initial±10°,±12°,
to current ±17°. Due to enlarged tilt angles, the illumination cone angle at the mirror
and the étendue of the illumination system are increased, contributing to a higher
system efficiency. The mirror pitch has been reduced from the initial 17 µm to current
5.4 µm realized in the Texas Instruments DLP3010, enabling a high-resolution DMD.
The optical efficiency of the DMD depends on several factors including the window
transmission, fill factor, diffraction effect, and mirror reflectance. To calculate the
system-level efficiency, we should consider more factors such as étendue mismatches
between the illumination and projection system, which are described in more detail
in Sec. 5.2. As for contrast ratio, it cannot be described for the DMD itself, which
is different from LCOS. The system contrast ratio can be defined once the projection
lens or system aperture stop constraining a solid angle is considered. The dominant
factors affecting the system contrast ratio consist of illumination angle, mirror gap,
numerical aperture, and coating quality [72]. Illumination angle and surface coating
796 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Figure 6
Schematic of (a) a single DMD pixel and (b) two DMD pixels with opposite statuses
and one micromirror at the parked state. (c) Diffraction pattern from the DMD at
normal incidence with all mirrors in the on state. (a) Reprinted with permission from
[70]. ©Texas Instruments. (b) and (c) Reprinted with permission from Scholes et al.,
Opt. Eng. 59, 041202 (2019) [71]. ©2019 SPIE.
can affect the distribution of reflected light and scattered light, and mirror gap can
lead to diffracted light. Numerical aperture determines how much of the reflected,
scattered, or diffracted light is collected by a projection lens and thereby plays a
significant role in influencing system contrast ratio as well. Figure 6(c) shows the
far-field diffraction pattern from the DMD at normal incidence when all mirrors are in
the “on state” [71]. The rectangle marked (0,0) represents the zeroth-order diffraction,
whereas the remaining rectangles represent other diffraction orders that may cause a
loss of efficiency and a reduction of system contrast ratio.
4.2. Self-Emissive Displays
4.2a. OLED
The basic structure of a low-voltage OLED device was invented in 1987 by Tang and
VanSlyke [2]. The electrons and holes are injected to the organic emission material
for light emission. However, at the initial stage, the OLED efficiency is limited by
the low quantum efficiency, carrier injection efficiency, and light extraction efficiency
(LEE). In the past few decades, different types of emission mechanism (fluorescent,
phosphorescence, triplet–triplet fluorescent, and thermally active delayed fluorescent)
have been developed to significantly boost the quantum efficiency [73–76]. In addition,
by applying doping techniques in the hole and electron injection layers, the carrier
injection efficiency is enhanced to almost 100%. Nowadays, the OLED efficiency is
mainly limited by the low extraction efficiency resulting from the TIR at the interfaces
between organic material, glass, and air [77,78]. From the perspective of display
configurations, an OLED display consists of two parts: (1) OLED devices (front
plane), which can be patterned red, green, and blue (RGB) OLED devices (Fig. 7(a))
or white OLED (WOLED) with RGB color filters (Fig. 7(b)); (2) electrical driving
(backplane) which can be TFTs fabricated on the glass substrate or the complementary
metal oxide semiconductor (CMOS) on the silicon backplane. According to two types
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 797
Figure 7
Schematics of (a) RGB OLED, (b) tandem WOLED, (c) RGB µLED, and (d) color
conversion µLED display. (c) Reprinted from Lee et al., J. Soc. Inf. Disp. 29, 360–369
(2021) [79]. Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with
permission. (d) Reprinted from [80] under a Creative Commons license.
of front planes and two types of back planes, there are four combinations for achieving
a full-color OLED display, but each has its own pros and cons.
For an OLED display, the brightness of each pixel is determined by the driving current.
Therefore, the complicated driving circuits, which consist of eight transistors, are
proposed to drive the OLED device to a precise gray level. Compared with TFT, CMOS
with a higher carrier mobility and narrower linewidth can minimize the occupied
layout area of the driving circuit and support a higher resolution density. However,
due to the fabrication process and cost, the panel size of OLED-in-silicon is normally
limited to about 1.5 inches [81]. In general, the patterned RGB OLED devices are
fabricated by material depositions through a fine metal mask (FMM) with resolution
density below 600 PPI, which is inadequate for near-eye headsets [82]. Therefore, an
advanced patterning method for the organic material is required. In contrast, for white
OLED with RGB color filters, the pixel pitch is defined by the resolution density of
color filters, which can be fabricated with high resolution density by conventional
photolithography methods [83]. However, the color filters absorb about two-thirds
of the emission light from the white OLED, resulting in a lower optical efficiency.
In addition, to generate white light, the tandem WOLED with two or more stacked
emission layers connected in series with charge generating units is widely applied.
To design an advanced WOLED device, it is important to balance the driving current
density, current efficiency, and peak brightness between RGB subpixels because the
weakest subpixels restrict the overall display performance [84]. For example, in the
798 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
two-stacked (blue florescent and yellow phosphorescent) WOLED, the blue OLED
[85] has a lower current efficiency that limits the peak brightness and lifetime of the
WOLED displays. As a result, tandem WOLEDs with more than two stacks are also
proposed to balance the performance between RGB subpixels, but at the expense of
higher driving voltage and higher fabrication cost.
4.2b. µLED
The GaAsP-based red LED was first invented by Holonyak and Bevacqua in 1962
[86]. Afterwards, different color LEDs were subsequently introduced but the commer-
cialization of blue LEDs faltered for decades until the 1990s, when high-power blue
LEDs were developed by Akasaki, Amano, and Nakamura (see Ref. [87]). After that,
high-power and high-efficiency LEDs have significantly replaced traditional lamps as
illuminating systems for light-modulating display systems, such as LCDs, and pro-
jection displays. Millimeter-scale LED chip sizes are used to provide high-brightness
emission, and the emitted light is further modulated through high-resolution density
LC panels or DMD to display vivid images [88]. In the early 2000s, the LED chip size as
small as 10 µm has been demonstrated [89]. Therefore, high-resolution-density LED
arrays can directly generate decent images without the need of an additional light-
modulating element. Like an OLED display, the active-matrix µLED also requires
front and back planes. For the front plane, it could be an array of RGB µLED chips
(Fig. 7(c)) [79] or an array of blue µLEDs with a patterned QD color conversion layer
(Fig. 7(d)) [80]. However, unlike OLEDs which can be fabricated directly on a glass
substrate, the µLED chips are first fabricated on a specific substrate, such as sapphire
for green and blue LEDs and GaAs for red LEDs, and then transferred to a back plane.
Because RGB LED chips are fabricated on different substrates, they need to be picked
and placed onto the designated positions of a glass substrate. For a 4K2K display,
there are 24M subpixels, i.e., 24M µLED chips need to be transferred. The fabrica-
tion yield and defects repair jointly affect the cost. Recently, various pick-and-place
techniques based on elastomer stamping, electrostatic/electromagnetic transfer, laser-
assisted transfer, or fluid self-assembly are proposed to improve the manufacturing
time and yield [90]. However, restricted by the transfer technologies the resolution
density of these methods is normally below or around 1500 PPI. At 2021 SID Display
Week, PlayNitride demonstrated a full-color RGB µLED prototype with 1411 PPI
by the pick-and-place mass transfer technology. To achieve high-resolution density
µLED arrays, wafer-level transfer approaches have been proposed: flip-chip bonding
and wafer bonding [91], which can work at subpixel pitches less than 3 µm. However,
for the monolithic integration methods, the display panels only support a single color.
Therefore, to generate full colors, the color conversion layer is widely used on top of
blue LED array or using an X-prism [92] to combine the images from three single-
color R/G/B µLED panels. At 2022 SID Display Week, PlayNitride demonstrated a
full-color µLED prototype (panel size 0.49-inch, resolution 1920 ×1080, pixel size
2.5 µm, and pixel pitch 5.6µm) with 4536 PPI using subpixel rendering arrangement
and QD color conversion layer. At 2020, Jade Bird Display demonstrated a full-color
µLED prototype (panel size 0.31-inch, resolution 1280 ×720, and pixel pitch 5 µm)
with 5000 PPI using an X-prism and three single-color R/G/B µLED panels. For the
back plane, similar to the OLED devices, both TFT and CMOS can be employed to
precisely control the driving current level or light emitting time to accurately modulate
the brightness of the µLED chips.
4.3. Light Scanning Displays
So far, we have introduced two kinds of display engines used in AR/VR, which are
panel-based microdisplays composed of a two-dimensional (2D) pixel array. In this
section, LBS microdisplays are introduced. In contrast to panel-based microdisplays,
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 799
LBS exhibits the advantages of small size, high brightness, and high contrast ratio.
Small size arises from the fact that they are not limited by the étendue conservation,
because there is no object plane unlike a panel-based microdisplay. However, the major
drawbacks of LBS microdisplays are low resolution and limited frame rate because
each individual pixel is displayed in a time-sequential manner, analogous to raster
scan. Moreover, laser speckles should be suppressed for practical applications.
A LBS system is composed of a laser module and a beam scanner. A laser mod-
ule usually consists of laser sources (red, green, and blue laser diodes), collimation
optics, and a dichroic beam combiner. For the beam scanning [93–95], in this review
paper we focus on MEMS scanners. To illustrate the working principles of a MEMS
scanner, here we briefly describe the mirror configurations and scanning methods.
The scanning methods include raster scan and bi-resonant (Lissajous) scan. For the
raster scan, the horizontal axis is driven in resonance with a high sinusoidal fre-
quency (around 1 kHz), and the vertical axis is driven in a linear motion with a
low frequency (60–120Hz). The horizontal scanning can be performed unidirection-
ally or bidirectionally. Unidirectional scanning has more equidistant horizontal lines,
but bidirectional scanning (Fig. 8(a)) can provide a higher resolution. For the bi-
resonant scanning (Fig. 8(b)), both axes are driven with high sinusoidal frequencies,
generating a so-called Lissajous trajectory. Due to a higher scanning frequency in
the vertical direction, Lissajous scanning outperforms raster scanning in the vertical
FoV, but the pixel density is less uniform [96]. Mirror configurations can also be
divided into two kinds: two one-dimensional (1D) MEMS mirrors (Fig. 8(c)) and
one 2D MEMS mirror (Fig. 8(d)). The 2D MEMS mirror reduces alignment steps
and power consumption, but the cross talk between two axes will be unacceptable
for high-resolution microdisplays, leading to the adoption of a bulkier two-mirror
architecture.
5. DISPLAY METRICS
According to the inherent requirements of HVS mentioned previously, the five metrics
resolution density, response time, efficiency/brightness/lifetime, HDR, and compact-
ness are proposed to evaluate the performance of light engines for AR/VR applications.
In the following, we compare different light engines based on these metrics one by
one.
5.1. Resolution Density
In an AR/VR device, unless specified, the FoV usually refers to the diagonal direction.
Without considering PPD degradation due to MTF mismatch, for a VR headset with
100°FoV, to meet a normal human visual acuity (60 PPD), a 6K resolution panel is
required for each eye. On the other hand, a 2K resolution panel is required to support an
AR device with 40°FoV. Moreover, a smaller microdisplay panel size leads to a smaller
optics so that the formfactor of the headset will be more compact and lightweight.
Therefore, for VR and AR devices, the display resolution of 6K×6K or 2K×2K should
be packed into a 2-inch or 0.3-inch panel, which correspond to 2121 PPI and 4715
PPI, respectively. Such a demand challenges every display technology, including LCD,
OLED, and micro-LED. It not only increases the manufacturing difficulty, but also
has side effects on display performance, such as lower efficiency, more pronounced
angular color shift, and reduced image quality. Recently, several advanced designs
have been proposed to boost the resolution density of all display technologies. Next,
we discuss them one by one.
5.1a. LCD
Presently, transmissive LCDs are widely used as light engines for VR headsets due to
800 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Figure 8
(a) Raster scan (bidirectional scan); (b) bi-resonant scan; (c) two 1D MEMS mirrors;
(d) one 2D MEMS mirror. (a) and (b) Reprinted with permission from Reitterer et al.,
Proc. SPIE 11765, 1176504 (2021) [96]. ©2021 SPIE.
their higher PPI, higher brightness, and lower cost than the competing active-matrix
OLED displays, because the latter require more TFTs and compensation circuits to
provide steady current. However, the LCD’s resolution density is still insufficient to
eliminate the screen door effect. For present commercial VR products, the LCD’s
resolution density is in the range of 500–800 PPI. As a result, there is still clearly a
long way to go for the targeted 2000 PPI or higher. For a high PPI display, the subpixel
width is restricted by the TFT manufacturing process. Recently, Innolux has developed
process capability of narrow linewidth (1.5µm) and line space (1.5 µm), as shown in
Fig. 9(a) [97]. The corresponding subpixel size is minimized to 6 µm (1411PPI). A
resolution density over 2000 PPI could be achieved by the subpixel rendering. However,
the challenge of high-resolution density is not merely improving the manufacturing
process to achieve narrower linewidths. As LCD resolution density increases, many
challenges and issues arise. (1) Flicker effect, which refers to changes in the brightness
of displayed image over time [98]. In a LCD, the storage capacitor is formed by the
sandwich structure between pixel electrode, passivation layer, and common electrode.
Furthermore, its capacitance is proportional to the cross-sectional area of the pixel
and common electrode. As a result, the smaller the pixel size, the smaller the LC
capacitance. The insufficient storage capacitance is unable to maintain the target
driving voltage in entire frame time so that the pixel’s brightness is susceptible to
the voltage variation from data and scan lines. As a result, the image flicker is more
pronounced [99]. To increase the storage capacitance, Innolux proposed to use top
Vcom structure and thinner passivation layer thickness [97]. Moreover, Huo et al. [100]
used an additional passivation layer and common electrode to form a compensate
capacitor at the opposite side of the pixel electrode to increase the LC pixel’s storage
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 801
Figure 9
(a) Schematics of a LCD subpixel. (b1) Schematics of aperture ratio comparison
between thin and thick black matrices. (b2) Aperture ratios of displays with different
black matrix thicknesses and resolution densities. (c1) Schematics of color mixing
effect in LCDs. (c2) Angular color shifts of LCDs with different color filter thicknesses
and resolution densities. The image quality comparison is between (d1) a rounded
resin black matrix and (d2) a sharp metal black matrix. (b1) and (b2) Reprinted from
Yoshida et al., J. Soc. Inf. Disp. 30, 413–422 (2022) [101]. Copyright Wiley-VCH
Verlag GmbH & Co. KGaA. Reproduced with permission. (d1) and (d2) Reprinted
from Yao et al., SID Symp. Dig. Tech. Pap. 52, 735–737 (2021) [102]. Copyright
Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission.
802 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
capacitance, which helps reduce the flicker ratio by 50%. In other words, the brightness
variation during the frame time is reduced by 50%. (2) Small aperture ratio, which
refers to the ratio of transmission area to the total subpixel area. The structure of
each LCD subpixel is shown in Fig. 9(a), which consists of a TFT, driving line, and
aperture area. Normally, a black matrix is overlaid on top of the electric circuit to
prevent the current leakage of TFT originated from backlight illumination and on top
of the driving line to prevent the light leakage from LCD by blocking the uncontrolled
area. As a result, the aperture ratio of the LCD with 1411 PPI is only 15%.
Figure 9(b1) illustrates that if the color filter includes a thin black matrix (by using a
low-reflectivity metal multilayer film), then the gap between TFT substrate and black
matrix layer can be reduced. As a result, a smaller black matrix area and a larger aper-
ture ratio can be achieved. Figure 9(b2) demonstrates the aperture ratio improvement
of thin black matrix as a function of different display resolution density [102]. (3)
Color mixing effect, which refers to the light in one pixel passing through the adjacent
color filters [103,104]. For example, as shown in Fig. 9(c1), the emitted light from
the red subpixel crosses to the adjacent blue and green subpixels. Therefore, the color
purity is degraded at large viewing angles. This color mixing effect is more pronounced
when the thickness of the LC cell increases and the resolution density increases. From
the perspective of display resolution density, Fig. 9(c2) illustrates that the 1210-PPI
LCD (solid blue line) shows a more noticeable angular color shift than the 1058-PPI
LCD (dashed blue lines). From the perspective of cell thickness, a thinner black matrix
helps to reduce the overcoat thickness in the CFs, thereby reducing the CF thickness
from 3.3 µm to 2 µm. A thinner CF mitigates the color mixing effect so that the upper
threshold angle of indistinguishable color shift (∆u’∆v’ <0.02) extends from 24°to
33°as illustrated in Fig. 9(c2). (4) Rounded and non-uniform black matrices signif-
icantly degrade the image quality of high-resolution-density LCDs. The thin metal
black matrix (100–200 nm) with a much sharper edge than the traditional resin BM
significantly improves the clarity of the displayed images, as shown in Fig. 9(d) [102].
5.1b. OLED
Although the high-resolution-density WOLED microdisplays have been commercial-
ized by Sony and Kopin, to reduce the approximately 80% light loss caused by color
filters, novel methods have been proposed to pattern RGB organic light-emitting mate-
rial with high resolution density. In the traditional evaporation method, the organic
material is sputtered through a FMM to obtain the desired pattern. However, the res-
olution density of the FMM fabricated by the wet etching process is inadequate, and
the thick FMM (20 µm) will also form a wide shadow area during the evaporation pro-
cess. Overall, this approach limits the display resolution density to about 600 PPI. To
improve the resolution density of the FMM, as shown in Fig. 10(a1), Kim et al. [105]
used a multi-scan femtosecond pulsed laser to ablate the pattern shapes in a metal
mask. Such an ultrashort pulsed laser significantly avoids the deformation caused by
overheating. In addition, the metal foil is thinned down from 20 to 2µm to reduce
shadowing effects during evaporation. Based on the above improvements, an 1057-
PPI FMM (Fig. 10(a2)) has been produced successfully [106]. To further increase
the resolution density of the FMM, the beam size of the laser could be reduced.
By replacing the infrared laser with an UV laser, the beam size of the laser can be
reduced from 15 to 2 µm, and the resolution density of FMM increases to 2400 PPI.
In addition to metal shadow masks, Jiang et al. [107] also fabricated a high-resolution
density (2000 PPI) shadow mask from a silicon nitride film as shown in Fig. 10(b1)
and 10(b2). Silicon nitride films are suitable for photolithographic methods to achieve
high resolution density patterns and are thin enough (<2µm) to significantly reduce
the shadowing effects. However, to pattern OLED devices with high resolution density,
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 803
Figure 10
(a1) Schematics of FMM fabricated by a multi-scanned ultrafast laser. (Inset: SEM
image showing drilling hole fabricated by different laser pulses.) Reprinted from Kim
et al., SID Symp. Dig. Tech. Pap. 49, 1011–1013 (2018) [82] and Kim et al., J.
Soc. Inf. Disp. 28, 668–679 (2020) [105]. Copyright Wiley-VCH Verlag GmbH &
Co. KGaA. Reproduced with permission. (a2) SEM image of fabricated 1057-PPI
FMM. Reprinted from Kim et al., SID Symp. Dig. Tech. Pap. 52, 131–134 (2021)
[106]. Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permis-
sion. (b1) Schematics of shadow mask fabricated by silicon nitride films. Reprinted
from Jiang et al., SID Symp. Dig. Tech. Pap. 49, 1011–1013 (2020) [107]. Copyright
Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission. (b2) PL image
of patterned organic material evaporated through the silicon nitride films. Reprinted
from Jiang et al., SID Symp. Dig. Tech. Pap. 49, 1011–1013 (2020) [107]. Copy-
right Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission. (c1) EL
image and (c2) OLED device-degraded patterned organic material fabricated by the
photolithographic method. (c1) Reprinted from Malinowski et al., J. Soc. Inf. Disp.
26, 128–136 (2018) [109]. Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Repro-
duced with permission. (c2) Reprinted from Ke et al., SID Symp. Dig. Tech. Pap. 52,
127–130 (2021) [110]. Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Repro-
duced with permission. (d1) Schematics and (d2) SEM image of solution-processed
organic material patterned by direct photolithography. Reprinted by permission from
Gather et al., Adv. Funct. Mater. 17, 191–200 (2007) [111]. Copyright Wiley-VCH
Verlag GmbH & Co. KGaA. Reproduced with permission. (e1) Schematics and (e2)
SEM image of electrohydrodynamic printing OLEDs. Reprinted from Ref. [112] with
permission from the Royal Society of Chemistry.
it is not enough to simply increase the resolution density of the shadow mask. The
evaporation sources should also be extended from point, line, to area so that the angle
of incidence from the source to the mask can be reduced, which helps to suppress
the shadowing effects [108]. In addition to the patterning process though a shadow
804 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
mask, IMEC and Fujifilm [109] have also developed a photographic method to pattern
organic light-emitting materials with 1-µm linewidth, as shown in Fig. 10(c1). During
the photographic patterning process, the organic material film is first evaporated in
a vacuum chamber. Then, a substrate is removed from the vacuum chamber, and a
photoresist material forms a design pattern on the organic material film. By reactive
ion etching, the designed photoresist pattern is transferred to the organic material thin
film. After that, the patterned organic thin film is loaded back to the vacuum chamber
to evaporate the second color organic material, and then the same patterning process
repeats. Organic light-emitting materials for RGB colors are patterned one by one
to achieve full-color display. However, the exposure of organic materials in a non-
vacuum environment, etching process, residual photoresist, etc., all lead to defects in
the organic materials, resulting in degraded optical properties of OLED devices, such
as reduced lifetime and increased driving voltage (∼2 V increasement at a brightness
of 1000 nits) [110]. The degradation of patterned OLED devices can be observed by
comparing the L–I–Vcurves between the non-patterned and patterned OLED devices
as shown in Fig. 10(c2).
Different from the photographic process that transfers the pattern on the photoresist
to the organic material thin film through the etching process, Gather et al. [111] used
oxetane-functionalized crosslinkable electroluminescent polymer as the functional
photoresist. Crosslinking in films is induced by UV exposure. Therefore, as shown in
Fig. 10(d1), by performing patterned UV exposure on the desired regions and rinsing
the film with organic solvents to dissolve the unexposed areas, patterned organic light-
emitting pattern with a 2-µm linewidth can be achieved (Fig. 10(d2)). Printing methods
with fewer materials waste are also developed for patterning organic light-emitting
materials. More recently, electrohydrodynamic jet printing [112,113], which controls
the electric field between the nozzle and the substrate (Fig. 10(e1)), enables printing
inks with a narrow linewidth down to around 5 µm, as demonstrated in Fig. 10(e2).
However, ink sticking to the nozzle, print uniformity (no coffee-ring effect), and time-
consuming printing process are the main challenges for its widespread application.
Overall, the high-resolution-density patterning method by increasing the shadow mask
resolution density and improving the evaporation source is most compatible with
existing fabrication processes. Moreover, other advanced patterning techniques may
break the vacuum environment and expose susceptible organic materials to moisture
and air during the fabrication processes. Therefore, the lifetime issue of OLED devices
fabricated by these novel patterning techniques, which disrupt the vacuum environment
during the fabrication should be further evaluated.
5.1c. µLED
As mentioned previously, the µLED chips need to be transferred from a wafer to
a display substrate. As a result, the resolution density of an RGB µLED display is
determined by two factors: the chip size and the gap between them. The chip size is
determined by the mesa manufacturing process of the LED chip and can be as small
as 2 µm and the gap is determined by the transferring technologies. Compared with
the “pick-and-place” approaches, monolithic integration methods achieving a gap of
about 1 µm shows advantages for high-resolution-density microdisplay applications.
Two monolithic integration methods have been implemented: flip-chip bonding and
wafer-bonding style. The major difference between these two methods is whether the
µLED chips are defined before or after the bonding process. For flip-chip bonding,
the µLED chips are well defined on the LED substrate, and the metal-bonding arrays
are fabricated on the CMOS substrate. Then, the µLED array is aligned and bonded
to the CMOS backplane by a thermal-compression method. For wafer-bonding style,
the LED epitaxy is firstly bonded to the entire CMOS substrate. Then, the µLED are
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 805
Figure 11
(a1) Schematics of µLED array fabricated by flip-chip bonding and wafer bonding.
(a2) Cross talk effect in the flip-chip bonding µLED array. Reprinted with permission
from [114]©The Optical Society. (a3) LEE of flip-chip µLED with different incli-
nation angle. (Red, black, and blue lines indicate the LED chip with diameter 5µm,
5µm, and 20 µm and thickness 2.4 µm, 1 µm, and 2.4 µm, respectively). Reprinted from
[115] under a Creative Commons license. (b) RGB µLED microdisplays combined by
a trichroic prism. Reprinted by permission from Zhang et al., J. Soc. Inf. Disp. 26,
137–145 (2018) [116]. ©2018 Wiley (c) Color conversion micro-LED microdisplays.
Reprinted from Kawanishi et al., J. Soc. Inf. Disp. 29, 57–67 (2021) [117]. Copy-
right Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission. (d1)
Schematics of vertically stacked µLED film. (d2) DBR interlayer in stacked µLED
film for avoiding light loss. (d3) The pixelized vertically stacked µLED connected to
the driving circuit. (d4) SEM image of pixelated µLED array. (d1)–(d4) Reprinted
from Ref. [123] with permission from the Royal Society of Chemistry. (e1) L–I–V
curve of the multi-wavelength MQWs µLED and its emitted colors under different
driving currents. (e2) PWM method for modulating the brightness of a multi-color
LED. (e1) and (e2) Reprinted with permission from [124]©The Optical Society.
806 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
formed by the photolithography process followed by plasma etching. Both methods
have pros and cons, and the manufacturing process and yield challenges have been
summarized in [91]. From the perspective of optical performance, Fig. 11(a1) shows
that the inverted trapezoidal geometry µLED chips fabricated by flip-chip bonding
process are more convenient to extract the emission light to the normal direction than
the trapezoidal geometry fabricated by wafer-bonding process. However, Fig. 11(a2)
demonstrates a severe optical cross talk of the flip-chip µLED with connected n-GaN
between each LED chip [114]. A complete isolation trench between each pixel is
required to eliminate the optical cross talk. Moreover, the inclination angle of the
inverted trapezoidal geometry should be tailored for improving the LEE of the LED
chips at normal direction, as shown in Fig. 11(a3) [115].
The µLED microdisplays fabricated by the monolithic integration methods are nor-
mally single color, so various methods have been developed to obtain full colors.
The most convenient way is to combine the displayed images of three separate RGB
µLED microdisplays through a trichroic prism [92], which is then projected by a
lens. Because the RGB subpixels are provided individually by three separate panels,
rather than side by side in a single panel as shown in Fig. 11(b) [116], the LED
pitch is the same as the pixel pitch, and the resolution density can be as high as
10,000 PPI. However, the aligning and driving three display panels synchronously
are still challenging. In addition, from the optics perspective, the trichroic prism is
bulky and the distance between the projection lens and the emission panel is relatively
long, which also reduces the FoV of the AR device. Another solution is to assemble
color conversion materials on top of the blue or UV µLEDs as shown in Fig. 11(c)
[117]. Numerous QD pattering processes have been proposed to pattern a uniform
QD array with high resolution density, sufficient film thickness, and reasonable sta-
bility [118,119]. Inkjet printing and photolithography are the two main methods for
patterning QD arrays due to their ability to form reasonable film thicknesses (few
micrometers). However, the uniformity and nozzle clogging are common issues of
inkjet printing technologies. For the photolithography, QD material degradation, due
to exposure to solvents and water, reduces the efficiency and stability of the QD
arrays. Other methods, such as microcontact printing, electron-beam lithography, and
light-driven ligand crosslinker, could provide a narrower linewidth which is favorable
for high-resolution density display applications, but the insufficient film thickness
[120–122] results in a severe blue light leakage and low color conversion efficiency.
Therefore, the challenges of patterning high-resolution-density QD arrays are twofold:
to achieve not only narrow linewidths but also high aspect ratios (thickness-to-width
ratios).
Figures 11(d1)–11(d4) shows the device structure of a vertically stacked µLED display
on a driver circuit through monolithic 3D integration [123]. The driver circuits, metal
lines, and electrode pads of each RGB pixel are prefabricated before the monolithic
integration process. Then, to avoid the alignment process, thin film transfer was
performed to bond the uniformly stacked µLED thin film layer on the driving substrate,
as shown in Fig. 11(d1). It is worth noting that for the stacked µLED thin film layer, due
to the high absorption of InGaP-based red LED, it is usually placed at the bottom of the
stack. Furthermore, to reflect the back-emitted green and blue light, a high distributed
Bragg reflector (DBR) film is deposited on the interface between the blue and red LEDs
(Fig. 11(d2)). After that, Fig. 11(d3) illustrates that high-resolution-density vertically
stacked µLED pixels can be defined by photolithographic methods, including the
cathode and anode of each layer of LEDs. Finally, the cathode and anode of each LED
are interconnected with the prefabricated driver circuit common and bus lines. The
pixel size of LEDs can be as small as 0.5 µm as shown in Fig. 11(d4), regardless of
driving capability. Overall speaking, vertically stacked RGB µLEDs [125–130] can
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 807
triple the resolution density, and the monolithic 3D integration could achieve precise
alignment, resulting in high resolution density. However, the complicated fabrication
process still hinders its widespread applications.
In addition to stacking RGB LEDs, multi-wavelength MQWs have also been explored.
For example, one MQW with a higher concentration of indium emits red light and the
other MQW with a lower concentration of indium emits blue light [131]. As shown
in Fig. 11(e1), the emission wavelength of the µLED is determined by modulating the
driving current, and the brightness is determined by modulating the emission time.
Similar to stacked RGB LEDs, multi-wavelength MQW full-color µLED chips can
emit RGB light on the same µLED chip, providing a three-fold increase in resolution
density as compared to the side-by-side RGB µLED displays. However, there are two
main drawbacks for this kind of devices. (1) Color purity. As both MQWs emit at
different driving currents, the color purity of such µLEDs has yet to be improved.
(2) Complicated driving circuit. As shown in Fig. 11(e2) [124], because the driving
current for blue light is much higher than that for red light, the blue light-emitting
time (0.6%) should be much shorter than the red light-emitting time (100%) to have
the same brightness. If we further consider generating 10-bit gray levels, the driver
is a big burden to overcome. Despite the above-mentioned challenges, these µLED
displays still attract a lot of attention because they do not require complex transfer
and stacking processes. For example, Porotech demonstrated a single-pixel full-color
adjustable display prototype at SID 2022 and won the I-Zone Best Prototype Award.
5.1d. LCOS
The resolution density of a LCOS is mainly limited by the FFE and the limited voltage
swing [132]. When the pixel pitch of a LCOS device is comparable with or even smaller
than the cell gap, unequal voltages on adjacent pixels generate a pronounced horizontal
electric field, which causes the FFE. For this reason, a straightforward method to
mitigate FFE is to employ a thinner cell gap by using a higher birefringence LC
mixture [133–135]. To demonstrate the FFE in an amplitude LCOS panel, we simulate
the FFE of MTN and VA modes using TechWiz LCD 3D program. Figure 12(a) shows
the simulated reflectance for the 6-µm-pitch VA cell and 6-µm-pitch 90°-MTN cell
when the three adjacent pixels are operated at dark–bright–dark state. The overall dark
state in the 90°-MTN cell is quite good, except for a small lobe at the right-hand side
of the dark pixel. In contrast, the FFE in the VA cell is rather strong: the fringing
fields induce a broad dark stripe near the right edge of the bright pixel. Due to the
negative dielectric anisotropy, the LC directors tend to reorient perpendicular to the
incident plane. At a certain position, most of the LC directors are orthogonal to the
incident polarization direction, leading to the unchanged polarization state. Under such
a condition, the incident light is completely blocked, and a dark stripe splits the bright
pixel into two unequal parts. In summary, the FFE can reduce the optical efficiency
and contrast ratio of an amplitude LCOS device.
Several strategies have been proposed to suppress the FFE of amplitude LCOS devices
since the early 2000s. In 2002, Fan-Chiang et al. optimized the slope of pixel electrodes
to reduce FFE [137]. For the 80°-MTN and 45°-TN modes, the optimal ITO electrode
slope is 1; however, changing the slope can hardly mitigate the FFE of the VA mode.
In 2005, Fan-Chiang et al. adopted a circularly polarized (CP) light to eliminate the
FFE of the VA mode [136] (Fig. 12(b)). At x=xb, most LC directors are aligned along
the ydirection. Therefore, little phase retardation is accumulated at x=xbwhen the
incident light is linearly polarized along the xdirection, leading to a dark region. A
CP light will help solve this issue because the phase retardation is still accumulated no
matter where the LC director is orientated. Later, Li et al. proposed a double mirror
structure and butterfly-like pixel shape to compensate the deformation of LCs [138].
808 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Figure 12
(a) Simulated FFE of the MTN and VA modes in an amplitude LCOS device and (b)
CP light employed to eliminate the FFE of the VA modes. Reprinted with permission
from Chiang et al., Appl. Phys. Lett. 87, 031110 (2005) [136]. Copyright 2005, AIP
Publishing LLC. (c) FFE without patterned pretilt angle (left) and the reduced FFE
with patterned pretilt angle (right). Reprinted from [132] under a Creative Commons
license.
Vithana developed a VA mode with large pretilt angles to remove the disclinations and
introduce twisted structure to keep a high contrast ratio [139], enabling a pixel pitch of
∼3µm. Liao proposed patterned pretilt angles to mitigate the FFE [140] (Fig. 12(c)).
In this method, the pixel region is divided into two zones with a FFE border width,
and such an inhomogeneous LC alignment helps mitigate FFE. The pretilt angle of the
optimized inner zone and outer zone is 88°and 85°, which can significantly reduce
the FFE as shown in the right-hand side of Fig. 12(c).
For a phase LCOS device, which is commonly called spatial light modulator (SLM),
the required phase change is 2π. As a result, the FFE is more severe than that in an
amplitude LCOS device because of the twice thicker cell gap. To demonstrate the FFE
in a SLM, we use TechWiz LCD 3D to calculate the LC director distribution and then
calculate the accumulated phase change by [141]
∆φ(x)=4π
λ∫d
0
neno
n2
o+(n2
e−n2
o)sin2(θ(x,z))
dz,(1)
where noand neare ordinary and extraordinary refractive indices of the LC material,
respectively, dis the cell gap, and θis the tilt angle of the LC directors. The simulated
configuration is as follows: three adjacent pixels operate at on–off–on states, constitut-
ing binary phase gratings. The on state refers to the maximum phase change and the off
state represents the minimum phase change. Figure 13(a) shows that the FFE blurs the
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 809
Figure 13
(a) Simulated FFE of the HG mode in a phase LCOS device. (b) Relation between
FoV and pixel pitch in a holographic display.
phase edges due to a transition region connecting the on and off states. Owing to the
smeared phase profiles, the diffraction efficiency in holographic displays is reduced.
In particular, 1 µm or less pixel pitch is required to achieve a wide FoV holographic
display [14,35] (Fig. 13(b)). To suppress the strong FFE for the 1-µm pixel pitch, the
dielectric shield wall structure was proposed [142], but the LC alignment remains
a major challenge [143]. Other strategies aim at integrating a modeled or measured
SLM response [144–146] (point spread function) into holographic calculations.
In addition to FFE, the limited voltage swing of the LCOS poses another challenge
to achieve 1-µm pixel pitch. In such a small pixel area, the CMOS backplane can
only support a limited voltage swing. For a LCOS panel with more than 5000 PPI,
the driving voltage is usually below 3.3 V [147]. Optimizing material parameters
can be an effective solution to reduce the operating voltage while keeping a fast
response time. Some main parameters of a LC mixture include birefringence (∆n),
dielectric anisotropy (∆ε), elastic constants (Kii), and rotational viscosity (γ1). A
higher ∆nenables a steeper slope of the voltage-dependent phase curve. On the other
hand, elastic constant and dielectric anisotropy jointly determines the Freedericksz
threshold voltage; a larger dielectric anisotropy helps to reduce the threshold voltage
and operating voltage. On the other hand, response time is determined by the cell gap
and viscoelastic constant. Therefore, these parameters are interrelated.
To illustrate the optimization process, we use DIC A4907 [148] as an example. The
voltage-dependent phase curves corresponding to different material parameters can
be simulated via TechWiz LCD. When the dielectric anisotropy remains unchanged
(∆ε=7.3), a low operating voltage of 3.3 V can be obtained if ∆nincreases to 0.23.
However, it is challenging to obtain such a high birefringence while retaining a low
viscosity. Although tolane compounds help realize this goal, their ultraviolet (for her-
metic sealing the SLM) and blue photostability remains a concern. The other method
to reduce the operating voltage is to decrease the threshold voltage. For this method,
the slope of the voltage-dependent phase curve basically remains unchanged, and the
curves shift towards the lower voltage side because of its lower threshold voltage, if
the dielectric anisotropy increases from 7.3 to 16.3. However, the challenge for this
method is to obtain such a high dielectric anisotropy without increasing the viscos-
ity too significantly. The difluoromethoxy bridge (–CF2O) [149] and isothiocyanato
group (–NCS) [150] could improve ∆εwhile retaining a low viscosity. However, the
molecular π–πconjugation is broken in the former, leading to a low ∆n, whereas
the latter decreases the voltage holding ratio because the –NCS dipole tends to trap
ions. If we combine these two mechanisms to lower the operating voltage to 3.3 Vrms,
810 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Figure 14
(a) (b)
3.3 V
(a) Operating voltage and (b) threshold voltage as a function of ∆nand ∆ε.
numerous combinations of ∆nand ∆εcan be searched in the parameter space. Fig-
ure 14(a) shows how the operating voltage varies with ∆nand ∆ε: as both ∆nand ∆ε
increase, the operating voltage is dramatically decreased from 5 V to 2.5 V. A certain
operating voltage of 3.3V can be realized by different combinations of ∆nand ∆ε, as
indicated by the same color corresponding to different ∆nand ∆ε. Figure 14(b) shows
how the threshold voltage varies with ∆nand ∆ε: the threshold voltage decreases as
∆εincreases. Among all the combinations of ∆nand ∆εthat can generate 3.3 V oper-
ating voltage, a medium ∆εis desirable due to a larger V2/Vth ratio. For example, if
∆n=0.21 and ∆ε=10.3, then the operating voltage is 3.3V and the threshold voltage
is 1.05 V. Such a LC mixture is not too difficult to obtain.
5.1e. DLP
For DMD, the resolution density is mainly limited by the gap between micromirrors.
Presently, the smallest pixel pitch is 5.4 µm as implemented in the Texas Instruments
DLP3010. A certain gap distance should be maintained to allow each micromirror to
rotate independently without colliding with each other. In comparison with DMD, the
gap between pixels in a LCOS panel can be much smaller because the light modulation
is through the liquid crystal layer, instead of mechanically moving mirrors. When the
pixel size decreases, the gap between mirrors may be comparable with the mirror size,
leading to a strong diffraction. The diffraction dramatically reduces the efficiency and
the contrast ratio of the DMD, making an ultrahigh resolution density difficult.
5.1f. LBS
The resolution density of the LBS is determined by the number of resolvable spots.
Under the paraxial approximation, the number of resolvable spots along the horizontal
direction can be calculated by the following equation [151]:
Nh=
θoptz
aλz/D
=
θopt
aλ/D
=
θoptD
aλ,(2)
where zis the distance from the MEMS scanner to the screen, θopt is the full optical
scan angle, Dis the clear mirror aperture, ais a shape factor determined by the beam
profile, mirror shape, and the amount of spot overlap between adjacent spots. For
example, a=1.22 if the shape of mirror is circular and laser beam profile approaches
plane wave. The practical avalue ranges from 0.75 to 2, depending on the laser beam
profile, mirror shape, and so on. Some relevant parameters are sketched in Fig. 15(a).
The horizontal pixel size is typically chosen to be equal to the spot size, aλz/D. From
Eq. (2), the product of θopt and Ddetermines the maximum horizontal resolution. On
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 811
Figure 15
(a) Schematic of the laser beam scanning. (b) Schematic of the dynamic mirror
deformation.
the other hand, the number of resolvable spots along the vertical direction is calculated
as
Nv=fhKubKrt
Fr
,(3)
where fhis the horizontal scanner frequency, Fris the frame rate, Kub is 1 for uni-
directional scanning and 2 for bidirectional scanning, and Krt considers the retrace
time of the vertical scanner. From Eq. (3), the ratio of the horizontal scan frequency
to the frame rate determines the maximum vertical resolution. Bidirectional scanning
and a large Krt relieve the requirement for the horizontal scanner frequency. Overall,
a large mirror aperture and a wide scan angle are desired to increase the horizontal
resolution; a high horizontal scanner frequency is demanded to obtain the high ver-
tical resolution. There is also a trade-off between frame rate and vertical resolution.
In real applications, a larger mirror aperture can worsen mirror deformation as shown
in Fig. 15(b), leading to image distortions and decreased scanning frequency [152].
According to Brosens’s formula [153], the mirror deformation relates with the mirror
aperture as
δmax =0.217 ρf2D5θmech
Et2
m
,(4)
where fis the scanner frequency, ρis the material density, Dis the mirror aperture,
Eis the modulus of elasticity, tmis the mirror thickness, and θmech is the zero-to-peak
mechanical scan angle. To ensure a good image quality, the mirror size is limited by
the maximum tolerable mirror deformation, thereby limiting the horizontal resolution.
5.2. Fast Response Time
The display with fast response time has two main functions in an AR/VR display
system: one is to alleviate motion blur from the displayed image and the other is to
generate multi-depth images to mitigate the VAC effect. To alleviate motion blurs, a
fast MPRT is desired. To achieve MPRT <1.5 ms, both high frame rate and low duty
ratio (the backlight-on time in a frame) can be considered. However, as the frame rate
increases from 60 to 120 and then to 240 Hz, the MPRT value will decrease noticeably
first but then gradually saturate. Therefore, a low duty ratio (10–20%) illumination
plays a key role to suppress the motion blurs. On the other hand, to generate multi-
depth images, because the content of images at different depths is diverse, a high
812 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
frame rate is required to quickly update the image information at different depths. Each
display system has different constraints to achieve fast response time. For example, the
response time of LCDs and LCOS microdisplays is mainly limited by the employed
LC material. On the other hand, for OLED and µLED displays, because the device
response time is in the microsecond and nanosecond range, the constraints come from
the backplane (driving circuit). Furthermore, there is a trade-off between high frame
rate and bit depth in DLP displays, and between high frame rate and display resolution
in LBS displays. In next section, we dive into each display system one by one.
5.2a. LCD
High frame rate (240 Hz) LCDs have been commercialized for gaming monitors.
However, most of these commercial products still have a LC response time of about
5 ms if the overdrive and undershoot voltage [154] is not applied. With overdrive and
undershoot, the LC response time can be reduced to about 1 ms. Under such condition,
if the duty ratio remains at 100%, then the MPRT is still ∼4 ms, which is mainly
determined by the 240-Hz frame rate. Such a MPRT is still far from the target 1.5 ms
for motion-blur-free CRT-comparable images [155]. In addition, the overdrive function
may further cause inaccurate gray levels and degraded image quality. Therefore, a low
duty ratio (<30%) has been widely employed to reduce the MPRT of the display.
Figure 16(a) illustrates the driving time chart of a LC panel with duty illumination,
which consists of three parts: scanning time, LC response time (LCRT), and backlight
emission time (backlight is on). In other words, the sum of scanning time, LC response
time, and backlight emission time should be smaller than the frame time. As shown
in Fig. 16(b), if we set the scan time for 2160 gates to 5 ms and take an emission
duty ratio of 25% as an example, the LCRT of the 120 Hz panel is 1.5 ms and the
corresponding MPRT matches the target value: 1.5 ms. Although using a smaller
emission duty ratio (e.g., 10%) allows more time for the LC to response (1.5–2.5 ms),
an instantaneously brighter backlight results in a higher driving current and higher
power consumption. In addition, one can use a lower frame rate (e.g., 90 Hz as shown)
and a smaller duty cycle (e.g., 15%) to achieve the same MPRT (1.5 ms). At such
a frame rate and duty ratio, the remaining LCRT is about 4.5 ms, which is friendly
to LC devices. However, as mentioned previously, low-frame-rate displays are not
suitable for multi-depth applications. Thus, to simultaneously achieve high frame rate
that generates multi-depth images and small duty ratio that mitigates motion blurs, a
fast-response LC device is required. In the following discussions, we focus on novel
LC modes and electrode designs to reduce LCRT without the overdrive/undershoot
function to maintain decent image quality and simple driving circuit. According to
Eq. (5), LCRT can be reduced by using a thinner cell gap and a LC material with a
smaller viscoelastic coefficient (γ1/Kii):
τ0=γ1d2
π2Kii
.(5)
In Eq. (5), the corresponding elastic constant Kii depends on the employed LC mode.
Taking FFS mode as an example, the corresponding Kii is the twist elastic constant
(K22), whose value is about two or three times smaller than the splay and the bend
elastic constants (K11 and K33), which corresponds to TN and VA LCDs. However, the
negative dielectric anisotropy LC material used in VA mode usually exhibits a higher
rotational viscosity because the polar groups are in lateral positions.
To further improve LC response time, various approaches for forming virtual walls
have been proposed and widely applied [156–160]. Yoon et al. [161] demonstrated a
three-electrode vertically aligned LC panel with a longitudinal electric field at the top
of the gap region and fringe field at the gap peripheral region (VA-FFS mode). The LC
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 813
Figure 16
(a) Timeline of LCD driving. (b) Calculated maximum allowable LCRT and the
corresponding MPRT as a function of duty ratio. The black dashed lines represent
a cathode ray tube (CRT)-like MPRT =1.5 ms. (c) Schematics and transmittance
distribution VAFFS mode LCD. Reprinted with permission from [161]©The Optical
Society. (d) FFS mode LCD with 0°rubbing angle and non-zero pretilt angle. Reprinted
from Matsushima et al., J. Soc. Inf. Disp. 29, 221–229 (2021) [160]. Copyright
Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission.
directors in the gap region do not reorient, forming virtual walls to provide a strong
restoring force. The device structure and its transmittance distribution are depicted
in Fig. 16(c). Furthermore, Fig. 16(d) illustrates that the FFS LCD can easily form
virtual walls by using zero degree rubbing angle and non-zero pretilt angle [160]. The
LC directors on top of electrodes are fixed, and on both sides of the electrode rotate
clockwise and counterclockwise, respectively. When the virtual wall is formed, three
anchoring forces are present simultaneously: top and bottom alignment layers, and
the virtual walls. Therefore, the LC response time is no longer governed solely by the
cell gap, but also by the distance between virtual walls. Taking the short-range lurch
control in-plane switching proposed by Japan Display Inc. as an example, its LC decay
time is determined by [160]
τ0=γ1
π2K22
d2+K11
l2.(6)
Under normal conditions, K11 ≈2K22 [162]. If the distance between two dead zones is
the same as the cell gap, the LC response time would be reduced by a factor of three.
Despite the benefits of fast response time mentioned previously, in the virtual wall
region, the incident light on the non-rotating LC molecules has no accumulated phase
and is, thus, further blocked by the top analyzer. Overall, these advanced LC devices
have improved response time at the expense of efficiency.
5.2b. LCOS
For LCOS devices, color sequential operation is widely adopted for achieving high
resolution. Therefore, high frame rate is essentially desirable to avoid color break-up.
The frame rate of LCOS devices is mainly limited by the LC response time. Due to
814 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
the reflective operation in the LCOS, its response time is approximately four times
faster than the corresponding transmissive LCD. To reduce response time further, high
birefringence and low viscosity LC mixtures [163] are particularly desirable. Chen
et al. [164] developed a fluorinated high-birefringence and low-viscosity negative
dielectric anisotropy LC mixture to enable a VA LCOS with submillisecond response
time. Later, Peng et al. [165] reported a LC mixture with low viscosity and high
clearing point for a color sequential MTN LCOS. The average gray-to-gray rise time
is 0.5 ms and gray-to-gray decay time is 0.2 ms at T=55°C. Abeeluck et al. [166]
reported an amplitude LCOS device with a color subframe rate greater than 720 Hz to
avoid color break-up. On the other hand, the required phase range for a phase LCOS
SLM is 2π, which is about twice as large as its amplitude counterpart. Therefore, the
cell gap of an LCOS SLM is twice as thick if the same LC material is employed. As a
result, the response time for a phase LCOS device is approximately four times slower
than its amplitude counterpart [167]. For this reason, it is challenging to obtain the
LCOS SLM with a high frame rate. Polymer network liquid crystals can be a potential
solution to achieve submillisecond response [168], but the high driving voltage limits
its adoption in the LCOS SLM. Recently, by using a high birefringence (∆n≈0.21) yet
photo-stable LC mixture, a LCOS SLM with 2πphase change at λ=633 nm and ∼3
ms response time (rise +decay) [169], which enables the LCOS SLM to be operated
at 300 Hz, has been demonstrated. To further decrease the response time, ferroelectric
liquid crystal (FLC) was used in LCOS devices because of its microsecond response
time [170,171]. To realize the inherent advantages of FLCs, photoalignment by nano-
size azo-dye was employed [172]. Three electrooptic modes, namely surface stabilized
FLC, deformed helix ferroelectric (DHF), and electrically suppressed helix, have been
developed. For FLC-based LCOS microdisplays, the DHF mode stands out because
of its low driving voltage, high resolution, and ultrafast response time. However, such
a FLC has no threshold voltage to block the TFT voltage fluctuation, and the driving
circuits are rather complicated. On the other hand, for holographic displays, the first-
order diffraction efficiency is only about 40.5% due to its binary phase modulation
[173].
5.2c. DLP
Among all the light engines, DMD has the highest frame rate, which is enabled by a fast
electromechanical response (∼µs) of micromirrors. Because the micromirror operates
at a binary (on/off) state, only black and white images can be generated without any
processing. To deliver an n-bit gray-scale image, PWM at each pixel is usually used.
The key point is that each frame is divided into n sequentially projected bit-planes.
The gray level of each frame depends on the ratio of the whole on-state time to one
frame time because the luminance is averaged by human eyes due to the persistence of
vision. The exposure time of each bit-plane is sequentially t, 2t, . . . , 2(n–1)tµs, which is
weighted by the corresponding power of 2. Figure 17(a) shows that how an 8-bit gray-
scale image is generated with PWM. It is noted that the minimum bit-plane exposure
time is limited by the DMD’s refresh time. With block-based control, the minimum
bit-plane exposure time can be as little as 2 µs [174]. To project a color image, the
above sequential bit-planes are repeated for each color channel, thus tripling the frame
time to 3 ×(2n– 1)t. When the bit depth increases, the total frame time exponentially
increases, leading to the exponentially decreasing frame rate. This trade-off between
frame rate and bit depth is demonstrated in Fig. 17(b). One solution to break the
exponential relationship is to use intensity-modulated light with the intensity of 2(n
–i) multiplied by the maximum intensity of light source Lat the ith bit-plane. For
this method, each bit-plane has the same exposure time twith different illumination
intensity (Fig. 17(c)), leading to a linear relationship between the total frame time of
n×tµs and the bit depth. However, the maximum averaged brightness is significantly
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 815
Figure 17
(a) Schematic of an 8-bit gray-scale image with the PWM method. (b) Trade-off
between frame rate and bit depth with the PWM method (t=2µs) (c) Schematic of
an 8-bit gray-scale image with the intensity-modulated light (d) Trade-off between
maximum brightness and bit depth with the intensity-modulated light.
Figure 18
(a) Schematic of a volumetric near-eye display composed of a HDR LED, a DLP
projector, a focus-tunable lens, and a combiner lens. ©2018 IEEE Computer Society.
Reprinted, with permission, from Rathinavel et al., IEEE Trans. Visual. Comput.
Graphics 24, 2857–2866 (2018) [175]. (b) Schematic of tomographic near-eye displays
consisting of a FSAB module, a relay optics, a LCD module, and a focus-tunable lens.
Reprinted from [176] under a Creative Commons license.
reduced, which can be calculated as
Lmax =L(2−(n−1)+· · · +2−1+1)t
nt
=2
n(1−1
2n)L.(7)
For example, when the bit depth n=16, the maximum brightness is reduced to about
L/8, as shown in Fig. 17(d). For AR applications calling for high brightness, this
solution is impractical.
The high-speed response makes the DMD a perfect candidate for a multifocal AR
enabled by time multiplexing. Rathinavel et al. [175] achieved a full-color volumetric
near-eye display with a focus ranging from 20 cm to 4 m via the synchronized operation
of the HDR LEDs, the DMD and the focus-tunable lens, as shown in Fig. 18(a). The
number of single-color binary images decomposed from a color image is determined
816 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
by the ratio of the DMD’s refresh rate to the target display refresh rate. Their DMD’s
refresh rate and target display refresh rate is 16,800Hz and 60 Hz, respectively, which
can support 280 binary images distributed at different depths. Later, Lee et al. [176]
presented a tomographic near-eye display enabled by the synchronization of focus-
tunable lens and a fast spatially adjustable backlight (FSAB). Figure 18(b) shows a
benchtop prototype consisting of a FSAB module, relay optics, LCD module, and
a focus-tunable lens. At the heart of the FSAB module is a DMD, which can show
binary images at a very high frame rate. The relay optics provides a magnified image
of the DMD panel at the LCD plane. The LCD module determines the gray levels and
the resolution. The complementary relations between the FSAB module and the LCD
module breaks the trade-off between frame rate and bit depth and, thus, support 24-bit
depth color as well as a high frame rate for 80 focal planes.
5.2d. LBS
The frame rate of LBS is usually 60 Hz, which might cause image flickering. In raster
scan, the frame rate is the same as the vertical scanning frequency. According to Eq. (3),
there is a trade-off between frame rate and vertical resolution. Bidirectional writing
schemes and a short retrace time of the vertical scanner [177] help increase the frame
rate at a given vertical resolution. On the other hand, increasing horizontal scanner
frequency can also boost the frame rate at a given vertical resolution. Reducing the
mirror size is a straightforward way to obtain a higher horizontal scanner frequency,
but the diffraction effect may limit the horizontal resolution.
5.2e. µLED and OLED
The device response time of OLED and µLED is in the microsecond and nanosecond
level, respectively, which enables a single device to be driven at megahertz or gigahertz
for optical communication applications. However, the situation is completely different
for display applications where millions of pixels should be driven to an accurate
brightness within a frame time. Considering a high-resolution and high-frame-rate
display, the addressing time for each row is extremely short. In general, the driving
timeline of a self-emissive display consists of three stages: compensation, data input,
and emission.
During the compensation period, the driving circuit detects the variations of each TFT,
device, etc., and precisely modulates the corresponding driving current for each device
to improve its image uniformity. At data and emission stages, compensated driving
data is delivered to each pixel, and then at emission stage devices are turned on. The
image performance under various compensation time (2, 4.3, and 33.3 µs) is illustrated
in Fig. 19(a). The presently developed compensation pixel circuits with progressive
emission, as Fig. 19(b) depicts, need to execute the compensation and data input
operations within the line time. Here, we take a 120Hz FHD (1920 ×1080) OLED
as an example, which has a line time of 4.3 µs (=1 s/120/1920), so the compensation
time in progressive emission approach must be less than 4.3 µs, e.g., 2 µs. It is clearly
shown in Fig. 19(a) that when the compensation time is reduced, the compensation
performance degrades, resulting in a non-uniform image. To resolve this issue, several
compensation methods including simultaneous emission (Fig. 19(c)) [178–180], block
emission (Fig. 19(d)) [181], parallel addressing (Fig. 19(e)) [182,183], and external
compensation, have been proposed to extend the time for compensation.
For the simultaneous emission method, all the pixels are compensated at the same time
and then data voltages, which are determined by the desired images, are input to pixels
row by row. Finally, all pixels are turned on simultaneously. Because the compensation
and data input operations are separated, the compensation time is adjustable and not
restricted by the line time. When the line time decreases with the increased frame
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 817
Figure 19
(a) Displayed image of an OLED panel with different compensation times. ©2019
IEEE Solid-State Circuits Society. Reprinted, with permission, from Lin et al., IEEE
J. Solid-State Circuits 54, 489–500 (2019) [184]. Driving diagram of (b) progressive
emission, (c) simultaneous emission, (d) block emission, and (e) parallel addressing.
rate, simultaneous emission can still yield an excellent performance of compensation
to enhance the image quality. However, this method has less emission time than the
progressive emission method, resulting in lower luminance. In addition, global signals
that are shared by all pixels are commonly used in the simultaneous emission method.
Once these signals are activated, all pixels are turned on and illuminated at the same
time, increasing power consumption, and generating a large current to flow through
the parasitic resistance on the power lines. This causes a serious drop in voltage, which
in turn degrades the image uniformity.
To extend the emission time for the reduction in the generated currents, the block
emission method, which combines the simultaneous emission and progressive emis-
sion methods, has been developed. The block emission method divides the whole
panel into Nblocks, and each block contains pixels of Mrows where N×Mis the
total number of the row lines. The operation of pixels in each block is the same as the
simultaneous emission method, ensuring sufficient compensation time. Each block is
activated sequentially, which is the same as the progressive emission method, enabling
that the emission time using the block emission method is longer than the simulta-
neous emission method. Consequently, adopting the block emission method has not
only enough compensation time but also longer emission time, which helps improve
the image quality. Nevertheless, a cross talk occurs between each block due to the
parasitic capacitor of the adjacent pixels, causing the Mura defect. The phenomenon
is more severe especially for those high PPI displays.
818 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
In contrast, the parallel addressing method also divides the compensation operation
from the data input operation, and no data voltages are used for the compensation,
enabling that the compensation time of adjacent pixels to overlap and be adjusted to a
proper duration for a better compensation performance. As the frame rate increases,
the compensation using parallel addressing method does not need to complete within
a shorter line time, acquiring an adequate time to provide an accurate compensation.
Therefore, the high uniformity of images is ensured despite the high frame rate of
displays. Nevertheless, more components including TFTs and capacitors, and more
complicated control signals are utilized to implement the parallel addressing method.
In general, it requires about six to seven transistors, which is almost doubled as
compared with the other three driving methods. Therefore, parallel addressing method
may increase the difficulty for achieving high pixel density and decrease the stability
of the pixel circuits. In addition to the compensation by the internal pixel circuits,
another approach is to use external compensation. The external compensation uses
a detection circuit out of the internal pixel with compensation algorithm to make
driving currents uniform. Therefore, the structure of internal pixels can be simplified.
This method performs compensation when displays are offline, so its compensation
time is independent of the line time, making the external compensation to provide
an accurate compensation even though the frame rate is greatly increased. However,
the required algorithm is complicated and additional memories are needed, which
undoubtedly increases the complexity and cost of the display system. The above-
mentioned methods can deal with the issue of insufficient compensation time for use
in high-frame-rate displays, but they have their own problems to be conquered.
5.3. Efficiency
When a microdisplay is used as the light engine for a see-through AR system, high
brightness is required due to image washout caused by strong ambient light. In general,
the output luminance of the AR systems should be at least 500 nits and 10,000 nits,
for indoor and outdoor applications, respectively. If we further consider the light
efficiency of the optical combiner (e.g., ∼1% for the waveguide approach), the required
microdisplay brightness should be up to 1 million nits. This is indeed the case for
some commercial prototypes demonstrated. On the other hand, for VR systems, since
ambient light is blocked by the enclosure, the desired brightness on the viewer’s end is
usually set at about 150 nits. However, to eliminate motion blur, a small duty emission
ratio of 10–20% is commonly used, which means the OLED or LCD panels should
provide 10 times the instantaneous display brightness. In addition, considering the 75%
optical loss of a decent pancake VR display, the display should be a further four times
brighter. Therefore, microdisplays of about 6000 nits are required. To deliver enough
brightness in an AR/VR system, the overall efficiency of the projection system should
be improved. Otherwise, the limited capacity of battery would significantly restrict
the device operation time. In an AR/VR system, the light emitted by the display is
first received by the projection system to generate an enlarged image for the HVS.
Here, we separate the display system into two categories: one is panel-based display
devices and the other is scanning display. Now, let us first focus on the efficiency of
panel-based projection systems. The optical efficiency of a system depends not only on
the efficiency of the panel itself, but also on the proportion of emitted light collected
by the projection system, which could be estimated and defined by the well-known
étendue conservation [185]. As shown in Fig. 20(a), the étendue of the projection
system in AR/VR displays can be defined by the FoV of the display system and the
surface area (AEP in Fig. 20(a)) of the eyebox (VR displays) or optical combiner (AR
displays). On the other hand, the étendue of the display panel may be determined by the
product of the display radiation pattern and the display surface area. The bottom part
of Fig. 20(b) shows the broad emission cone display typically produced by an OLED
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 819
Figure 20
(a) Étendue definition in a projection display system. (b) Schematic of applying broad
(bottom part of display) and narrow (top part of display) emission cone display in
a projection display system. (P1, P2, and P3 indicates top, center, and bottom pupil
position, and its corresponding received image is shown in the right.) (c) Schematic of
matched (top part of display) and unmatched (bottom part of display) display chief ray
with a projection display system. (d) Schematic of a LBS projection display system.
or µLED display. On the other hand, the upper part of Fig. 20(b) shows the narrow
emission cone display typically produced by LCD, LCOS, and DLP. For the broad
emission cone display (red light path), the étendue of the display is greater than the
étendue the projection system, i.e., part of the light emitted by the display is wasted,
which means that the coupling efficiency is less than 1. The benefit is that it provides
uniform brightness in the different pupil position within the entrance pupil area. On
the other hand, for the narrow emission cone display (orange light path), the étendue
the display is less than that of the projection system, and the coupling efficiency is
1, which means that there is no power loss during coupling, but the brightness of
the display will change as eye saccades in the entrance pupil area. Therefore, the
optimal design of the projection system is to match the étendue of the display panel
and projection system to maximize the optical efficiency without compromising image
quality. In Figs. 20(a) and 20(b), a telecentric projection system is used. However, in
some projection systems, because telecentric design is not employed, the chief rays
of the projection system’s display emission and reception cones may not match at
the edges of the display panel, as shown in Fig. 20(c). In such projection display
systems, the radiation pattern of the display should be locally modulated to match the
corresponding receiver cone to maximize the optical efficiency [52,186].
For the panel-based projection systems, the overall efficiency of the projection system
can be defined by
η=ηuD ×ηE,(8)
820 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
where ηuD is the optical efficiency of the display panel and ηEis the coupling efficiency
between the display panel and the projection system.
Furthermore, for a self-emissive display (OLED and µLED), the optical efficiency of
the display panel is determined by
ηuD =ηLED/OLED ×ηcolor ×Tfilm,(9)
where ηLED/OLED is the optical efficiency of the emissive devices and Tfilm is the
transmittance of the optical films. Color filters are employed in a white OLED to
generate RGB colors and in QD-based µLED displays to purify colors. Depending
on the spectral match between the color filter transmission spectra and the device
emission spectra, varying degrees of optical loss may occur. For RGB OLEDs and
RGB µLEDs, color filters are not usually needed, leading to a higher ηcolor.
For the non-emissive display, the optical efficiency of the display panel is determined
by
ηuD =ηS×ηI×ηp×ηm×ηcolor ×Tfilm,(10)
where ηSis the efficiency of the illumination source, ηIis the efficiency of the integrator,
ηPis the polarization management efficiency, ηmis the optical efficiency of the light
modulator (LCD panel or DMD), and ηcolor is the color-generating efficiency. More
specifically, the integrator efficiency represents the optical losses during light shaping
and homogenization, such as in LCD backlight units, and in integrator rods and fly-eye
lenses in LCOS/DLP microdisplays. The polarization management loss is due to the
polarizer in the illumination system that generates polarized light for the LC device.
Novel compact polarization recycling systems help reduce polarization management
losses. The optical efficiency loss of light modulators may come from diffraction,
reflection, a limited fill factor, and so on. In LCDs, color filters are used to generate
colors, leading to about two-thirds optical loss. For LCOS and DMD, color sequential
operation is adopted to eliminate color filters, leading to a higher ηcolor.
Different from panel-based projection displays, the LBS microdisplay usually does
not require the collimation or projection lens in an AR display [187], as shown in
Fig. 20(d) Therefore, the overall efficiency ηis equal to the efficiency of the LBS
ηuD. For the AR based on a waveguide combiner, the scanned laser beam can be
directly coupled into the waveguide via an input coupler because the pixel in the LBS
is already in the angular domain. The LBS can also enable the Maxwellian display
based on a free space combiner [36], thus exhibiting a focus-free advantage [188]. To
summarize, the small étendue of the LBS helps minimize the coupling loss. In the
remaining section, we focus on reviewing the optical efficiency of each light engine
based on the above parameters.
5.3a. LCD
LCD has been developed for decades, but its low optical efficiency has always been the
Achilles heel. Fortunately, many methods have been proposed to solve this issue. For
example, dual brightness enhancement films are employed to increase the polarization
management efficiency by 60%, and optical losses in CFs can be mitigated by QDs
with narrow emission spectra. In addition, brightness enhancement films are used to
concentrate the emitted light onto the desired viewing cone, much like what we desire
to do in projection systems, matching the display’s emission cone to the projection
system’s acceptance cone. Based on the topic of this review paper, here let us pay
special attention to the LCDs used in AR/VR display systems. In AR/VR display
systems, the requirements for high-resolution density and fast-response LC panels,
results in small aperture ratios and dead zones, respectively. As a result, compared
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 821
Figure 21
Imaging results of a Fresnel VR system with (a1) conventional BL and (a2) directional
BL at center fovea. (a1) and (a2) Reprinted with permission from [189]. ©The
Optical Society (b1). Device structure of a directional backlight unit consisting of three
microprism structures on three surfaces of light guide plate. Reprinted with permission
from [190]. ©The Optical Society. (b2) Directional backlight unit consisting of grating
structure with different spatial densities on top of light guide plate for directional
extraction of light. Reprinted by permission from Opt. Commun. 459, Zhang et al.,
Directional backlight module based on pixelated nano-gratings,125034, Copyright
2020 [191], with permission from Elsevier. The optical efficiency distribution of a
Fresnel VR system (c1) without and (c2) with a spatially radiation pattern modulator.
(c1) and (c2) Reprinted with permission from [186]©The Optical Society.
with conventional direct-view LCDs, the LCDs intended for AR/VR systems are much
less efficient.
On the other hand, compared to conventional LCDs, the fixed eyebox region in AR/VR
systems also provides a chance to improve the coupling efficiency. The small pupil
size of HVS constrains the étendue of the projection system, so only a narrow emission
cone can finally be received by the HVS. Figure 21(a) compares the displayed image
of a VR headset with a LCD light engine implementing directional or conventional
backlighting [189]. The LCD light source with a directional backlight shows brighter
images. The directional backlight unit can be achieved by applying microlens arrays
(MLAs), micropyramid films, or reflective microstructures. As shown in Fig. 21(b1),
Wang et al. [190] proposed a directional backlight by using microprism arrays, whereas
as shown in Fig. 21(b2) Zhang et al. [191] demonstrated a grating method. The
collimated light source is steered into the lightguide plate, and then the traditional
scattering particles are replaced by the surface relief grating (SRG). The grating pitch
is designed to diffract the light in lightguide upward to surface normal. Moreover,
the density of the grating increases from the near side to the far side of the light
source to keep a good brightness uniformity. The angular divergence is 6.17°, and
the brightness uniformity exceeds 90% for a phone-sized backlight unit. Recently,
Maimone and Wang [52] demonstrated holographic films based on emissive backlights
for a compact VR system. In this application, a light beam from a point source at the
edge of the lightguide propagates, and it is diffracted by the holographic films in the
lightguide to form an extended source. Furthermore, Fig. 21(c1) illustrates that in a
822 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
VR display system without a telecentric design (Fig. 20(c)), because the acceptance
cones of the projection systems are spatially different, applying a directional backlight
in a VR display system may result in a severe vignetting effect, which indicates that
in a projection system, pixels in the center of the display may appear brighter than
those at the edges. In 2021, Hsiang et al. [186] proposed a patterned inverted prism
film for spatially modulating the radiation pattern of the LCD backlights to locally
match the chief ray of display and projection system. Figure 21(c2) illustrates that they
could obtain uniform optical efficiency distribution (without vignetting effect) from
the center to the edge of the display panel. Notably, the complicated backlight unit of
LCD provides room for modulation of the display radiation pattern before the pixel-
defining layer. As a result, image quality can be better maintained than modulating
the display radiation pattern after the pixel-defining layer typically used for OLED or
µLED displays.
5.3b. OLED
A high-brightness OLED device with a high driving current density usually suffers
from shorter lifetime and image burn-in issues. Thus, there is urgent need to develop
high-brightness OLED [192] with a low current density so that its lifetime can be
extended to a reasonable range. The total efficiency of an OLED device is governed
by two factors: internal quantum efficiency (IQE) and LEE. The IQE of an OLED
device is mainly determined by the emission mechanism of the emitter, such as a
fluorescence emitter (IQE ∼25% (single) or ∼75% (triple)), phosphorescence emitter
(∼100%), and thermally activated delayed fluorescence emitter (∼100%). Considering
high-quantum-yield emitters and their reasonable lifetimes, high-IQE OLED devices
typically use fluorescent emitters for blue color and phosphorescent emittersfor green,
red, and other colors. As the IQE of OLED devices has approached 100%, recent
research has focused on improving the LEE of OLED devices, initially limited to
20% due to TIR. For an OLED microdisplay, due to the opaque silicon backplane,
the OLED devices must be a top-emitting device. Furthermore, to display clear and
sharp images, the added light extraction components should not sacrifice the resolution
density of the microdisplay. For a top-emitting OLED, the strong cavity effect formed
by the top and bottom metal electrodes leads to a large surface plasmon polariton
(SPP) mode lost (air mode ∼25%; waveguide mode ∼35%; SPP mode ∼40%) [193].
Several methods, such as scattering layers, MLA, and corrugated structures, have
been widely used to extract the power from OLEDs in waveguide and SPP modes
[77]. Scattering layers are widely used in OLED devices for lighting applications,
which consist of nanoparticles, reflective diffusers, or nanoporous films to break
the TIR and to extract the substrate or waveguide mode [194–196]. However, for
high-resolution-density displays, image blurring caused by the scattering layers is the
most critical issue. Furthermore, scattering layer widens the emission cone of the
OLED display, approaching the Lambertian distribution, which is also undesirable in
projection display systems.
A MLA can be laminated directly on the top electrode of OLED devices to extract the
waveguide mode or on the top glass to extract the substrate mode. However, the former
approach may damage the vulnerable OLED devices and require a high-refractive-
index MLA to match the high-refractive-index organic materials or ITO, whereas
the latter method may result in server blurred images due to a thicker substrate.
Laminating a MLA onto a microdisplay is much more challenging than using it
in traditional displays or lighting apparatus [197,198]. For traditional displays, one
OLED device corresponds to an array of microlenses. In contrast, for microdisplay
applications, because the light-emitting area of each OLED device is very small
(∼µm), the microlens size should be in the same level. In other words, one microlens
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 823
will correspond to one OLED device. In addition, a microlens with high aperture
ratio, high aspect ratio, and small gap between the emissive layers is required to
reduce cross talk and to maximize efficiency. Sony [199] proposed to assemble an
inverted microlens above a top-emitting OLED device, which not only increases the
brightness in normal direction by a factor of 1.8 (Fig. 22(a1)) but also reduces cross talk
between subpixels. Furthermore, as shown in Fig. 22(a2), the offset between microlens
and OLED subpixel shifts the chief ray of the emitted light by ±20°from the center
[200]. Therefore, by optimizing the arrangement between the microlens and OLED
subpixels, spatially different radiation patterns can be achieved for improving the
overall optical efficiency and image quality of the OLED projection display system.
Finally, the OLED projection system with >1,000,000 nits (green light) has been
demonstrated (Fig. 22(a3)). Another approach is corrugated structure. Figure 22(b1)
shows the device structure of a corrugated OLED [201]. The main challenge of the
corrugated structure is that even though the corrugated structure can efficiently extract
the trapped waveguide and SPP modes through Bragg diffraction, the corrugated
structure simultaneously weakens the cavity effect due to its low reflectivity. Moreover,
the extracted waveguide and SPP modes cause a fierce angular intensity variation
(Fig. 21(b2)) [202].
In a projection system, a higher LEE does not mean a better outcome. Preferably, we
only want to extract the emitted light of the display whose emission cone falls into
the acceptance cone of the projection system; the other light extracted outside the
acceptance cone of the projection system will become stray light, which may degrade
the image quality. Therefore, how to realize OLED devices with directional angular
intensity distribution is a hot topic. There are two main approaches: one is to increase
the reflectivity of the top cathode through the DBR layers to obtain a stronger cavity;
the other is to optimize the thickness of each organic layer to achieve a highly direct-
emitting OLED device (Fig. 22(c1)) [203]. Although the higher the intensity of the
microcavity, the more concentrated the light is in the normal direction, the EQE of
the OLED device first increases and then decreases. Therefore, Hsiang et al. [186]
systematically optimized OLED devices to achieve high efficiency, high color gamut,
and low color non-uniformity for VR display applications. The second is to design the
cavity of the OLED device with strong waveguide mode first, and then to eliminate
the air mode, and then use the nanostructure to form Bragg diffraction to extract the
waveguide mode with an extremely narrow emission cone to the normal direction
(Fig. 22(c2)) [204].
Multiple emissive layers (RGB layers or BY layers) can be used to realize a white light
emitting OLED device [205]. The electrons and holes of the OLED device are injected
from the cathode and anode, respectively. The thicknesses and electron/hole mobilities
of the different emissive layers determine the recombination rates of different emissive
layers. For example, due to the low hole mobility of the blue emissive layer, most of
the recombination is close to the bottom blue emissive layer (near the anode) and
results in an undesired emission spectrum with dominant blue light and insufficient
green and red lights. Therefore, to achieve electron–hole balance and desired white
light emission spectra, the exciton recombination rates of different emission layers
should be carefully designed by adjusting the thickness and position of each emission
layer and intercalation interlayer. Furthermore, from the driving voltage perspective,
a p-type doped hole injection layer (HIL), and a thinner emissive layer with low
hole mobility, are required to reduce the overall operating voltage. After the device
configuration is optimized, the L–I–Vcurves and emission spectra of the multilayer
WOLED are shown in Fig. 23(a). The current efficiency is 6 cd/A, and the brightness
at 3.3 V is 4200 nits, which is about 100 times higher than the initial state.
824 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Figure 22
The optical performance of microlens OLED devices: (a1) efficiency improvement;
(a2) radiation pattern shift by offset microlens; and (a3) projected image of green writ-
ing. (a1)–(a3) Reprinted from Itonaga et al., SID Symp. Dig. Tech. Pap. 51, 683–686
(2020) [200]. Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with
permission. (b1) Schematic and (b2) radiation pattern of a corrugated OLED (inset:
SEM image of the corrugated structure). (b1) Reprinted with permission from Amoah
et al., ACS Appl. Mater. Interfaces 14, 9377–9385 (2022) [201]. Copyright 2022
American Chemical Society, https://doi.org/10.1021/acsami.1c21128. (b2) Reprinted
with permission from [202]©The Optical Society. (c1) Schematic of resonant OLED
devices (inset: radiation pattern of resonant OLED devices). Reprinted from Liang
et al., Adv. Opt. Mater. 10, 2101642 (2022) [203]. Copyright Wiley-VCH Verlag
GmbH & Co. KGaA. Reproduced with permission. (c2) Spatial patterns from a con-
ventional OLEDs and a waveguide emission OLEDs. Reprinted from Fu et al., Adv.
Mater. 33, 2006801 (2021) [204]. Copyright Wiley-VCH Verlag GmbH & Co. KGaA.
Reproduced with permission.
Tandem OLED devices have long been widely used for TV applications. When a
voltage is applied, the current-generating layer generates an extra pair of electron and
hole, so when more OLED devices are connected in series, more carriers are generated
to improve the current efficiency (cd/A). For example, OLEDWorks [84] demonstrate
3-stack, 4-stack, and 5-stack tandem WOLED devices with current efficiency of 12
(cd/A), 15 (cd/A), and 20 (cd/A), respectively. Thus, a tandem OLED device exhibits a
higher peak brightness and longer lifetime in comparison with the single-layer OLED
because it can deliver higher optical power by a lower current density. But the driving
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 825
Figure 23
(a) Optical performance of OLED devices with optimized interlayer thickness. (Inset:
the emission spectra of the optimized OLED devices.) Reprinted with permission from
Lee et al., J. Ind. Eng. Chem. 105, 132–137 (2022) [205]. ©2022 The Korean Society
of Industrial and Engineering Chemistry. (b) Optical performance comparison of tan-
dem OLED devices with thin and thick cavity designs. Reprinted with permission from
Bae et al., ACS Appl. Electron. Mater. 3, 3240–3246 (2021) [208]. Copyright 2021
American Chemical Society, https://doi.org/10.1021/acsaelm.1c00406. Schematic of
OLED devices with customized resonant cavity by RGB devices with (c) different
IZO (CCL) thicknesses and (d1) different reflective metasurface. (d2) Photolumi-
nance image of the metasurface OLED devices. (c) Reprinted by permission from
Org. Electron. 87, Kim et al., “Primary color generation from white organic light-
emitting diodes using a cavity control layer for AR/VR applications,” 105938 [211],
Copyright 2020, with permission from Elsevier. (d1) and (d2) Reprinted by permission
from Joo et al., Science 370, 459–463 (2020) [214]. ©2020 AAAS.
voltage required to connect OLEDs in series is also proportional to the number of
stacks. In a high-resolution-density microdisplay, small transistors limit the maximum
operating voltage. To reduce the driving voltage, p- and n-doped transport layers are
applied to increase the carrier mobility [206]. However, the side effects of the doped
transport layer in electrical cross talk between each subpixel should also be considered
[207]. Another way to reduce the driving voltage is to design all the emissive layers to
resonate at the first antinode to achieve a thinner device profile as shown Fig. 23(b).
Bae et al. [208] demonstrated that the driving voltage for achieving 1000 nits was
reduced from 9.5 to 5.8 V, despite the lower current efficiency of a thin OLED device.
If we set 9 V as the maximum driving voltage for the backplane, the thin OLED device
shows 10 times the brightness. In addition, because the color filter absorbs about 70%
of the emitted light from the WOLED device, how to remove CFs or reduce the optical
loss of CFs becomes crucial. By resonating the white light emission spectrum of
the OLED device to the desired color corresponding to RGB subpixels, the emission
spectrum of the OLED device can be matched with the transmission spectrum of
the color filter, thereby reducing the light loss caused by the color filter [209–213].
Strong microcavity design in top-emitting OLED devices is a widely used approach to
achieve narrow emission spectra. However, unlike conventional RGB OLEDs, which
826 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
have a single color in the microcavity, WOLED devices have two or three colors in
the cavity at the same time. In 2020, Kim et al. [211] proposed to apply different
thicknesses of IZO as a cavity control layer (Fig. 23(c)), and Joo et al. [214] used a
reflective metasurface as a phase modulation layer (Figs. 23(d1) and 23(d2)), both of
which successfully resonated RGB lights from WOLED devices. However, due to the
wide emission spectrum of the WOLED devices, the color purity is not good enough
to completely remove the color filters, but the efficiency is improved.
5.3c. µLED
As the size of a µLED chip decreases, its EQE decreases, indicating that surface defects
at the periphery of the LED chip results in a larger non-radiative Shockley–Read–Hall
(SRH) recombination. This phenomenon can be evaluated by the ABC model as
follows:
IQE =Bn2
An +Bn2+Cn3;A=A0+vs
P
A,(11)
where nis the carrier concentration, and A,B, and Care the coefficients related to SRH
recombination, radiation, and Auger recombination, respectively. In addition, the SRH
coefficient (A) consists of two parts: the bulk SRH coefficient (A0) and the product
of the surface recombination velocity (vs) and the ratio of perimeter to surface area.
As the size of LED devices decreases, the ratio of perimeter to surface area increases
and leads to a larger non-radiative SRH recombination, and this effect is more severe
in LED devices with larger surface recombination velocity [215]. The most common
way to mitigate the size effect of µLEDs is to apply dielectric sidewall passivation in
the LED chips to reduce defects generated during the etching process. The passivation
process is developed for all kinds of µLEDs, such as AlGaInP red LEDs [216–219],
GaN blue and green LEDs, and nanorod LEDs [220–222]. In addition, by reducing the
ratio of electrode size to LED mesa size or adding a current confinement layer to avoid
non-radiative recombination on the sidewalls of the LED chip, the IQE of the µLEDs
can also be improved. In addition to the surface defects, studies have been conducted
to discuss the influence of device surface recombination velocity on the size effect of
the µLEDs. Smith et al. [223] analyzed the size-dependent IQE for GaN-based green
and blue µLEDs smaller than 10 µm, as shown in Fig. 24(a). It shows that for LED
chips >10 µm, blue µLEDs are more efficient than green µLEDs. It is well understood
that the low efficiency of green µLEDs (high indium doping) is caused by high defects
densities and a strong quantum-confined Stark effect. Surprisingly, however, blue
µLEDs are less efficient than green µLEDs when the chip size is <10 µm. In other
words, the size-dependent efficiency drop for blue µLEDs is more severe than for green
µLEDs, and it can be explained by the lower surface recombination velocity of green
µLEDs [224]. According to the above comparison, both bulk SRH coefficient and
surface recombination velocity of the LED device affects the efficiency of the µLED
chip, so that a more efficient LED device at a large chip size does not guarantee a better
performance as the chip size shrinks. Similar considerations should be taken when
determining whether AlGaInP or InGaN µLEDs are more efficient for red µLEDs. High
indium doping (>30%) in InGaN red µLEDs result in a high bulk SRH coefficient
and a low surface recombination velocity. As a result, InGaN red µLEDs exhibit a low
IQE [225,226], but its efficiency is independent of the chip size [227]. On the other
hand, AlGaInP LEDs with better lattice matching have low bulk SRH coefficients but
high surface recombination velocity [228], which results in high-efficiency red LEDs
accompanied by severe size-dependent efficiency drop [229]. Figure 24(b) illustrates
the size effect of two types of red µLEDs. The debate persists until now, with both
teams improving their devices to achieve the target EQE (5%) for the small µLED chip
size (5 µm×5µm) [230].
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 827
In addition to the size-dependent IQE of µLED devices, the LEE of µLEDs also
depends on the die size. An inverted trapezoidal LED chip shape with reflective
sidewalls was used in the optical simulation. The small LED chip size improves
the interplay between the dipole emission and the oblique sidewall, which redirects
the trapped waveguide mode to the air mode. In contrast to the IQE drop, LEE
typically increases as the LED chip size decreases (Fig. 24(c)) [231,232]. Like an
OLED microdisplay, the directional-emission µLED devices are also favored in the
projection display system. Since the 2000s, different types of diffractive photonic
crystals (PhC) have been applied on LED devices to extract waveguide modes for
a higher LEE [233–235]. By diffracting the guided modes to the normal direction,
LED devices with an efficient directional light emission can be realized. However, a
thick LED chip (few micrometers) simultaneously supports many guided modes with
various in-plane vectors, so a constant reciprocal lattice vector of PhC cannot diffract
all the guided modes in the normal direction at same time and achieve non-directional
emission. Therefore, the thickness (<1µm) of the LED device and the location of the
MQW are tailored to concentrate the power into one or two guided modes. Then, PhC
is applied to diffract the high-power guided modes to the normal direction, as shown
in Fig. 24(d) [236]. Drawbacks of using PhC on LED devices are also evident, such
as defects caused by drilling holes on LED devices, and the lower current injection
efficiency of PhC LED devices.
Figure 24(e) depicts a parabolic LED chip with reflective sidewalls [237–239]. It is
also proposed to concentrate the emitted light in the normal direction. However, the
difficulty of fabricating parabolic LED chips below 10µm remains a challenge. In
addition, the parabolic shape design can only work well for the luminescence at the
center of the MQW, and the luminescence at the edge of the chip may interfere with the
directionality of the radiation pattern. The resonant cavity LED (RC-LED) with strong
cavity, resonating the emitted light in the LED device through the high reflectivity
upper and lower mirrors, is also used to improve the directivity of the radiation pattern
as shown in Fig. 24(f) [240,241]. However, due to the multiple reflections of light back
and forth in the RC-LED device, a tiny optical loss in the cavity can eventually lead to
a severe efficiency drop. To reduce absorption by metal mirrors, DBR and nanoporous
GaN with high reflectivity and low absorption are used to form cavities [242,243].
Furthermore, as in designing a PhC-LED, a thin LED device also needs to limit the
number of resonant modes to achieve a higher directionality radiation pattern.
5.3d. LCOS
To analyze the efficiency of a LCOS microdisplay, we can turn to Eq. (10) to consider
all the factors involved. Some general analyses have been discussed above. The oper-
ation of a LCOS requires a polarized light. Therefore, the polarization management
efficiency ηpis an important part for determining the final efficiency. Next, we focus
on the parameters ηmand ηpin the LCOS. We can estimate ηmas follows:
ηm,LCOS =(Twindow)2×FF ×Rmirror,LC ×ηdiffraction.(12)
In a LCOS device, the window transmittance refers to the transmittance of the cover
glass. The fill factor is the fractional coverage of aluminum electrode in a pixel. The
gap between pixels can be very narrow (<0.2 µm), leading to a large fill factor (>90%).
The reflectance in the LCOS projection system is determined jointly by the mirror
reflectance and the LC mode. In contrast, the reflectance in the DMD system is only
determined by the mirror. Despite a small gap between pixels, the diffraction still
exists because of the formation of amplitude gratings [244]. A great deal of efforts
has been made to increase the optical efficiency of LCOS devices. Li et al. [138]
proposed a double mirror structure to realize a 100% fill factor and to eliminate the
828 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Figure 24
(a) Size effect of InGaN green and blue µLEDs. Reprinted with permission from Smith
et al., Appl. Phys. Lett. 116, 071102 (2020) [223]. Copyright 2020, AIP Publishing
LLC. (b) Size effect of InGaN and AlInGaP red µLEDs. Reprinted from [230] under
aCreative Commons license. (c) LEE of µLED with different chip size. Inset: SEM
image of the µLED chip. Reprinted with permission from Ley et al., Appl. Phys.
Lett. 116, 251104 (2020) [232]. Copyright 2020, AIP Publishing LLC. (d) Direc-
tional emission of PhC-LED. Red dashed lines indicate the Lambertian distribution.
Reprinted with permission from Rangel et al., Appl. Phys. Lett. 98, 081104 (2011)
[236]. Copyright 2011, AIP Publishing LLC. (e) Schematic of a parabolic surface
µLED. Inset: SEM image of the µLED chip. Reprinted from Henry et al., SID Symp.
Dig. Tech. Pap. 47, 747–750 (2016) [238]. Copyright Wiley-VCH Verlag GmbH & Co.
KGaA. Reproduced with permission. (f) Schematic and simulated radiation pattern
of a resonant LED. Reprinted from Khaidarov et al., Laser Photon. Rev. 14, 1900235
(2020) [240]. Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with
permission.
diffraction. In such a device configuration, a bottom mirror is introduced, and the top
mirror and bottom mirror are separated by a dielectric layer with a certain height.
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 829
Figure 25
(a) Amplitude LCOS integrated with AR systems. (b) Basic PCS with a PBS. (c)
Conventional PCS with a rod integrator. (d) Conventional PCS with lens integrators. (e)
complex-ordered PG-based PCS. (f) PCS with a PG and a patterned QWP sandwiched
between lens integrators.
Light shining on the pixel electrode region is reflected by the top mirror; light shining
on the pixel gap region is reflected by the bottom mirror. In this way, double mirrors
behave as a plane mirror instead of pixelated mirrors. Abeeluck et al. [166] reported
a high-performance LCOS microdisplay with an ultra-small pixel pitch, a fill factor
of 93.5%, low diffraction losses and high reflectance. Such a high fill factor can still
be achieved at an ultra-small pixel pitch, reducing absorption and diffraction losses.
High reflectance arises from an optimized DBR laminated between the backplane and
the LC layer.
Most LC devices [245,246], including LCOS, operate based on a polarized light,
whereas many light sources are unpolarized, such as arc lamps and energy-efficient
LEDs. Polarized lasers are rarely used in LCOS due to the cost issue. One way to
obtain polarized light from an unpolarized source is to absorb one polarization by
an absorptive polarizer, but it leads to ηp<50%. In general, two approaches are
commonly used to improve the conversion efficiency. A polarization recycling system
leveraging a RP [247] with hundreds of coatings (e.g., 3M’s DBEF) randomize light
polarization and contribute to the conversion efficiency statistically through multiple
roundtrips. In contrast, a polarization conversion system (PCS) is deterministic, where
light only passes through the system once and the orthogonal component follow
distinctly different paths. A typical LCOS integrated with a PCS is illustrated in
Fig. 25(a). Furthermore, a basic PCS is sketched in Fig. 25(b). Light from the source
encounters a PBS and is divided into two beams with orthogonal polarizations (s and
p), respectively. The transmitted beam remains its original polarization, whereas the
other is reflected by the PBS and passing through a half-wave plate (HWP) orientated
45°to the polarization vector. Thus, all output light shares the same polarization, but
the final projected images will suffer from severe luminance non-uniformity, where
two bright regions are centered on the top half and bottom half of the output image.
In most cases, a PCS will include either a lenslet or a rod integrator to shape the
circular beam to a rectangular format and homogenize the output. A PCS with the
combination of two PBSs and a rod integrator [248] is illustrated in Fig. 25(c). The
830 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
incident unpolarized light is separated into s and p polarizations by the first PBS and
achieves the same polarization after one beam passes through the HWP. Both beams
are sent into the mixing rod for uniform illuminance. The rod can be hollow and
coated with a highly reflective material, or it may be a solid, and the incident light
propagates through TIRs. Owing to multiple reflections, the position of each ray at
input and output ports are uncorrelated and therefore the illuminance at the output can
be spatially homogenized. Empirically, the uniformity of light intensity distribution
is proportional to the length of rod integrator and numerical aperture of the incident
beam and is reversely proportional to the beam diameter and refractive index of the
rod. Only detailed raytracing with many rays can assure rod design with acceptable
uniformity at the output end. Although in theory the rod integrator is not polarization
preserving, but in practice the relatively short rod length preserves the polarization
well enough and the phase difference between s and p polarizations remains relatively
small even at a large incident angle.
In a typical PCS with a lenslet integrator [249] (Fig. 25(d)), the first lenslet array
images the source onto the corresponding lenslet in the second array, forming an array
of virtual sources. The PBS array and the HWPs split each source into two paths
and convert them into the same linear polarization state. Each lenslet in the second
array images the lenslet in the first array onto the microdisplay. Although the light
distribution of each lenslet is asymmetric, the summation largely cancels each other,
and the overlaying creates a much more uniform illumination than each individual
lenslet does. Too few lenslets would deteriorate the output uniformity, but too many
lenslets could cause the critical alignment issue.
As the form factor of LCOS shrinks down, such a traditional PCS with PBS cannot
be simply scaled down accordingly due to its inherent bulkiness and fabrication
difficulties. LC-based Pancharatnam–Berry optical elements (PBOEs) are polarization
sensitive to orthogonal CP lights and exhibit a high diffraction efficiency [250]. By
controlling the 2D pattern of the LC molecular orientation, an arbitrary phase pattern
can be achieved, such as polarization gratings (PGs) and PB lenses. A polarization
conversion device based on PG was proposed by Du et al. [251] in Fig. 25(e). A
complex-ordered multidomain LC polymer film is fabricated with photo-alignment
technique. Each domain is square-shaped and equally splits into four subdomains
of PGs, where each grating vector points to the center of the domain. LCP and
RCP lights are spatially separated at a short distance from the polymer film. An
additional patterned QWP with ±45°optical axis is placed to convert both CP lights
to the same linearly polarized light. Although the device shows ≥90% conversion
efficiency, it requires precise alignment. In addition, only highly collimated input light
is accepted, and the output illuminance is not uniform. Kim et al. [252] demonstrated a
compact PCS with a uniform PG and a patterned QWP, sandwiched in a pair of lenslet
integrators, as shown in Fig. 25(f). However, the lenslet integrator cannot guarantee a
uniform output when the LCOS form factor keeps decreasing.
5.3e. DMD
The DMD belongs to light modulating display; thus, Eq. (10) works for the DLP
microdisplay as well. Compared with LCOS, DMD does not require a polarized light,
leading to ηp=100%. Therefore, here we only focus on the parameter ηm. According
to TI’s published white paper [253], ηmcan be estimated using
ηm,DMD =(Twindow)2×FF ×Rmirror ×ηdiffraction.(13)
From Eq. (13), the overall optical efficiency is a product of window transmittance
Twindow, fill factor (FF), the mirror reflectance Rmir ror, and diffraction efficiency
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 831
ηdiffraction. The DMD windows depend on the packaging types: type a uses Corn-
ing glass; wafer-level packaging uses Eagle XG glass. Both types adopt anti-reflection
thin film coatings to increase transmittance. Let us assume each coated surface has
an approximately 99% transmittance. Therefore, the transmittance after a single pass
and a double pass will be 98% and 96%, respectively. The fill factor is the fractional
mirror coverage (on-state mirrors) as viewed from the illumination direction, which
depends on the micromirror size, micromirror gap, and tilt angle. For commercialized
DMD products from TI, the typical on-state FF is larger than 90%. The FF less than
100% causes a small amount of light absorbed and thus reduces the optical efficiency.
The third factor is the mirror reflectance. The aluminum-based mirrors employed in
the DMD usually exhibit an 89% reflectance in the visible range. Window trans-
mittance, fill factor, and mirror reflectance should be as high as possible to obtain
maximum optical efficiency. Diffraction effect in the DMD is a main limiting factor
for achieving a higher optical efficiency, which is related to the pixel pitch, fill factor,
and incident wavelength and angle. For the flat state micromirrors, the DMD acts as a
pure binary amplitude grating because of the gap between the micromirrors. For the
on-state micromirrors, the DMD behaves as a coupled amplitude and phase grating
because position-dependent phase change is formed because of an inclined configu-
ration [71,254]. For this reason, maximum diffraction efficiency may be achieved by
searching a possible “blaze” condition. Recently, Deng et al. [255] reported a high
single-order diffraction efficiency via optimizing the incident angle of the illumina-
tion light. However, the relatively large gap between the micromirrors still poses a
constraint on further improving overall DMD efficiency. To summarize, overall DMD
efficiency can be estimated by taking the above four factors into account. The total
photopic efficiency is approximately 60–70%, depending on the DMD pitch, tilt angle,
and f/# of the projection lens.
5.3f. LBS
For the LBS microdisplay, there is usually no projection lens in the AR system.
Therefore, the efficiency of LBS is determined by the laser light sources and the
MEMS scanner. Developing high-efficiency laser light sources [256,257] and high-
efficiency MEMS scanners [258] is critical for the next-generation low-power AR
glasses. For the laser sources, RGB laser diodes are employed to generate a color laser
beam. In particular, the efficiency of direct emission green diode lasers [259–261]
is crucial for long usage time of the AR glasses. It is noted that to assure a fast-
switching speed, the lasers still work in the subthreshold current when the pixel is
black, thus consuming additional power. In addition to laser sources, the scanning
process consumes electrical power as well. The MEMS mirror driver sends electrical
signals to deliver the accurate control of the MEMS mirrors. The MEMS mirror(s) can
be two separate 1D MEMS mirrors or one 2D MEMS mirror. The 2D MEMS mirror
can exhibit a lower power consumption than the two 1D MEMS mirrors. However,
the possible cross talk between the two axes is an issue for the high-resolution LBS
microdisplay. The methods to actuate the MEMS mirror include electrostatic actuation,
electromagnetic actuation, and piezoelectric actuation. Among these three actuation
methods, piezoelectric actuation exhibits the highest actuation forces and the highest
power efficiency [151]. Recently, Boni et al. [262] demonstrated the piezoelectric
MEMS mirrors working at the target FoV of 56°×32°(diagonal 65°). Their measured
power consumption was less than 20 mW. Lastly, there is a trade-off between the
light source efficiency and the resolution [263]. The light source can be on during the
entire pixel time (pixel duty cycle is 1) to maximize the light source efficiency, but the
scanner motion during the entire pixel time may blur the spot along the scan direction
and sacrifice the resolution.
832 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
5.4. High Dynamic Range
HDR displays have been constantly pursued and improved to deliver a better visual
experience than SDR displays since the early 2000s [264]. Dynamic range is deter-
mined by the peak brightness, contrast ratio, bit depth, and color gamut. The peak
brightness of at least 1000 cd/m2, contrast ratio over 200,000:1, and gray levels more
than 10 bits have been proposed as the HDR standard. Many efforts have been devoted
to HDR direct-view LCDs such as mini-LCD backlit TVs, desktop monitors, laptop
computers, and pads [4,265]. For near-eye displays, it is more challenging to achieve
HDR because of additional optical elements are involved [266]. For instance, in an
immersive VR headset, the brightness and contrast ratio are jointly determined by the
light engine and the imaging lens (Fresnel lens or “pancake” lens). For a see-through
AR, the contrast ratio is greatly reduced by the imaging lens, optical combiner, and
ambient light. It is challenging to achieve an ACR over 5:1 under strong day light
condition. Meanwhile, there is a strict constraint on the weight and volume of the
AR/VR devices. For this reason, some HDR methods suitable for large space displays
cannot be transferred directly to near-eye displays. Overall, HDR is a critical metric
in AR/VR and more endeavors are urgently needed to achieve HDR near-eye displays.
In the following, we focus on how light engines help achieve HDR near-eye displays
with an emphasis on novel designs and strategies.
5.4a. Dual Modulation Display
To improve the dynamic range of a display system, a dual modulation display system
is proposed. Figure 26(a) illustrates a simplified device configuration for a dual mod-
ulation display system, where the first modulation panel (subpanel) can be a LC panel
without color filters, DLP, LBS or mini-LED array, and the second modulation panel
Figure 26
(a) Simplified schematic of a dual modulation display system. (b) DLP-based HDR
display system. (c) Folded LBS-based HDR display system.
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 833
can be a LCD with color filters (main panel). The severity of visual artifacts mainly
depends on three factors: the dimming zone size of the first modulation panel (P1), the
distance between the dual panels, and the range of the acceptance cone for the near-eye
display (α). Compared with direct-view displays with a tiny viewing cone (α∼0.1°),
the viewing cone in VR display systems is several degrees, resulting in a greater cross
talk in the same panel space (d) between subpanel and main panel. For example, as
shown in Fig. 26(a), both zone A and zone B simultaneously illuminate the target pixel
in main panel and enter the pupil range. Therefore, reducing the dimming zone size of
the subpanel (P1) and space between the dual panels is critical to mitigate the visual
artifacts caused by dual-modulation display systems. Here, we analyze the pros and
cons of three pairs of advanced HDR near-eye displays: DLP microdisplay+LCD;
mini-LED array +LCD, and LBS microdisplay+LCD.
(1) Combination of DLP microdisplay (subpanel) and LCD (main panel). As shown
in Fig. 26(b), due to the bulky illumination system of the DLP microdisplay, the
distance (d) between the two panels is in the millimeter range, so relay optics
is used to reimage the intensity distribution on the subpanel to the main panel.
However, such a bulky display system is impractical for NED applications.
(2) Local dimming mini-LED array has been widely used as a subpanel for direct-
view HDR LCDs, and its compact system configuration is suitable for NED
applications. However, Tan et al. [267,268] pointed out that when the native
contrast ratio of a LC is 2000:1, the viewing cone of the dimming zone should
be less than 0.5°to eliminate the halo artifacts. The corresponding dimming zone
size is about 400 µm, and the required number of mini-LEDs is about 40,000,
which is much larger than advanced mini-LED LCDs with only a few thousand
local dimming zones in commercial projects.
(3) To achieve a dual-modulation display system with a high-resolution subpanel and
a compact design, in Fig. 26(c) we show a guided LBS projection system that Zhao
et al. [266] proposed as a subpanel, which eliminates the relay optics, and has a
compact form factor. The optical system consists of a projection unit, lightguide,
turning film, and a diffuser. The lightguide is a critical element to replace the
relay optics and enables a compact form factor via multiple reflections. Tilting
the system can further reduce the volume, and the inclined rays will be deflected
by a thin turning film composed of micro prisms. Owing to the magnification
of the projection system, the numerical aperture of the relayed image of primary
modulation is reduced, calling for a diffuser at the secondary modulation to achieve
an acceptable eyebox. The superior HDR images achieved by the proposed display
system are shown in Fig. 26(c). Overall, research into HDR NED applications is
still in its infancy. Many issues still deserve more study, such as halo artifacts
caused by an inadequate number of dimming zones, image degradation caused by
the cross talk between two panels, cross talk variation during pupil swimming,
and compact form factors.
5.4b. HDR Optics
The optics with limited MTF and stray light also cause image degradation in the
AR/VR display systems. For example, the folded optics (called pancake eyepiece) has
degraded image quality due to stray light (ghost image). To analyze the stray light, the
optical path of the pancake eyepiece is shown in Fig. 27(a). Compared with the ideal
situation shown in Fig. 3(b), which is composed of ideal CP light and linearly polarized
light, the polarization state of the light in an actual pancake eyepiece is composed of
elliptically polarized light. It is this undesired elliptically polarized light that causes
the straylight, which in turn decreases the image contrast of the VR system. For the
834 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Figure 27
(a) Simplified schematic of folded optics (pancake eyepiece) VR displays. The ray
path of (b) signal, (c) stray light (type 1), and (d) stray light (type 2). (e1) Ray path
and (e2) intensity distribution of signal, and stray light (type 1). The pixel position is
off axis.
display panel, a point source is used to present the light emitted from pixels, and the
circular polarizer is laminated on top of the point source to generate the CP light.
The point light source from the display panel passes through the pancake eyepiece
and becomes a collimated beam. Then, the human eye ideally images the signal beam
(Fig. 27(b)) to a point on the retina. The straylight can be distinguished by the time
the light travels through the pancake lens. Type 1 straylight refers to the stray light that
is not reflected by the RP but passes directly through it (Fig. 27(c)). Compared with
the signal beam (Fig. 27(b)) passing through the pancake lens thrice, the straylight
passes through the pancake lens only once. Therefore, it has less optical path in the
pancake lens, that is, it experiences a smaller optical power. Such an insufficient lens
power causes the divergent beam to hit the eye lens, and the divergent beam is focused
further away from the retina. In other words, at the retina, the straylight is not a sharp
focus point, but a circular illumination area (Fig. 27(c)). Type 2 straylight is reflected
twice by the RP and experiences a stronger lens power than the signal light path. As
a result, the ray path is focused in front of the retina (Fig. 27(d)). To reduce ghosting
and preserve the HDR of the display panel, tailoring the angular distribution of the
display panel is an effective solution. Figure 27(e) illustrates that the chief ray of the
emission cone of the signal beam and the stray light are different. Therefore, the effects
of stray light can be reduced if we can carefully modulate the radiation pattern of the
display to match the emission cone of the signal beam and avoid overlapping with the
emission cone of the ghost image.
5.4c. HDR LCOS
For the amplitude-modulating LCOS device, the contrast ratio is greatly affected by the
FFE. Chiang et al. investigated the contrast ratios of five LC modes at different elec-
trode slopes [137]. The contrast ratio is defined as the ratio of the average reflectance
at the bright pixel to that at the dark pixel. When all the pixels are turned on and off,
the VA mode exhibits a calculated contrast ratio as high as 3300:1 at the electrode
slope of 0.3. For most of displayed contents, it is rare to turn on/off all pixels. For the
dark/bright/dark pixel configuration, the 90°-MTN mode exhibits a higher contrast
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 835
ratio than the VA mode due to a weaker fringing field. In addition to the FFE, the con-
trast ratio of the PBS, and the f/# of the projection lens also affect the system contrast
ratio. Recently, Abeeluck et al. [166] presented an amplitude LCOS microdisplay with
about averaged 3000:1 measured by a lens with an f/# =2.5. In real AR applications,
the dynamic range of the outdoor scenes can possibly be as high as 14 orders of magni-
tude. It is challenging to achieve such a HDR just using a single LCOS panel. For this
reason, Xu and Hua [269] proposed a HDR head-mounted display (HMD) based on a
pair of LCOS panels, which is composed of a HDR image generator and the viewing
optics as shown in Fig. 28(a). In a near-eye display, the spacing between the LCOS
panels with the same spatial resolution should be as small as possible to avoid cross
talk and to improve the accuracy of dynamic range enhancement. Therefore, the relay
optics with a 1:1 magnification was introduced to optically overlay the modulation lay-
ers (the gap can be as small as a few micrometers) and thereby enabled pixel-by-pixel
modulation on dynamic range. More specifically, a double telecentric relay system was
adopted because it can provide uniform illumination, uniform optical magnification,
and uniform light efficiency. The LCOS1 was equipped with the LED source and the
wire grid film (WGF). The LED source served as the illumination for the HDR image
generator, and the WGF generated a high-efficiency linearly polarized light. The LED
source and WGF polarizer were removed from the LCOS2. To demonstrate the perfor-
mance of dual LCOS panels, a HDR image was first synthesized by capturing a HDR
scene with different camera exposure time. The HDR scene mainly consisted of a
desk lamp and two resolution targets. To display the synthesized HDR image using the
HDR-HMD, it was then re-rendered into two low-dynamic-range modulation images.
The two modulation images were geometrically corrected and then displayed by cor-
responding LCOS panels. In this way, the HDR image could be successfully displayed
by the HDR-HMD system. Figure 28(b) shows that the details and dynamic range in
the HDR scene are preserved. In contrast, they are lost in a low-dynamic-range HMD
based on one LCOS panel, as shown in Fig. 28(c). Although the proposed HDR HMD
can display HDR contents, it undoubtedly increases the bulkiness and cost of the AR
headset.
5.4d. HDR DLP
From a system perspective, several factors such as illumination angle, coating quality,
mirror gap, and numerical aperture play significant roles in determining the contrast
ratio of the DMD system [72]. Illumination angle greatly affects the distribution and
the amount of reflected and scattered light. In particular, as the mirrors are in the off
state, there is a larger probability for the light to hit the edges of the mirrors and mirror
vias, leading to more scattered light. Coating quality influences the reflected light
from all surfaces such as the surfaces of a lens and a prism. The reflected light may
travel through the projection lens by a certain optical path, reducing the contrast ratio.
Mirror gap determines the diffracted light that may be collected by the projection lens
as well. Finally, the numerical aperture of the projection system controls the amount
of collected light including stray light. Based on these factors, several strategies have
been proposed to increase the contrast ratio [270–274]. A straightforward method is
to use dark (non-reflecting) metal layers below each mirror. The reduced reflected
or scattered light helps boost the contrast ratio to exceed 1000:1. In addition to dark
metal layers, several other methods have also been proposed to improve the DMD’s
contrast ratio. Meuret [272] proposed two methods to improve the contrast ratio:
cutting off the cone of light and shifting the cones of light to larger angles. The
reference contrast ratio degraded by the scattering was 430:1, which was simulated by
the optical ray-tracing program ASAP. Truncating the light cones with the adjusted
aperture stop in the projection lens improved the full-on/full-off contrast ratio to 750:1.
For the second method, the full-on/full-off contrast ratio could be improved to 755:1
836 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Figure 28
(a) Schematic of a HDR HMD based on dual LCOS panels. (b) Re-rendered HDR
image displayed by the HDR-HMD. (c) Low dynamic range image displayed by a
low-dynamic-range HMD. (a)–(c) Reprinted with permission [269]©The Optical
Society.
by increasing the cones of illumination light by 3°with the well-adjusted asymmetric
aperture stop in the projection lens. If we combine any of the two methods with the
dark metal layer, a higher contrast ratio can be expected. Later, Pan and Wang [273]
proposed a novel prism design to increase the contrast ratio while maintaining the
optical efficiency. The prism was composed of three smaller prisms with the same
material. The first TIR surface was used for eliminating the interference between the
illumination beam and the reflected beam at the on state of the DMD. The image rays
directly travelled through the second TIR surface; however, the stray light would be
deflected away from the projection lens because of the TIR, thus greatly enhancing
the contrast ratio. Ding and Pan [274] designed a freeform surface lens to achieve a
high contrast ratio without reducing system efficiency. The freeform surface lens was
designed to eliminate the overlap effect between the on state and the flat state caused
by the anamorphic phenomenon.
The DMD itself is a binary device; therefore, the gray level is generated by sequen-
tially projecting bit-planes. Traditional PWM poses a constraint on the maximum
achievable bit depth due to the exponential relationship between the frame rate and
the bit depth. Intensity-modulated light can achieve both high frame rate and high bit
depth, but the maximum brightness is greatly sacrificed. Although multiple cascaded
DMDs can also increase the bit depth, its efficiency loss, sophisticated calibration,
and increased cost limit its widespread adoption. To achieve a high bit-depth DLP
microdisplay with high frame rate and high brightness, Chang et al. [174] proposed
a hybrid light modulation (HLM) by combining PWM and intensity-modulated light
as schematized in Fig. 29(a). For the least-significant bit-planes (0–3 in Fig. 29(a)),
intensity-modulated light encoded in n1bits is applied to alleviate increased exposure
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 837
Figure 29
(a) Schematic for realizing an 8-bit gray-scale image with HLM. (b) 8-bit projected
image with the traditional PWM method. (c) 16-bit projected image with the HLM
method. (a)–(c) Reprinted with permission from [174]. ©The Optical Society.
time from PWM; for more significant bit-planes (4–7 in Fig. 29(a)), PWM encoded in
n2bits is employed to mitigate the brightness losses from intensity-modulated light.
The total frame time is n1t+(2n2−1)t, and the maximum averaged brightness can be
calculated as
Lmax =L(2−n1+· · · +2−1)t+t(1+21+· · · +2n2−1)L
n1t+(2n2−1)t,(14)
where Lis the maximum intensity of light source used in the PWM, and tis the
minimum bit-plane exposure time. Owing to an exponential growth, 2n2is usually
much larger than n1, leading to Lmax ≈L. By carefully selecting n1, the HLM helps
achieve a 16-bit depth DLP microdisplay with adequate frame rate and high brightness.
A 16-bit image has 16 million times more RGB colors than an 8-bit image, thus
preserving more subtle details. Figures 29(b) and 29(c) show an 8-bit projected image
using the traditional PWM method and a 16-bit projected image using the HLM
method, respectively. The bushes and tree trunks marked in Fig. 29(c) are more visible,
demonstrating the advantage of high bit depth. Finally, by combing some strategies to
achieve high contrast ratio and the HLM, the DLP microdisplay can simultaneously
exhibit high contrast ratio, high bit depth, high frame rate, and high brightness.
5.4e. HDR LBS
The contrast ratio of a LBS microdisplay is mainly determined by the diffraction
limitation and scattered light in the optical system [275]. For a dark region surrounded
by white regions, the luminance of the dark region is not zero due to the diffraction
spread from all the white pixels. A larger mirror size may increase the contrast ratio
838 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
by reducing the diffraction, but the dynamic mirror deformation may be increased.
Therefore, an optimal mirror size should be searched. The gray level in the LBS is
realized by varying the luminance of each pixel. Laser diodes can be directly modulated
by the driving current to generate the current-dependent luminance. It is noted that
the pixel time at the center and edge of the display is different due to the sinusoidal
movement in the horizontal/vertical scan. For central pixels with a shorter pixel time,
a higher laser modulation rate is desired, especially for a high-resolution microdisplay.
For example, Holmstrom et al. [151] calculated the maximum required pixel clock
frequency for supporting HD1080 (1920 ×1080), which was about 260MHz in the
bidirectional raster scanning architecture and 60Hz refresh rate. Although external
modulators such as acousto-optic modulators may realize such a high pixel rate, the
cost, and the size of the LBS will be increased. Therefore, high-speed laser drivers
are still essential, and the maximum achievable bit depth is limited by the laser driver
and laser diodes. Recently, Petrak et al. [276] presented a LBS microdisplay with
maximum supported 10-bit depth and 200 MHz pixel clock frequency. Wide color
gamut is another advantage of the LBS [277,278]. The color gamut of the LBS is
determined by the central wavelengths of three laser sources and the full width at
half maximum of their emission spectra. By employing a green laser source with the
central wavelength of 530 nm, its color gamut can be much larger than the sRGB. The
green laser sources were realized by frequency doubled infrared laser diodes or direct
emission green laser diodes. The latter truly enables a miniaturized LBS microdisplay
and accelerates the development of the full-color compact LBS light engine for AR
applications. For the LCOS and the DMD systems, the backlight source can be LEDs
or lasers. If the laser backlight is adopted, the color gamut of the LCOS or the DMD
can be comparable with that of the LBS.
5.4f. HDR OLED and µLED
OLED and µLED emissive displays are current-driven devices, so a pixel circuit
is required to generate the desired current to achieve a certain brightness. In the
pixel circuit, data voltages that correspond to the desired gray levels determine the
magnitude of the driving currents, exhibiting the different luminance. To enhance the
quality of HDR, a wider range of data voltage for maximizing display luminance is
favorable. On the other hand, an extremely low luminance interval for distinguishing
the gray scale of a black image is also beneficial to HDR. Although OLED, mini-LED,
and µLED can provide a higher quality of black images, their pixel circuits cannot
ensure these devices to be turned off completely. Because each pixel circuit needs to
perform compensation in addition to data input and emission operations in each frame
to improve the uniformity of the displayed images, a current may flow through these
emissive devices during the compensation and data input periods, causing flicker to
reduce the dynamic range. Hence, to achieve HDR, the developed pixel circuits should
provide no leakage current flowing through such emissive devices except during the
emission period when a black image is displayed. Furthermore, after the images are
illuminated, there are some charges stored at the emissive devices. Nevertheless, these
charges may slightly activate these devices, degrading the quality of the displayed
pure black images. To resolve this issue for a higher dynamic range, the charges must
be released before each emission. Thus, pixel circuits require to reset the anode or
cathode of these emissive devices before the generation of the driving currents. By
designing pixel circuits to prevent the emissive devices from flickering, a better quality
of HDR can be attained.
5.5. Compactness
The form factor of light engines is much more demanding for AR than VR. Presently,
several microdisplay technologies, such as µLED, µOLED, LCOS, DLP, and LBS have
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 839
Figure 30
A qualitative comparison of volume size of each light engine in AR at each FoV. Data
points of µLED [279], LCOS [280], and LBS [281] are marked out.
been developed for AR. Each technology has its own pros and cons. Figure 30 shows a
generic comparison between the volume and the FoV in panel-based displays, and data
points of state-of-the-art µLED [279], LCOS [280], and LBS [281] are marked out.
Among them, quantum-dot based µLED and µOLED displays are both self-emissive
and can achieve very high pixel density and full color on a single panel, which seems
ideal for compact light engines. However, the trade-off between brightness and lifetime
remains to be overcome for µOLED to extends its application to AR because of the high
brightness requirement [282]. µLED aims to preserve all the advantages of µOLED
and mitigate the brightness-lifetime issue, but mass production is still in the infancy
stage [283]. Although these two panel-based display technologies still have limited
market penetration in present AR headsets, they remain strong contenders because the
above-mentioned issues are gradually being overcome. LCOS and DLP are both non-
emissive panel displays and have reached a matured stage for mass production after
decades of investments in standard CMOS technology [170], but an extra illumination
system usually leads to a larger form factor. In the illumination system, the light from an
external illuminating source (e.g., LEDs) usually needs to be homogenized by passing
through a pair of fly-eye lenses or a rod integrator before reaching the LCOS/DLP
panel. In a traditional LCOS display [132], a PBS functions as both a polarizer and an
analyzer. In the telecentric DLP display [270], a light separator (prism) is employed to
direct the uncontrolled beam away from the projection lens. To further reduce the form
factor, these bulky optics should be shrunk or even removed. The transmissive LCD,
like LCOS, relies on liquid crystal for amplitude modulation, but it uses a backlight
illumination. At first glance, the transmissive property seems intriguing to make a
simpler optical design, but the lower fill factor (∼20% due to black matrices) limits
its pixel density to be ∼2000–3000 PPI, which is lower than its reflective counterpart
(>4000 PPI). On the other hand, LBS consisting of a tiny laser module and MEMS
mirrors, has a very attractive form factor and is still pushing its limit, but its frame rate
and scanning uniformity remain to be improved. Next, we discuss recent progress in
reducing the form factor of LCOS and LBS displays.
A conventional LCOS display consists of a color combiner, a homogenizer, a PBS, and
a LCOS panel, as shown in Fig. 31(a). This configuration is referred to as free-space-lit
LCOS, because light mainly propagates in free space. Similar to the edge-lit backlight
840 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Figure 31
(a) Free-space color-sequential LCOS design with an X-cube, a homogenizer, and a
PBS. (b) Backlit LCOS with collimation optics replaced by a light guide. (c) Front-lit
LCOS where the PBS is removed. (d) Light propagation process in a front-lit LCOS.
(e) New slim LCOS with an enlarged FoV. HG, homogenizer; PL, projection light;
LG, light guide; FLP, front-lit plate.
[284] for a direct-view LCD, a back-lit LCOS [285] was proposed to reduce to from
factor of the illumination system, as plotted in Fig. 31(b). The whole illumination
system could be replaced by a lightguide plate with LEDs on the edge. The lightguide
mixes the input light to attain uniform illumination. Still, a PBS cube, which is about
1 cm3, is required to polarize and direct the illumination toward the LCOS panel, and
then analyze the spatially modulated light from the LCOS panel. If the LCOS panel is
directly illuminated from top, then the bulky PBS can be replaced by a planar polarizer.
Later, a so-called front-lit LCOS [286,287] was introduced by placing a flat plate in
front of LCOS to eliminate both illumination system and PBS, as shown in Fig. 31(c).
The idea of “front-lit” could trace back to the reflective LCD direct-view displays
[288] in mid-1990s to solve the readability issue in low-light condition. The flat plate
is only 1 mm thick in total and the detailed structure of front-lit LCOS is illustrated
in Fig. 31(d). The front-lit plate contains a lightguide on top of the LCOS module.
The LED is located on the edge of the lightguide, and a polarizer is inserted between
the LED and the lightguide. On top of the lightguide, a RP is placed as an analyzer.
Assuming the polarizer attached to the LED transmits s-polarized light, the RP should
reflect s-wave and let p-wave pass. The s-wave propagates in the lightguide by TIR
process or reflecting from optical elements such as dielectric mirrors. When the TIR
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 841
condition is not satisfied, the ray escapes from the lightguide and is modulated and
reflected by the LCOS panel. The p-wave in the reflected ray passes the RP and forms
an image. The downside of this design is that the viewing cone is limited to about 30°.
In the original design [286], the LCOS is operating in color-filter mode illuminated by
a white LED. If operating in color sequential mode [287], an additional hollow rod is
used for RGB color mixing by means of multiple reflections, but the overall thickness
of the front-lit plate is still slim (1.5 mm).
Recently, a new LCOS architecture [289,290] specially designed for large-FoV AR
applications, radically removes the light guide in the front-lit design, as shown in
Fig. 31(e). In this design, the LCOS panel modulates a CP light. The RGB LEDs
are placed side-by-side in the light source region. A RP is attached to LEDs for
polarization recycling. A linear polarizer (LP) attached with a QWP converts the
incident light to LCP state. A lens is placed below QWP and functions as a Fourier
transform of the LED light. The LCOS panel at the focal plane of the lens receives the
angular spectrum of light from LEDs, where each pixel corresponds to a plane wave
component at a different propagation angle. When the light is reflected, the spatial
pixel is transformed to far field after passing through the lens for a second time, which
is exactly what a waveguide combiner needs. A double circular polarizer, a linear
polarizer sandwiched between two crossed QWPs, is inserted between the lens and
the LCOS, functioning as a polarizer and analyzer. The orientation of the slow axis
of QWP1 and QWP3 should be aligned, whereas that of QWP2 is orthogonal to the
other two. In the on state, the LCOS panel modulates the incident light so that the
polarization state of the reflected light remains unchanged, which is still LCP and
passes through two linear polarizers without any loss ideally. The circular polarization
helps suppress stray light caused by Fresnel reflection which will flip the handedness
of incident light. For example, the stray light is RCP and will be absorbed by the LP1.
Such a compact LCOS design shortens the distance between the collimation lens to the
waveguide combiner and therefore enlarges the FoV. With an improved LCOS design,
some commercial products with volume close to 1 cm3have been launched [280].
An LBS display can be generally divided into two parts including a laser illumination
module and a sets of MEMS mirrors. Unlike panel-based display systems, the form
factor of a LBS display remains unchanged when increasing the pixel density and FoV
[277], because there is no real object plane and pixel information is encoded in the
angular domain. Conventionally, in the laser module, the separated RGB laser beams
are collimated and then combined before sending it to the scanning MEMS mirror.
Figure 32
(a) Laser module with separate hermetically sealed RGB laser diodes, separate col-
limation lens, and a dichroic beam combiner. (b) RGB laser diodes are integrated in
one package and share collimation optics without a beam combiner.
842 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
Table 1. Comparison of Different Light Engines for AR/VR
VR Light
Engine
Resolution
Density
Frame Rate Efficiency Contrast Ratio Form Factora
LCD Low Max:
∼2000 PPI
(with subpixel
rendering)
Medium
(90–120 Hz;
limited by LC
response time
and driving
scan time)
Low (can be
tripled by field
sequential
colors)
Low ∼500:1
(can be
enhanced by
local dimming)
Medium
µOLED*
display
Medium
WOLED:
∼3000 PPI
RGB OLED:
∼2645 PPI
Medium
(limited by
scan and
compensation
time)
Medium (can
be improved by
tandem
structure)
High (>104:1) Small (self-
emissive)
AR Light
Engine
Resolution
Density
Frame Rate Efficiency Contrast
Ratio
Form
Factor*
LCOS
display
High 8500 PPI
(∼3µm)
Medium
(≥720 Hz)
Medium (can
be improved by
laser backlight)
Medium
(∼500:1)
Small (new
designs
without PBS)
DLP
display
Medium Texas
instruments:
4700 PPI
(5.4 µm)
High
(improved by
hybrid
modulation)
Medium
(without
polarization
loss)
Medium (a few
thousand to
one)
Medium (TIR
prism)
µLED
display
Medium
PlayNitride:
4536 PPI (color
conversion)
JBD: ∼10,000
PPI (single
color)
Medium
(similar to
µOLED
display)
Medium (red is
worse: EQE
∼1%)
High (>104:1) Small (self-
emissive,
single panel)
LBS
display
Low (HoloLens
2; worse image
quality)
Low (trade-off
with resolution
density)
High (simple
optical
structure: only
mirror loss)
High (>104:1) Small
(independent
of FoV and
pixel density)
aµOLED can be used on both AR and VR displays.
Combiners can be a simple X-cube, or a series of mirrors or prisms with dichroic
coatings allowing wavelength-selective reflection and transmission of collimated RGB
laser beams, as shown in Fig. 32(a). A combiner system occupies the valuable space
in light engine and requires precise assembly and alignment processes. A new design
[96] is proposed to eliminate any additional combining optical elements by correcting
the angular offset of RGB laser beams with software compensation in the time domain,
as shown in Fig. 32(b). Two common lenses are used for collimating the RGB laser
beams, and the non-coaxial beams intersect at the MEMS mirror plane with a tilt
angle. The three laser diodes are shifted accordingly in the propagation direction
to compensate for the wavelength-dependent back focal length of lenses. A widely
used MEMS mirror configuration is to cascade two 1D MEMS mirrors adopting
raster scanning method. The first MEMS mirror has a small diameter and is driven in
resonance frequency for horizontal scan. The 1D picture is then sent to a much larger
second MEMS mirror for linear scan in the vertical axis. The advantages of this design
are the wider angular swing space and faster scan speed. However, the drawback is
a larger form factor because two MEMS mirrors need to be aligned and the driving
electronics are more complex. The size of the LBS light engine can be dramatically
reduced by using a single 2D MEMS mirror, which can scan in both axes, as shown
in Fig. 8(d). The trade-offs are the possible cross talk between two axes and the lower
frame rate. A LBS with volume less than 1cm3has been demonstrated [281].
6. DISCUSSION AND CONCLUSION
In this paper, we have first reviewed the optical configuration of AR/VR displays, and
the requirements of our HVS. Then, we described the operation principle and highlight
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 843
the pros and cons of six advanced light engines by five human-centric display metrics.
Here, we briefly summarize the pros and cons of each light engine, as listed in Table 1.
These analyses on the pros and cons of each light engine can serve as a basis for
further developing new light engines to enhance the optical performance of AR/VR
displays.
(1) Transmissive LCDs: The aperture ratio, which is governed by the TFT size and
black matrices, determines the maximum resolution density of the active-matrix
LCD panel for VR headsets. In 2022, a 2016 PPI transmissive LCD prototype has
been demonstrated by Innolux. To further increase the resolution density, single-
crystal silicon transistors can be utilized, but the FFE needs to be considered as
well. Furthermore, the LC response time is about 3 ms, and its optical efficiency
is lower than that of LCOS microdisplays due to its larger cell gap and smaller
aperture ratio. However, its relatively mature and cheap manufacturing still offers
a great advantage for VR applications.
(2) LCOS microdisplay: Its pixel pitch can be as small as 3 µm. Its fast response
time enables field sequential color operation, so that the pixel pitch remains at
3µm for a full-color pixel. It is a mature high-resolution density microdisplay.
However, the demand for higher resolution density (<1µm) for phase modulation
applications and for reducing panel size has not stopped. To achieve these goals,
the bottleneck of severe FFE needs to be overcome. In addition, a novel front-lit
illumination system and PCS have been developed to realize efficient and compact
LCOS microdisplays.
(3) DLP microdisplay: It exhibits the highest frame rate among all microdisplay
technologies considered because of its fast electromechanical response of the
micromirrors, thus supporting multiplane displays. By applying HLM methods
and prism designs, a single DMD can simultaneously achieve high contrast ratio
and high bit depth. However, the optics of DLP is still bulky and difficult to further
improve. It shows the worst compactness of all the six microdisplay technologies
discussed here.
(4) OLED microdisplays: High current efficiency is critical for OLED devices to
achieve a high brightness with a reasonable lifetime. Tandem WOLED devices
and patterned RGB OLED devices have been developed to achieve this goal.
The RGB OLEDs without color filters shows high optical efficiency and better
color performance. To achieving high-resolution-density RGB OLED devices,
the development of organic material patterning methods has recently flour-
ished. Tandem WOLED devices generate additional carrier pairs for higher
current efficiency, but the downside is its higher driving voltage. Advanced
research has focused on reducing the driving voltage, balancing the carrier
recombination for white light emission, and optimizing the cavity design to
improve the optical efficiency of tandem WOLED devices. In addition, the
frame rate of OLED microdisplays is mainly determined by the compensa-
tion method of the driving circuit and similar restrictions also apply to µLED
microdisplays.
(5) µLED microdisplays: Although color-conversion µLEDs, vertically stacked
µLEDs, and multi-color MQW µLEDs have been proposed and demonstrated
as prototypes, there is still no mature fabrication process for full-color high-
resolution-density microdisplays. Combining three separate RGB µLED displays
with an X-prism is promising, but its volume increases noticeably. In addition,
advanced LED chip designs have been developed to mitigate the size effects and
narrow the emission cones, thereby increasing the optical efficiency for projection
displays. Unlike traditional LEDs, where green color is the least efficient, the least
efficient µLEDs are red.
844 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
(6) LBS microdisplay: There is a trade-off between resolution density and frame rate.
Therefore, to maintain a decent resolution density, the frame rate is limited to
about 60 Hz. However, LBS microdisplays offer the advantage of providing high
optical coupling in projection systems and are more suitable for some AR display
systems such as Maxwellian displays.
FUNDING
Nichia Corporation; Meta.
ACKNOWLEDGMENT
The authors are indebted to Dr. Ming-Yang Deng for useful discussions.
DISCLOSURES
The authors declare no conflicts of interest.
DATA AVAILABILITY
All data needed to evaluate the conclusions in the paper are present in the paper.
Additional data related to this paper may be requested from the authors.
REFERENCES
1. M. Schadt, “Milestone in the history of field-effect liquid crystal displays and
materials,” Jpn. J. Appl. Phys. 48, 03B001 (2009).
2. C. W. Tang and S. A. VanSlyke, “Organic electroluminescent diodes,” Appl.
Phys. Lett. 51, 913–915 (1987).
3. J. Y. Lin and H. X. Jiang, “Development of microLED,” Appl. Phys. Lett. 116,
100502 (2020).
4. E. L. Hsiang, Z. Yang, Q. Yang, Y. F. Lan, and S. T. Wu, “Prospects and challenges
of mini-LED, OLED, and micro-LED displays,” J. Soc. Inf. Disp. 29, 446–465
(2021).
5. Y. Huang, E. L. Hsiang, M. Y. Deng, and S. T. Wu, “Mini-LED, Micro-LED and
OLED displays: present status and future perspectives,” Light: Sci. Appl. 9, 105
(2020).
6. H. W. Chen, J. H. Lee, B. Y. Lin, S. Chen, and S. T. Wu, “Liquid crystal display
and organic light-emitting diode display: present status and future perspectives,”
Light: Sci. Appl. 7, 17168 (2018).
7. O. Cakmakci and J. Rolland, “Head-worn displays: a review,” J. Disp. Technol.
2, 199–216 (2006).
8. S. Choi, K. Jung, and S. D. Noh, “Virtual reality applications in manufacturing
industries: past research, present findings, and future directions,” Concurr. Eng.
23, 40–63 (2015).
9. X. Li, W. Yi, H. L. Chi, X. Wang, and A. P. C. Chan, “A critical review of virtual
and augmented reality (VR/AR) applications in construction safety,” Autom.
Constr. 86, 150–162 (2018).
10. K. Yin, E. L. Hsiang, J. Zou, Y. Li, Z. Yang, Q. Yang, P. C. Lai, C. L. Lin, and S.
T. Wu, “Advanced liquid crystal devices for augmented reality and virtual reality
displays: principles and applications,” Light: Sci. Appl. 11, 161 (2022).
11. J. Xiong, E. L. Hsiang, Z. He, T. Zhan, and S. T. Wu, “Augmented reality and
virtual reality displays: emerging technologies and future perspectives,” Light:
Sci. Appl. 10, 216 (2021).
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 845
12. D. Cheng, Q. Wang, Y. Liu, H. Chen, D. Ni, X. Wang, C. Yao, Q. Hou, W. Hou,
G. Luo, and Y. Wang, “Design and manufacture AR head-mounted displays: a
review and outlook,” Light: Adv. Manuf. 2, 336–369 (2021).
13. B. C. Kress, Optical Architectures for Augmented-, Virtual-, and Mixed-Reality
Headsets (SPIE, 2020).
14. C. Chang, K. Bang, G. Wetzstein, B. Lee, and L. Gao, “Toward the next-
generation VR/AR optics: a review of holographic near-eye displays from a
human-centric perspective,” Optica 7, 1563–1578 (2020).
15. Q. Yang, Z. Yang, Y. F. Lan, and S. T. Wu, “Low-diffraction transparent micro
light-emitting diode displays with optimized pixel structure,” J. Soc. Inf. Disp.
30, 395–403 (2022).
16. H. Kim, Y. T. Kwon, H. R. Lim, J. H. Kim, Y. S. Kim, and W. H. Yeo, “Recent
advances in wearable sensors and integrated functional devices for virtual and
augmented reality applications,” Adv. Funct. Mater. 31, 2005692 (2021).
17. H. G. Kim, H. T. Lim, and Y. M. Ro, “Deep virtual reality image quality assess-
ment with human perception guider for omnidirectional image,” IEEE Trans.
Circuits Syst. Video Technol. 30, 917–928 (2020).
18. M. R. Miller, H. Jun, F. Herrera, J. Y. Villa, G. Welch, and J. N. Bailenson,
“Social interaction in augmented reality,” PLoS One 14, e0216290 (2019).
19. D. Shin, “How does immersion work in augmented reality games? a user-centric
view of immersion and engagement,” Inf. Commun. Soc 22, 1212–1229 (2019).
20. R. T. Azuma, “Making augmented reality a reality,” in Imaging and Applied
Optics 2017 (3D, AIO, COSI, IS, MATH, PcAOP), OSA Technical Digest (Online)
(Optica Publishing Group, 2017), p. JTu1F.1.
21. J. Kim, Y. Jeong, M. Stengel, K. Akşit, R. Albert, B. Boudaoud, T. Greer, J.
Kim, W. Lopes, Z. Majercik, P. Shirley, J. Spjut, M. McGuire, and D. Luebke,
“Foveated AR: dynamically-foveated augmented reality display,” ACM Trans.
Graph. 38, 1 (2019).
22. T. Hirzle, M. Cordts, E. Rukzio, J. Gugenheimer, and A. Bulling, “A critical
assessment of the use of ssq as a measure of general discomfort in vr head-
mounted displays,” in Proceedings of the 2021 CHI Conference on Human
Factors in Computing Systems, CHI ‘21 (Association for Computing Machinery,
2021), pp. 530:1–530:14.
23. N. A. Dodgson, “Variation and extrema of human interpupillary distance,” Proc.
SPIE 5291, 36–46 (2004).
24. J. Schwiegerling, Field Guide to Visual and Ophthalmic Optics (SPIE, 2004).
25. J. W. Goodman, Introduction to Fourier Optics, 3rd ed. (Roberts and Company
Publishers, 2004).
26. J. min Cho, Y. do Kim, S. H. Jung, H. Shin, and T. Kim, “78-4: screen door effect
mitigation and its quantitative evaluation in VR display,” Dig. Tech. Pap. - Soc.
Inf. Disp. Int. Symp. 48, 1154–1156 (2017).
27. S. Reichelt, R. Häussler, G. Fütterer, and N. Leister, “Depth cues in human visual
perception and their realization in 3D displays,” Proc. SPIE 7690, 76900B (2010).
28. H. Hua, “Enabling focus cues in head-mounted displays,” Proc. IEEE 105,
805–824 (2017).
29. K. J. MacKenzie, D. M. Hoffman, and S. J. Watt, “Accommodation to multiple-
focal-plane displays: implications for improving stereoscopic displays and for
accommodation control,” J. Vis. 10(8), 22 (2010).
30. R. Zabels, K. Osmanis, M. Narels, U. Gertners, A. Ozols, K. R¯
utenbergs,
and I. Osmanis, “AR displays: next-generation technologies to solve the
vergence–accommodation conflict,” Appl. Sci. 9, 3147 (2019).
846 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
31. D. M. Hoffman, A. R. Girshick, K. Akeley, and M. S. Banks, “Ver-
gence–accommodation conflicts hinder visual performance and cause visual
fatigue,” J. Vis. 8(3), 33 (2008).
32. G. Kramida, “Resolving the vergence-accommodation conflict in head-mounted
displays,” IEEE Trans. Vis. Comput. Graphics 22, 1912–1931 (2016).
33. H. Huang and H. Hua, “High-performance integral-imaging-based light field
augmented reality display using freeform optics,” Opt. Express 26, 17578–17590
(2018).
34. X. Wang and H. Hua, “Depth-enhanced head-mounted light field displays based
on integral imaging,” Opt. Lett. 46, 985–988 (2021).
35. A. Maimone, A. Georgiou, and J. S. Kollin, “Holographic near-eye displays for
virtual and augmented reality,” ACM Trans. Graph. 36, 1–16 (2017).
36. S. B. Kim and J. H. Park, “Optical see-through Maxwellian near-to-eye display
with an enlarged eyebox,” Opt. Lett. 43, 767–770 (2018).
37. T. Zhan, J. Xiong, J. Zou, and S. T. Wu, “Multifocal displays: review and
prospect,” Photonis X 1, 10 (2020).
38. J. Miseli, “7.3: motion artifacts,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 35,
86–89 (2004).
39. B. Shneiderman, “Response time and display rate in human performance with
computers,” ACM Comput. Surv. 16, 265–285 (1984).
40. H. Pan, X. F. Feng, and S. Daly, “LCD motion blur modeling and analysis,” in
IEEE International Conference on Image Processing 2005 (2005).
41. J. Zhao, R. S. Allison, M. Vinnikov, and S. Jennings, “Estimating the motion-
to-photon latency in head mounted displays,” in 2017 IEEE Virtual Reality (VR)
(2017), pp. 313–314.
42. P. Lincoln, A. Blate, M. Singh, T. Whitted, A. State, A. Lastra, and H. Fuchs,
“From motion to photons in 80 microseconds: towards minimal latency for virtual
and augmented reality,” IEEE Trans. Vis. Comput. Graphics 22, 1367–1376
(2016).
43. H. Seetzen, W. Heidrich, W. Stuerzlinger, G. Ward, L. Whitehead, M. Trentacoste,
A. Ghosh, and A. Vorozcovs, “High dynamic range display systems,” ACM Trans.
Graph. 23, 760–768 (2004).
44. P. Lincoln, A. Blate, M. Singh, A. State, M. C. Whitton, T. Whitted, and H.
Fuchs, “Scene-adaptive high dynamic range display for low latency augmented
reality,” in Proceedings of the 21st ACM SIGGRAPH Symposium on Interactive
3D Graphics and Games, I3D ‘17 (Association for Computing Machinery, 2017),
pp. 15:1–15:7.
45. Y. H. Lee, T. Zhan, and S. T. Wu, “Prospects and challenges in augmented reality
displays,” Virtual Real. Intell. Hardw. 1, 10–20 (2019).
46. S. A. Cholewiak, Z. Başgöze, O. Cakmakci, D. M. Hoffman, and E. A. Cooper,
“A perceptual eyebox for near-eye displays,” Opt. Express 28, 38008–38028
(2020).
47. E. Hecht, Optics (Pearson, 2017).
48. T. Lin, T. Lin, T. Lin, T. Zhan, T. Zhan, J. Zou, F. Fan, and S. T. Wu, “Maxwellian
near-eye display with an expanded eyebox,” Opt. Express 28, 38616–38625
(2020).
49. J. Xiong, Y. Li, K. Li, and S. T. Wu, “Aberration-free pupil steerable Maxwellian
display for augmented reality with cholesteric liquid crystal holographic lenses,”
Opt. Lett. 46, 1760–1763 (2021).
50. T. Zhan, J. Zou, J. Xiong, X. Liu, H. Chen, J. Yang, S. Liu, Y. Dong, and S. T. Wu,
“Practical chromatic aberration correction in virtual reality displays enabled by
cost-effective ultra-broadband liquid crystal polymer lenses,” Adv. Opt. Mater.
8, 1901360 (2020).
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 847
51. B. A. Narasimhan, “Ultra-compact pancake optics based on ThinEyes super-
resolution technology for virtual reality headsets,” Proc. SPIE 10676, 106761G
(2018).
52. A. Maimone and J. Wang, “Holographic optics for thin and lightweight virtual
reality,” ACM Trans. Graph. 39, 67 (2020).
53. Y. Li, T. Zhan, Z. Yang, C. Xu, P. L. LiKamWa, K. Li, and S. T. Wu, “Broadband
cholesteric liquid crystal lens for chromatic aberration correction in catadioptric
virtual reality optics,” Opt. Express 29, 6011–6020 (2021).
54. K. Pulli, “11-2: invited paper: Meta 2: immersive optical-see-through augmented
reality,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 48, 132–133 (2017).
55. B. Lee, C. Yoo, J. Jeong, B. Lee, and K. Bang, “Key issues and technologies for
AR/VR head-mounted displays,” Proc. SPIE 11304, 1130402 (2020).
56. G. Westheimer, “The Maxwellian view,” Vision Res. 6, 669–682 (1966).
57. C. Jang, K. Bang, S. Moon, J. Kim, S. Lee, and B. Lee, “Retinal 3D: augmented
reality near-eye display via pupil-tracked light field projection on retina,” ACM
Trans. Graph. 36, 1–13 (2017).
58. D. Cheng, Y. Wang, H. Hua, and M. M. Talha, “Design of an optical see-through
head-mounted display with a low f-number and large field of view using a
freeform prism,” Appl. Opt. 48, 2655–2668 (2009).
59. D. K. Yang and S. T. Wu, Fundamentals of Liquid Crystal Devices (John Wiley
& Sons, 2006).
60. S. H. Lee, S. L. Lee, and H. Y. Kim, “Electro-optic characteristics and switching
principle of a nematic liquid crystal cell controlled by fringe-field switching,”
Appl. Phys. Lett. 73, 2881–2883 (1998).
61. M. F. Schiekel and K. Fahrenschon, “Deformation of nematic liquid crystals with
vertical orientation in electrical fields,” Appl. Phys. Lett. 19, 391–393 (1971).
62. D. H. Kim, Y. J. Lim, D. E. Kim, H. Ren, S. H. Ahn, and S. H. Lee, “Past,
present, and future of fringe-field switching-liquid crystal display,” J. Inf. Disp.
15, 99–106 (2014).
63. J. J. Lyu, J. Sohn, H. Y. Kim, and S. H. Lee, “Recent trends on patterned vertical
alignment (PVA) and fringe-field switching (FFS) liquid crystal displays for liquid
crystal television applications,” J. Disp. Technol. 3, 404–412 (2007).
64. S. S. Kim, B. H. You, J. H. Cho, D. G. Kim, B. H. Berkeley, and N. D. Kim, “An
82-in. ultra-definition 120-Hz LCD TV using new driving scheme and advanced
super PVA technology,” J. Soc. Inf. Disp. 17, 71–78 (2009).
65. M. N. Ernstoff, A. M. Leupp, M. J. Little, and H. T. Peterson, “Liquid crystal
pictorial display,” in 1973 International Electron Devices Meeting (IEEE, 1973),
pp. 548–551.
66. N. Collings, W. A. Crossland, P. J. Ayliffe, D. G. Vass, and I. Underwood,
“Evolutionary development of advanced liquid crystal spatial light modulators,”
Appl. Opt. 28, 4740–4747 (1989).
67. S. T. Wu and C. S. Wu, “Mixed-mode twisted nematic liquid crystal cells for
reflective displays,” Appl. Phys. Lett. 68, 1455–1457 (1996).
68. D. Dudley, W. M. Duncan, and J. Slaughter, “Emerging digital micromirror
device (DMD) applications,” Proc. SPIE 4985, 14–25 (2003).
69. L. J. Hornbeck, “Digital light processing for high-brightness high-resolution
applications,” Proc. SPIE 3013, 27–40 (1997).
70. Texas Instrument, “Introduction to ±12 degree orthogonal digital micromirror
devices (DMDs),” https://www.ti.com/lit/pdf/dlpa008b.
71. S. Scholes, R. Kara, J. Pinnell, V. Rodríguez-Fajardo, and A. Forbes, “Structured
light with digital micromirror devices: a guide to best practice,” Opt. Eng. 59, 1
(2019).
72. Texas Instruments, “DLP system optics,” https://www.ti.com/lit/pdf/dlpa022.
848 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
73. D. Y. Kondakov, “Characterization of triplet-triplet annihilation in organic light-
emitting diodes based on anthracene derivatives,” J. Appl. Phys. 102, 114504
(2007).
74. M. A. Baldo, S. Lamansky, P. E. Burrows, M. E. Thompson, and S. R.
Forrest, “Very high-efficiency green organic light-emitting devices based on
electrophosphorescence,” Appl. Phys. Lett. 75, 4–6 (1999).
75. H. Uoyama, K. Goushi, K. Shizu, H. Nomura, and C. Adachi, “Highly efficient
organic light-emitting diodes from delayed fluorescence,” Nature 492, 234–238
(2012).
76. Y. S. Park, S. Lee, K. H. Kim, S. Y. Kim, J. H. Lee, and J. J. Kim, “Exciplex-
forming co-host for organic light-emitting diodes with ultimate efficiency,” Adv.
Funct. Mater. 23, 4914–4920 (2013).
77. A. Salehi, X. Fu, D. H. Shin, and F. So, “Recent advances in OLED optical
design,” Adv. Funct. Mater. 29, 1808803 (2019).
78. W. K. Lee, Y. T. Chen, S. W. Wen, P. H. Liao, M. C. Lee, T. S. Hsu, Y. J. Chen, G.
D. Su, H. Y. Lin, C. C. Chen, W. Y. Lin, L. Xu, G. Yu, B. L. Kwak, R. J. Visser,
and C. C. Wu, “Three-dimensional pixel configurations for optical outcoupling
of OLED displays—optical simulation,” J. Soc. Inf. Disp. 27, 273–284 (2019).
79. S. L. Lee, C. C. Cheng, C. J. Liu, C. N. Yeh, and Y. C. Lin, “9.4-inch 228-ppi
flexible micro-LED display,” J. Soc. Inf. Disp. 29, 360–369 (2021).
80. J. Bae, Y. Shin, H. Yoo, Y. Choi, J. Lim, D. Jeon, I. Kim, M. Han, and S. Lee,
“Quantum dot-integrated GaN light-emitting diodes with resolution beyond the
retinal limit,” Nat. Commun. 13, 1862 (2022).
81. K. Parikh, J. Zhuang, K. Pallister, J. Jiang, and M. Smith, “40-1: invited paper:
next generation virtual reality displays: challenges and opportunities,” Dig. Tech.
Pap. - Soc. Inf. Disp. Int. Symp. 49, 502–505 (2018).
82. K. Kim, C. Kim, J. K. Park, D. H. Kim, and K. Jung, “75-4: invited paper: FMM
materials and manufacturing process: review of the technical issues,” Dig. Tech.
Pap. - Soc. Inf. Disp. Int. Symp. 49, 1011–1013 (2018).
83. Y. Onoyama, J. Yamashita, H. Kitagawa, E. Hasegawa, A. Makita, S. Yokoyama,
R. Asaki, E. Kambe, T. Nishimura, J. Yamada, K. Uchino, S. Makimura, Y.
Shiraishi, and T. Urabes, “70.4L: late-news paper: 0.5-inch XGA Micro-OLED
display on a silicon backplane with high-definition technologies,” Dig. Tech. Pap.
- Soc. Inf. Disp. Int. Symp. 43, 950–953 (2012).
84. J. Hamer, M. Kondakova, J. Spindler, R. Cupello, S. Hamer, M. Andre, S.
McClurg, A. Pleten, B. Primerano, D. Scott, G. Scherer, and S. Stickel, “High-
performance OLED microdisplays made with multi-stack OLED formulations
on CMOS backplanes,” Proc. SPIE 11473, 114730F (2020).
85. J. H. Lee, C. H. Chen, P. H. Lee, H. Y. Lin, M. K. Leung, T. L. Chiu, and C. F.
Lin, “Blue organic light-emitting diodes: current status, challenges, and future
outlook,” J. Mater. Chem. C 7, 5874–5888 (2019).
86. N. Holonyak and S. F. Bevacqua, “Coherent (visible) light emission from
Ga(As1−xPx) junctions,” Appl. Phys. Lett. 1, 82–83 (1962).
87. Y. Nanishi, “The birth of the blue LED,” Nat. Photonics 8, 884–886 (2014).
88. X. J. Yu, Y. L. Ho, L. Tan, H. C. Huang, and H. S. Kwok, “LED-based projection
systems,” J. Disp. Technol. 3, 295–303 (2007).
89. H. X. Jiang, S. X. Jin, J. Li, J. Shakya, and J. Y. Lin, “III-nitride blue
microdisplays,” Appl. Phys. Lett. 78, 1303–1305 (2001).
90. A. R. Anwar, M. T. Sajjad, M. A. Johar, C. A. Hernández-Gutiérrez, M.
Usman, and S. P. Łepkowski, “Recent progress in micro-LED-based display
technologies,” Laser Photonics Rev. 16, 2100427 (2022).
91. K. L. Liang, W. H. Kuo, H. T. Shen, P. W. Yu, Y. H. Fang, and C. C. Lin, “Advances
in color-converted micro-LED arrays,” Jpn. J. Appl. Phys. 60, SA0802 (2021).
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 849
92. Z. J. Liu, W. C. Chong, K. M. Wong, K. H. Tam, and K. M. Lau, “A novel
BLU-free full-color LED projector using LED on silicon micro-displays,” IEEE
Photonics Technol. Lett. 25, 2267–2270 (2013).
93. G. R. B. E. Römer and P. Bechtold, “Electro-optic and acousto-optic laser beam
scanners,” Phys. Procedia 56, 29–39 (2014).
94. G. Silva, F. Carpignano, F. Guerinoni, S. Costantini, M. De Fazio, and S. Merlo,
“Optical detection of the electromechanical response of MEMS micromirrors
designed for scanning picoprojectors,” IEEE J. Sel. Top. Quantum Electron. 21,
147–156 (2015).
95. J. Ma, “Advanced MEMS-based technologies and displays,” Displays 37, 2–10
(2015).
96. J. Reitterer, Z. Chen, A. Balbekova, G. Schmid, G. Schestak, F. Nassar, M.
Dorfmeister, and M. Ley, “Ultra-compact micro-electro-mechanical laser beam
scanner for augmented reality applications,” Proc. SPIE 11765, 1176504 (2021).
97. C. L. Yang, Y. H. Wu, I. A. Yao, Y. S. Tsou, C. H. Tsai, and J. S. Lin, “47-1:
invited paper: high resolution HDR VR display using Mini-LED,” Dig. Tech.
Pap. - Soc. Inf. Disp. Int. Symp. 52, 636–639 (2021).
98. M. S. Park, H. J. Youn, C. S. Lee, and S. H. Yoon, “P-156: flicker effect depending
on the density of charge impurities in the cell,” Dig. Tech. Pap. - Soc. Inf. Disp.
Int. Symp. 49, 1744–1747 (2018).
99. T. Nakamura, M. Tada, and H. Kimura, “A 550-PPI LCD using 1.5µm channel
width LTPS TFTs with low frame rate driving,” J. Soc. Inf. Disp. 23, 580–586
(2015).
100. P. Huo, L. Yao, B. Li, Y. Fang, Z. Wang, J. Xu, and J. Xie, “P-3.3: analysis and
improvement of flicker NG on VR display,” Dig. Tech. Pap. - Soc. Inf. Disp. Int.
Symp. 52, 721–723 (2021).
101. K. Yoshida, S. Ueda, A. Hachiya, K. Okada, A. Hirai, Y. Kawahira, K. Misaki,
H. Furukawa, and H. Katoh, “Thin and low-reflection metal black matrix for high
PPI liquid crystal display,” J. Soc. Inf. Disp. 30, 413–422 (2022).
102. L. Yao, F. Li, Y. Fang, W. Li, H. Su, L. Yan, Y. Gao, K. Li, C. Wang, L. Hou, and
J. Xie, “P-3.7: study on preparation and performance of new black matrix with
high PPI VR display,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 52, 735–737
(2021).
103. Y. Ma, H. Ikeno, P. Shen, L. Wu, H. Wu, Q. Song, and C. Tseng, “37-3: a novel
pixel structure for 847 PPI display with LTPS technology,” Dig. Tech. Pap. - Soc.
Inf. Disp. Int. Symp. 47, 477–479 (2016).
104. J. H. Kwon, S. H. Ji, S. H. Han, M. S. Kwak, M. Jun, and I. B. Kang, “P-184:
improvement of color mixing in high PPI mobile displays,” Dig. Tech. Pap. -
Soc. Inf. Disp. Int. Symp. 47, 1830–1833 (2016).
105. C. Kim, K. Kim, O. Kwon, J. Jung, J. K. Park, D. H. Kim, and K. Jung, “Fine
metal mask material and manufacturing process for high-resolution active-matrix
organic light-emitting diode displays,” J. Soc. Inf. Disp. 28, 668–679 (2020).
106. C. Kim, J. H. Jung, D. Y. Sung, J. S. Park, J. K. Park, T. Kim, D. H. Kim,
and K. Jung, “11-3: 2,400ppi RGB side-by-side OLED micro-display for AR
applications,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 52, 131–134 (2021).
107. Y. Jiang, B. S. T. Tam, S. C. Dong, and C. W. Tang, “61-2: 2-inch, 2,000-PPI
silicon nitride mask for patterning ultra-high-resolution OLED displays,” Dig.
Tech. Pap. - Soc. Inf. Disp. Int. Symp. 51, 909–912 (2020).
108. C. Hwang, S. S. Kim, S. W. Bang, S. H. Choi, B. D. Chin, and H. Ko, “67-
3: unique belt plane source evaporation techniques for the mass production of
2250ppi AMOLED and 77” QD OLED TV,” Dig. Tech. Pap. - Soc. Inf. Disp.
Int. Symp. 50, 949–952 (2019).
850 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
109. P. E. Malinowski, T. H. Ke, A. Nakamura, Y. H. Liu, D. Vander Velpen, E.
Vandenplas, N. Papadopoulos, A. J. Kronemeijer, J. L. van der Steen, S. Steudel,
C. C. Kuo, Y. Y. Huang, Y. H. Chen, M. H. Yeh, G. Gelinck, and P. Heremans,
“High resolution photolithography for direct view active matrix organic light-
emitting diode augmented reality displays,” J. Soc. Inf. Disp. 26, 128–136 (2018).
110. T. H. Ke, A. Glushkova, C. M. Sandeheng, G. M. Alvarez, E. Vandenplas, R.
M. Ciarnain, P. E. Malinowski, J. Mitard, N. Rassoul, A. Belmonte, Z. Tokei,
G. S. Kar, J. Genoe, and P. Heremans, “11-2: technology developments in
high-resolution FMM-free OLED and BEOL IGZO TFTs for power-efficient
microdisplays,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 52, 127–130 (2021).
111. M. C. Gather, A. Köhnen, A. Falcou, H. Becker, and K. Meerholz, “Solution-
processed full-color polymer organic light-emitting diode displays fabricated by
direct photolithography,” Adv. Funct. Mater. 17, 191–200 (2007).
112. K. Kim, G. Kim, B. Ram Lee, S. Ji, S. Y. Kim, B. Wan An, M. Hoon Song, and
J. U. Park, “High-resolution electrohydrodynamic jet printing of small-molecule
organic light-emitting diodes,” Nanoscale 7, 13410–13415 (2015).
113. L. Mu, C. Jiang, J. Wang, H. Zheng, L. Ying, M. Xu, J. Wang, J. Peng, and Y. Cao,
“34-3: OLED display with high resolution fabricated by electrohydrodynamic
printing,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 51, 485–488 (2020).
114. L. Qi, X. Zhang, W. C. Chong, P. Li, and K. M. Lau, “848 ppi high-brightness
active-matrix micro-LED micro-display using GaN-on-Si epi-wafers towards
mass production,” Opt. Express 29, 10580–10591 (2021).
115. K. A. Bulashevich, S. S. Konoplev, and S. Y. Karpov, “Effect of die shape and
size on performance of III-nitride micro-LEDs: a modeling study,” Photonics 5,
41 (2018).
116. L. Zhang, F. Ou, W. C. Chong, Y. Chen, and Q. Li, “Wafer-scale monolithic
hybrid integration of Si-based IC and III–V epi-layers—A mass manufacturable
approach for active matrix micro-LED micro-displays,” J. Soc. Inf. Disp. 26,
137–145 (2018).
117. H. Kawanishi, H. Onuma, M. Maegawa, T. Kurisu, T. Ono, S. Akase, S. Yam-
aguchi, N. Momotani, Y. Fujita, Y. Kondo, K. Kubota, T. Yoshida, Y. Ikawa,
T. Ono, H. Higashisaka, Y. Hirano, and S. Anzai, “High-resolution and high-
brightness full-colour ‘silicon display’ for augmented and mixed reality,” J. Soc.
Inf. Disp. 29, 57–67 (2021).
118. M. A. Triana, E. L. Hsiang, C. Zhang, Y. Dong, and S. T. Wu, “Luminescent
nanomaterials for energy-efficient display and healthcare,” ACS Energy Lett. 7,
1001–1020 (2022).
119. Z. Liu, C. H. Lin, B. R. Hyun, C. W. Sher, Z. Lv, B. Luo, F. Jiang, T. Wu, C. H.
Ho, H. C. Kuo, and J. H. He, “Micro-light-emitting diodes with quantum dots in
display technology,” Light: Sci. Appl. 9, 83 (2020).
120. T. H. Kim, K. S. Cho, E. K. Lee, S. J. Lee, J. Chae, J. W. Kim, D. H. Kim, J.
Y. Kwon, G. Amaratunga, S. Y. Lee, B. L. Choi, Y. Kuk, J. M. Kim, and K.
Kim, “Full-colour quantum dot displays fabricated by transfer printing,” Nat.
Photonics 5, 176–182 (2011).
121. R. Palankar, N. Medvedev, A. Rong, and M. Delcea, “Fabrication of quantum dot
microarrays using electron beam lithography for applications in analyte sensing
and cellular dynamics,” ACS Nano 7, 4617–4628 (2013).
122. H. L. Kang, J. Kang, J. K. Won, S. M. Jung, J. Kim, C. H. Park, B. K. Ju, M. G.
Kim, and S. K. Park, “Spatial light patterning of full color quantum dot displays
enabled by locally controlled surface tailoring,” Adv. Opt. Mater. 6, 1701335
(2018).
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 851
123. D. M. Geum, S. K. Kim, C. M. Kang, S. H. Moon, J. Kyhm, J. Han, D. S. Lee, and
S. Kim, “Strategy toward the fabrication of ultrahigh-resolution micro-LED dis-
plays by bonding-interface-engineered vertical stacking and surface passivation,”
Nanoscale 11, 23139–23148 (2019).
124. Z. Wang, S. Zhu, X. Shan, Z. Yuan, X. Cui, and P. Tian, “Full-color micro-LED
display based on a single chip with two types of InGaN/GaN MQWs,” Opt. Lett.
46, 4358–4361 (2021).
125. L. Li, G. Tang, Z. Shi, H. Ding, C. Liu, D. Cheng, Q. Zhang, L. Yin, Z. Yao,
L. Duan, D. Zhang, C. Wang, M. Feng, Q. Sun, Q. Wang, Y. Han, L. Wang, Y.
Luo, and X. Sheng, “Transfer-printed, tandem microscale light-emitting diodes
for full-color displays,” Proc. Natl. Acad. Sci. U. S. A. 118, e2023436118 (2021).
126. D. J. Kong, C. M. Kang, J. Y. Lee, J. Kim, and D. S. Lee, “Color tunable
monolithic InGaN/GaN LED having a multi-junction structure,” Opt. Express
24, A667–A673 (2016).
127. S. H. Mun, C. M. Kang, J. H. Min, S. Y. Choi, W. L. Jeong, G. G. Kim, J. S. Lee,
K. P. Kim, H. C. Ko, and D. S. Lee, “Highly efficient full-color inorganic LEDs
on a single wafer by using multiple adhesive bonding,” Adv. Mater. Interfaces 8,
2100300 (2021).
128. C. M. Kang, J. Y. Lee, D. J. Kong, J. P. Shim, S. Kim, S. H. Mun, S. Y. Choi, M.
D. Park, J. Kim, and D. S. Lee, “Hybrid full-color inorganic light-emitting diodes
integrated on a single wafer using selective area growth and adhesive bonding,”
ACS Photonics 5, 4413–4422 (2018).
129. K. Yadavalli, C. L. Chuang, and H. S. El-Ghoroury, “Monolithic and heteroge-
neous integration of RGB micro-LED arrays with pixel-level optics array and
CMOS image processor to enable small form-factor display applications,” Proc.
SPIE 11310, 113100Z (2020).
130. C. M. Kang, S. J. Kang, S. H. Mun, S. Y. Choi, J. H. Min, S. Kim, J. P. Shim,
and D. S. Lee, “Monolithic integration of AlGaInP-based red and InGaN-based
green LEDs via adhesive bonding for multicolor emission,” Sci. Rep. 7, 10333
(2017).
131. H. S. El-Ghoroury, M. Yeh, J. C. Chen, X. Li, and C. L. Chuang, “Growth of
monolithic full-color GaN-based LED with intermediate carrier blocking layers,”
AIP Adv. 6, 075316 (2016).
132. Y. Huang, E. Liao, R. Chen, and S. T. Wu, “Liquid-crystal-on-silicon for
augmented reality displays,” Appl. Sci. 8, 2366 (2018).
133. R. D˛abrowski, P. Kula, and J. Herman, “High birefringence liquid crystals,”
Crystals 3, 443–482 (2013).
134. S. Gauza, H. Wang, C. H. Wen, S. T. Wu, A. J. Seed, and R. Dabrowski,
“High birefringence isothiocyanato tolane liquid crystals,” Jpn. J. Appl. Phys.
42, 3463–3466 (2003).
135. J. Zou, Z. Yang, C. Mao, and S. T. Wu, “Fast-response liquid crystals for 6G
optical communications,” Crystals 11, 797 (2021).
136. K. H. Fan-Chiang, S. H. Chen, and S. T. Wu, “High-definition vertically aligned
liquid crystal microdisplays using a circularly polarized light,” Appl. Phys. Lett.
87, 031110 (2005).
137. K. H. Fan-Chiang, S. T. Wu, and S. H. Chen, “Fringing field effect of the
liquid-crystal-on-silicon devices,” Jpn. J. Appl. Phys. 41, 4577–4585 (2002).
138. Y. W. Li, W. H. Chen, K. H. Fan Chiang, H. C. Kuo, and H. C. Tsai, “18.2: color
filter LCOS with double mirror structure,” Dig. Tech. Pap. - Soc. Inf. Disp. Int.
Symp. 45, 223–225 (2014).
139. H. Vithana, “Vertically aligned nematic mode liquid crystal display having large
tilt angles and high contrast,” U.S. patent 9,551,901 (10 September 2017).
852 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
140. C. H. Liao, “Reducing fringe field effect for spatial light modulator,” U.S. patent
10,416,498 (21 June 2018).
141. L. Yang, J. Xia, and X. Zhang, “Optimization of fringing field effect using
dielectric separation and local electric field enhancement,” J. Disp. Technol. 11,
242–247 (2015).
142. Y. Isomae, Y. Shibata, T. Ishinabe, and H. Fujikake, “Design of 1-µm-pitch
liquid crystal spatial light modulators having dielectric shield wall structure for
holographic display with wide field of view,” Opt. Rev. 24, 165–176 (2017).
143. Y. Isomae, T. Ishinabe, Y. Shibata, and H. Fujikake, “Alignment control of liquid
crystals in a 1.0-µm-pitch spatial light modulator by lattice-shaped dielectric wall
structure,” J. Soc. Inf. Disp. 27, 251–258 (2019).
144. U. Efron, B. Apter, and E. Bahat-Treidel, “Fringing-field effect in liquid-crystal
beam-steering devices: an approximate analytical model,” J. Opt. Soc. Am. A
21, 1996–2008 (2004).
145. S. Moser, M. Ritsch-Marte, and G. Thalhammer, “Model-based compensation
of pixel crosstalk in liquid crystal spatial light modulators,” Opt. Express 27,
25046–25063 (2019).
146. M. Persson, D. Engström, and M. Goksör, “Reducing the effect of pixel crosstalk
in phase only spatial light modulators,” Opt. Express 20, 22334–22343 (2012).
147. H. M. Chen, J. P. Yang, H. T. Yen, Z. N. Hsu, Y. Huang, and S. T. Wu, “Pursuing
high quality phase-only liquid crystal on silicon (LCoS) devices,” Appl. Sci. 8,
2323 (2018).
148. J. Zou, Q. Yang, E. L. Hsiang, H. Ooishi, Z. Yang, K. Yoshidaya, and S. T. Wu,
“Fast-response liquid crystal for spatial light modulator and LiDAR applications,”
Crystals 11, 93 (2021).
149. S. H. Lee, S. S. Bhattacharyya, H. S. Jin, and K. U. Jeong, “Devices and mate-
rials for high-performance mobile liquid crystal displays,” J. Mater. Chem. 22,
11893–11903 (2012).
150. J. Li, Z. Peng, R. Chen, J. Li, M. Hu, L. Zhang, and Z. An, “Investigation
of terminal olefine in the isothiocyanatotolane liquid crystals with alkoxy end
group,” Liq. Cryst. 45, 1498–1507 (2018).
151. S. T. S. Holmstrom, U. Baran, and H. Urey, “MEMS laser scanners: a review,” J.
Microelectromech. Syst. 23, 259–275 (2014).
152. H. Urey, D. W. Wine, and J. R. Lewis, “Scanner design and resolution trade-offs
for miniature scanning displays,” Proc. SPIE 3636, 60–68 (1999).
153. P. J. Brosens, “Dynamic mirror distortions in optical scanning,” Appl. Opt. 11,
2987–2989 (1972).
154. S. T. Wu, “Nematic liquid crystal modulator with response time less than 100 µs
at room temperature,” Appl. Phys. Lett. 57, 986–988 (1990).
155. F. Peng, H. Chen, F. Gou, Y. H. Lee, M. Wand, M. C. Li, S. L. Lee, and S. T. Wu,
“Analytical equation for the motion picture response time of display devices,” J.
Appl. Phys. 121, 023108 (2017).
156. S. I. Jo, S. S. Yoon, S. R. Park, J. H. Lee, M. C. Jun, and I. B. Kang, “44-2:
invited paper: fast response time advanced high performance in-plane switching
(AH-IPS) mode for high resolution application,” Dig. Tech. Pap. - Soc. Inf. Disp.
Int. Symp. 49, 581–584 (2018).
157. F. Gou, H. Chen, M. C. Li, S. L. Lee, and S. T. Wu, “Submillisecond-response
liquid crystal for high-resolution virtual reality displays,” Opt. Express 25,
7984–7997 (2017).
158. W. K. Choi, C. W. Hsu, C. H. Tung, and B. K. Tseng, “Effects of electrode
structure and dielectric anisotropy on the performance of VA-FFS LC mode,”
Opt. Express 27, 34343–34358 (2019).
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 853
159. T. H. Choi, S. W. Oh, Y. J. Park, Y. Choi, and T. H. Yoon, “Fast fringe-field
switching of a liquid crystal cell by two-dimensional confinement with virtual
walls,” Sci. Rep. 6, 27936 (2016).
160. T. Matsushima, S. Kimura, and S. Komura, “Fast response in-plane switching
liquid crystal display mode optimized for high-resolution virtual-reality head-
mounted display,” J. Soc. Inf. Disp. 29, 221–229 (2021).
161. J. H. Yoon, S. J. Lee, Y. J. Lim, E. J. Seo, H. S. Shin, J. M. Myoung, and S. H.
Lee, “Fast switching, high contrast and high resolution liquid crystal device for
virtual reality display,” Opt. Express 26, 34142–34149 (2018).
162. M. Engel, G. Bernatz, A. Götz, H. Hirschmann, and S. K. Lee, “43.1: invited
paper: UB-FFS: new materials for advanced mobile applications,” Dig. Tech.
Pap. - Soc. Inf. Disp. Int. Symp. 46, 645–647 (2015).
163. H. Chen, M. Hu, F. Peng, J. Li, Z. An, and S. T. Wu, “Ultra-low viscosity liquid
crystal materials,” Opt. Mater. Express 5, 655–660 (2015).
164. Y. Chen, F. Peng, and S. T. Wu, “Submillisecond-response vertical-aligned liquid
crystal for color sequential projection displays,” J. Disp. Technol. 9, 78–81 (2013).
165. F. Peng, F. Gou, H. Chen, Y. Huang, and S. T. Wu, “A submillisecond-response
liquid crystal for color sequential projection displays,” J. Soc. Inf. Disp. 24,
241–245 (2016).
166. A. K. Abeeluck, A. Iverson, H. Goetz, and E. Passon, “58-2: Invited Paper:
high-performance displays for wearable and HUD applications,” Dig. Tech. Pap.
- Soc. Inf. Disp. Int. Symp. 49, 768–771 (2018).
167. S. T. Wu, “Design of a liquid crystal based tunable electrooptic filter,” Appl. Opt.
28, 48–52 (1989).
168. Y. Li, Z. Yang, R. Chen, L. Mo, J. Li, M. Hu, and S. T. Wu, “Submillisecond-
response polymer network liquid crystal phase modulators,” Polymers 12, 2862
(2020).
169. Q. Yang, J. Zou, Y. Li, and S. T. Wu, “Fast-response liquid crystal phase
modulators with an excellent photostability,” Crystals 10, 765 (2020).
170. Z. Zhang, Z. You, and D. Chu, “Fundamentals of phase-only liquid crystal on
silicon (LCOS) devices,” Light: Sci. Appl. 3, e213 (2014).
171. Q. Guo, K. Yan, V. Chigrinov, H. Zhao, and M. Tribelsky, “Ferroelectric liquid
crystals: physics and applications,” Crystals 9, 470 (2019).
172. V. Chigrinov, Q. Guo, and A. Kudreyko, “Photo-aligned ferroelectric liquid
crystal devices with novel electro-optic characteristics,” Crystals 10, 563 (2020).
173. T. Haist and W. Osten, “Holography using pixelated spatial light modula-
tors—part 1: theory and basic considerations,” J. Micro/Nanolith. MEMS
MOEMS 14, 041310 (2015).
174. J. H. R. Chang, B. V. K. Vijaya Kumar, and A. C. Sankaranarayanan, “216 shades
of gray: high bit-depth projection using light intensity control,” Opt. Express 24,
27937–27950 (2016).
175. K. Rathinavel, H. Wang, A. Blate, and H. Fuchs, “An extended depth-at-field
volumetric near-eye augmented reality display,” IEEE Trans. Visual. Comput.
Graphics 24, 2857–2866 (2018).
176. S. Lee, Y. Jo, D. Yoo, J. Cho, D. Lee, and B. Lee, “Tomographic near-eye
displays,” Nat. Commun. 10, 2497 (2019).
177. H. Urey, D. W. Wine, and T. D. Osborn, “Optical performance requirements for
MEMS-scanner-based microdisplays,” Proc. SPIE 4178, 176–185 (2000).
178. C. L. Lin, C. C. Hung, P. Y. Kuo, and M. H. Cheng, “New LTPS pixel circuit with
AC driving method to reduce OLED degradation for 3D AMOLED displays,” J.
Disp. Technol. 8, 681–683 (2012).
854 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
179. H. J. In and O. K. Kwon, “A simple pixel structure using polycrystalline-silicon
thin-film transistors for high-resolution active-matrix organic light-emitting diode
displays,” IEEE Electron Device Lett. 33, 1018–1020 (2012).
180. C. L. Lin, F. H. Chen, C. C. Hung, P. S. Chen, M. Y. Deng, C. M. Lu, and T.
H. Huang, “New a-IGZO pixel circuit composed of three transistors and one
capacitor for use in high-speed-scan AMOLED displays,” J. Disp. Technol. 11,
1031–1034 (2015).
181. N. H. Keum, K. Oh, S. K. Hong, and O. K. Kwon, “A pixel structure using
block emission driving method for high image quality in active matrix organic
light-emitting diode displays,” J. Disp. Technol. 12, 1250–1256 (2016).
182. C. L. Lin, C. C. Hung, W. Y. Chang, K. W. Chou, and C. Y. Chuang, “Novel
a-Si:H AMOLED pixel circuit to ameliorate OLED luminance degradation by
external detection,” IEEE Electron Device Lett. 32, 1716–1718 (2011).
183. S. Sambandan and A. Nathan, “Single-technology-based statistical calibration
for high-performance active-matrix organic LED displays,” J. Disp. Technol. 3,
284–294 (2007).
184. C. L. Lin, P. C. Lai, L. W. Shih, C. C. Hung, P. C. Lai, T. Y. Lin, K. H. Liu, and
T. H. Wang, “Compensation pixel circuit to improve image quality for mobile
AMOLED displays,” IEEE J. Solid-State Circuits 54, 489–500 (2019).
185. F. Fournier and J. Rolland, “Design methodology for high brightness projectors,”
J. Disp. Technol. 4, 86–91 (2008).
186. E. L. Hsiang, Z. Yang, T. Zhan, J. Zou, H. Akimoto, and S. T. Wu, “Optimizing the
display performance for virtual reality systems,” OSA Continuum 4, 3052–3067
(2021).
187. L. Noui, J. Reitterer, and M. Schöffmann, “Laser beam scanner and combiner
architectures,” Proc. SPIE 11765, 1176508 (2021).
188. M. Sugawara, M. Suzuki, and N. Miyauchi, “14-5L: late-news paper: retinal
imaging laser eyewear with focus-free and augmented reality,” Dig. Tech. Pap. -
Soc. Inf. Disp. Int. Symp. 47, 164–167 (2016).
189. J. Zou, T. Zhan, E. L. Hsiang, X. Du, X. Yu, K. Li, and S. T. Wu, “Doubling
the optical efficiency of VR systems with a directional backlight and a diffractive
deflection film,” Opt. Express 29, 20673–20686 (2021).
190. Y. J. Wang, S. H. Ouyang, W. C. Chao, J. G. Lu, and H. P. D. Shieh, “High
directional backlight using an integrated light guide plate,” Opt. Express 23,
1567–1575 (2015).
191. Y. Zhang, D. Yi, W. Qiao, and L. Chen, “Directional backlight module based on
pixelated nano-gratings,” Opt. Commun. 459, 125034 (2020).
192. A. Ghosh, I. Khayrullin, F. Vazan, F. Zhao, H. Lin, O. Prache, L. Sziklas, S.
Peruvemba, and A. Sculley, “Developing the world’s brightest WUXGA OLED
microdisplay,” Inf. Disp. 38, 19–23 (2022).
193. X. Fu, S. Yin, Y. A. Chen, L. Zhu, Q. Dong, C. H. Chang, and F. So, “Light
extraction in tandem organic light emitting diodes,” Appl. Phys. Lett. 119, 060504
(2021).
194. B. Pyo, C. W. Joo, H. S. Kim, B. H. Kwon, J. I. Lee, J. Lee, and M. C. Suh, “A
nanoporous polymer film as a diffuser as well as a light extraction component for
top emitting organic light emitting diodes with a strong microcavity structure,”
Nanoscale 8, 8575–8582 (2016).
195. J. Kim, Y. Qu, C. Coburn, and S. R. Forrest, “Efficient outcoupling of organic
light-emitting devices using a light-scattering dielectric layer,” ACS Photonics
5, 3315–3321 (2018).
196. J. Song, K. H. Kim, E. Kim, C. K. Moon, Y. H. Kim, J. J. Kim, and S. Yoo, “Lens-
free OLEDs with over 50% external quantum efficiency via external scattering
and horizontally oriented emitters,” Nat. Commun. 9, 3207 (2018).
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 855
197. M. Thomschke, S. Reineke, B. Lüssem, and K. Leo, “Highly efficient white top-
emitting organic light-emitting diodes comprising laminated microlens films,”
Nano Lett. 12, 424–428 (2012).
198. Y. Sun and S. R. Forrest, “Organic light emitting devices with enhanced out-
coupling via microlenses fabricated by imprint lithography,” J. Appl. Phys. 100,
073106 (2006).
199. Y. Motoyama, K. Sugiyama, H. Tanaka, H. Tsuchioka, K. Matsusaki, and H.
Fukumoto, “High-efficiency OLED microdisplay with microlens array,” J. Soc.
Inf. Disp. 27, 354–360 (2019).
200. K. Itonaga, S. Sudo, J. Nishikawa, K. Kimura, H. Uchiyama, M. Yagi, R. Sawano,
T. Matsuyama, K. Sasaki, K. Itatsu, T. Tsuchiya, Y. Nomura, and Y. Sato, “47-1:
the world smallest OLED microdisplay projection device design methodology,”
Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 51, 683–686 (2020).
201. S. Amoah, X. Fu, S. Yin, Q. Dong, C. Dong, and F. So, “Curved mirror arrays for
light extraction in top-emitting organic light-emitting diodes,” ACS Appl. Mater.
Interfaces 14, 9377–9385 (2022).
202. X. Fu, Y. A. Chen, D. H. Shin, Y. Mehta, I. T. Chen, N. Barange, L. Zhu, S.
Amoah, C. H. Chang, and F. So, “Recovering cavity effects in corrugated organic
light emitting diodes,” Opt. Express 28, 32214–32225 (2020).
203. J. Liang, Y. Du, K. Wang, A. Ren, X. Dong, C. Zhang, J. Tang, Y. Yan, and Y. S.
Zhao, “Ultrahigh color rendering in RGB perovskite micro-light-emitting diode
arrays with resonance-enhanced photon recycling for next generation displays,”
Adv. Opt. Mater. 10, 2101642 (2022).
204. X. Fu, Y. Mehta, Y. A. Chen, L. Lei, L. Zhu, N. Barange, Q. Dong, S. Yin, J.
Mendes, S. He, R. Gogusetti, C. H. Chang, and F. So, “Directional polarized light
emission from thin-film light-emitting diodes,” Adv. Mater. 33, 2006801 (2021).
205. S. J. Lee, C. mo Kang, J. W. Shin, D. H. Ahn, C. W. Joo, H. Cho, N. S. Cho,
H. M. Youn, Y. J. An, J. S. Kim, J. Lee, and H. Lee, “Three-wavelength white
organic light-emitting diodes on silicon for high luminance and color gamut
microdisplays,” J. Ind. Eng. Chem. 105, 132–137 (2022).
206. H. Cho, C. W. Byun, C. M. Kang, J. W. Shin, B. H. Kwon, S. Choi, N. S. Cho,
J. I. Lee, H. Kim, J. H. Lee, M. Kim, and H. Lee, “White organic light-emitting
diode (OLED) microdisplay with a tandem structure,” J. Inf. Disp. 20, 249–255
(2019).
207. L. Penninck, M. Diethelm, S. Altazin, R. Hiestand, C. Kirsch, and B. Ruh-
staller, “Modelling crosstalk through common semiconductor layers in AMOLED
displays,” J. Soc. Inf. Disp. 26, 546–554 (2018).
208. H. W. Bae, Y. W. Kwon, M. An, J. Kim, J. H. Kwon, and D. Lee, “High-color-
stability and low-driving-voltage white organic light-emitting diodes on silicon
with interlayers of thin charge generation units for microdisplay applications,”
ACS Appl. Electron. Mater. 3, 3240–3246 (2021).
209. H. Cho, C. W. Joo, S. Choi, C. mo Kang, G. H. Kim, J. W. Shin, B. H. Kwon, H.
Lee, C. W. Byun, and N. S. Cho, “Design of white tandem organic light-emitting
diodes for full-color microdisplay with high current efficiency and high color
gamut,” ETRI Journal 43, 1093–1102 (2021).
210. U. Vogel, B. Richter, P. Wartenberg, P. König, O. R. Hild, K. Fehse, M. Schober,
E. Bodenstein, and B. Beyer, “OLED microdisplays in near-to-eye applications:
challenges and solutions,” Proc. SPIE 10335, 1033503 (2017).
211. S. K. Kim, M. J. Park, R. Lampande, S. W. Jung, H.-U. Park, J. K. Jeong, and
J. H. Kwon, “Primary color generation from white organic light-emitting diodes
using a cavity control layer for AR/VR applications,” Org. Electron. 87, 105938
(2020).
856 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
212. H. Cho, C. W. Joo, S. Choi, C. mo Kang, B. H. Kwon, J. W. Shin, K. Kim, D. H.
Ahn, and N. S. Cho, “Organic light-emitting diode structure for high color gamut
in high-resolution microdisplay: over 90% color gamut based on BT.2020,” Org.
Electron. 101, 106419 (2022).
213. M. K. Fung, Y. Q. Li, and L. S. Liao, “Tandem organic light-emitting diodes,”
Adv. Mater. 28, 10381–10408 (2016).
214. W. J. Joo, J. Kyoung, M. Esfandyarpour, S. H. Lee, H. Koo, S. Song, Y. N. Kwon,
S. H. Song, J. C. Bae, A. Jo, M. J. Kwon, S. H. Han, S. H. Kim, S. Hwang, and
M. L. Brongersma, “Metasurface-driven OLED displays beyond 10,000 pixels
per inch,” Science 370, 459–463 (2020).
215. F. Olivier, S. Tirano, L. Dupré, B. Aventurier, C. Largeron, and F. Templier,
“Influence of size-reduction on the performances of GaN-based micro-LEDs for
display application,” J. Lumin. 191, 112–116 (2017).
216. S. Han, C. Xu, H. Li, S. Liu, H. Xu, Y. Zhu, A. Fang, and X. Wang, “AlGaInP-
based micro-LED array with enhanced optoelectrical properties,” Opt. Mater.
114, 110860 (2021).
217. M. S. Wong, J. A. Kearns, C. Lee, J. M. Smith, C. Lynsky, G. Lheureux, H. Choi,
J. Kim, C. Kim, S. Nakamura, S. Nakamura, J. S. Speck, S. P. DenBaars, and S. P.
DenBaars, “Improved performance of AlGaInP red micro-light-emitting diodes
with sidewall treatments,” Opt. Express 28, 5787–5793 (2020).
218. H. H. Huang, S. K. Huang, Y. L. Tsai, S. W. Wang, Y. Y. Lee, S. Y. Weng,
H. C. Kuo, and C. chung Lin, “Investigation on reliability of red micro-light
emitting diodes with atomic layer deposition passivation layers,” Opt. Express
28, 38184–38195 (2020).
219. J. Park, W. Baek, D. M. Geum, and S. Kim, “Understanding the sidewall pas-
sivation effects in AlGaInP/GaInP micro-LED,” Nanoscale Res. Lett. 17, 29
(2022).
220. M. S. Wong, D. Hwang, A. I. Alhassan, C. Lee, R. Ley, S. Nakamura, and S.
P. DenBaars, “High efficiency of III-nitride micro-light-emitting diodes by side-
wall passivation using atomic layer deposition,” Opt. Express 26, 21324–21331
(2018).
221. D. Chen, Z. Wang, F. C. Hu, C. Shen, N. Chi, W. Liu, W. Liu, D. W. Zhang, H.
L. Lu, and H. L. Lu, “Improved electro-optical and photoelectric performance of
GaN-based micro-LEDs with an atomic layer deposited AlN passivation layer,”
Opt. Express 29, 36559–36566 (2021).
222. M. S. Wong, C. Lee, D. J. Myers, D. Hwang, J. A. Kearns, T. Li, J. S. Speck, S.
Nakamura, and S. P. DenBaars, “Size-independent peak efficiency of III-nitride
micro-light-emitting-diodes using chemical treatment and sidewall passivation,”
Appl. Phys. Express 12, 097004 (2019).
223. J. M. Smith, R. Ley, M. S. Wong, Y. H. Baek, J. H. Kang, C. H. Kim, M. J.
Gordon, S. Nakamura, J. S. Speck, and S. P. DenBaars, “Comparison of size-
dependent characteristics of blue and green InGaN microLEDs down to 1µm in
diameter,” Appl. Phys. Lett. 116, 071102 (2020).
224. K. A. Bulashevich and S. Yu Karpov, “Impact of surface recombination on
efficiency of III-nitride light-emitting diodes,” Phys. Status Solidi. Rapid Res.
Lett. 10, 480–484 (2016).
225. Z. Zhuang, D. Iida, and K. Ohkawa, “InGaN-based red light-emitting diodes:
from traditional to micro-LEDs,” Jpn. J. Appl. Phys. 61, SA0809 (2022).
226. Z. Zhuang, D. Iida, M. Velazquez-Rizo, and K. Ohkawa, “630-nm red InGaN
micro-light-emitting diodes (<20 µm×20 µm) exceeding 1mW/mm2 for full-
color micro-displays,” Photonics Res. 9, 1796–1802 (2021).
227. P. Li, H. Li, H. Zhang, C. Lynsky, M. Iza, J. S. Speck, S. Nakamura, and S. P.
DenBaars, “Size-independent peak external quantum efficiency (>2%) of InGaN
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 857
red micro-light-emitting diodes with an emission wavelength over 600 nm,” Appl.
Phys. Lett. 119, 081102 (2021).
228. Y. Zhao, Y. Zhao, J. Liang, Q. Zeng, Y. Li, Y. Li, P. Li, P. Li, K. Fan, K. Fan,
W. Sun, W. Sun, J. Lv, Y. Qin, Q. Wang, J. Tao, J. Tao, W. Wang, and W. Wang,
“2021 PPI silicon-based AlGaInP red micro-LED arrays fabricated via wafer
bonding and epilayer lift-off,” Opt. Express 29, 20217–20228 (2021).
229. J. T. Oh, S. Y. Lee, Y. T. Moon, J. H. Moon, S. Park, K. Y. Hong, K. Y. Song,
C. Oh, J. I. Shim, H. H. Jeong, J. O. Song, H. Amano, and T. Y. Seong, “Light
output performance of red AlGaInP-based light emitting diodes with different
chip geometries and structures,” Opt. Express 26, 11194–11200 (2018).
230. P. Li, H. Li, M. S. Wong, P. Chan, Y. Yang, H. Zhang, M. Iza, J. S. Speck,
S. Nakamura, and S. P. Denbaars, “Progress of InGaN-based red micro-light
emitting diodes,” Crystals 12, 541 (2022).
231. S. Lan, H. Wan, J. Zhao, and S. Zhou, “Light extraction analysis of AlGaInP
based red and GaN based blue/green flip-chip micro-LEDs using the monte carlo
ray tracing method,” Micromachines 10, 860 (2019).
232. R. T. Ley, J. M. Smith, M. S. Wong, T. Margalith, S. Nakamura, S. P. Den-
Baars, and M. J. Gordon, “Revealing the importance of light extraction efficiency
in InGaN/GaN microLEDs via chemical treatment and dielectric passivation,”
Appl. Phys. Lett. 116, 251104 (2020).
233. A. A. Erchak, D. J. Ripin, S. Fan, P. Rakich, J. D. Joannopoulos, E. P. Ippen, G. S.
Petrich, and L. A. Kolodziejski, “Enhanced coupling to vertical radiation using
a two-dimensional photonic crystal in a semiconductor light-emitting diode,”
Appl. Phys. Lett. 78, 563–565 (2001).
234. A. David, H. Benisty, and C. Weisbuch, “Photonic crystal light-emitting sources,”
Rep. Prog. Phys. 75, 126501 (2012).
235. C. Hoepfner, “61.1: invited paper: PhlatLight photonic lattice LEDs for RPTV
light engines,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 37, 1808–1811 (2006).
236. E. Rangel, E. Matioli, Y. S. Choi, C. Weisbuch, J. S. Speck, and E. L. Hu,
“Directionality control through selective excitation of low-order guided modes
in thin-film InGaN photonic crystal light-emitting diodes,” Appl. Phys. Lett. 98,
081104 (2011).
237. E. L. Hsiang, Z. He, Z. Yang, Y. F. Lan, and S. T. Wu, “Tailoring the light distribu-
tion of micro-LED displays with a compact compound parabolic concentrator
and an engineered diffusor,” Opt. Express 29, 39859–39873 (2021).
238. W. Henry and C. Percival, “55-2: invited paper: ILED displays: next generation
display technology,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 47, 747–750
(2016).
239. S. Tanriseven, P. Maaskant, and B. Corbett, “Broadband quantum dot micro-
light-emitting diodes with parabolic sidewalls,” Appl. Phys. Lett. 92, 123501
(2008).
240. E. Khaidarov, Z. Liu, R. Paniagua-Domínguez, S. T. Ha, V. Valuckas, X. Liang,
Y. Akimov, P. Bai, C. E. Png, H. V. Demir, and A. I. Kuznetsov, “Control of
LED emission with functional dielectric metasurfaces,” Laser Photonics Rev.
14, 1900235 (2020).
241. J. Huang, Z. Hu, X. Gao, Y. Xu, and L. Wang, “Unidirectional-emitting GaN-
based micro-LED for 3D display,” Opt. Lett. 46, 3476–3479 (2021).
242. G. Y. Shiu, K. T. Chen, F. H. Fan, K. P. Huang, W. J. Hsu, J. J. Dai, C. F. Lai,
and C. F. Lin, “InGaN light-emitting diodes with an embedded nanoporous GaN
distributed bragg reflectors,” Sci. Rep. 6, 29138 (2016).
243. C. J. Wang, Y. Ke, G. Y. Shiu, Y. Y. Chen, Y. S. Lin, H. Chen, and C. F.
Lin, “InGaN resonant-cavity light-emitting diodes with porous and dielectric
reflectors,” Appl. Sci. 11, 8 (2020).
858 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
244. K. H. Fan Chiang, S. H. Chen, and S. T. Wu, “Diffraction effect on high-resolution
liquid-crystal-on-silicon devices,” Jpn. J. Appl. Phys. 44, 3068–3072 (2005).
245. Z. Yang, T. Zhan, and S. T. Wu, “Polarization independent guided-mode
resonance in liquid crystal-based polarization gratings,” OSA Continuum 3,
3107–3115 (2020).
246. Z. Yang, T. Zhan, and S. T. Wu, “Polarization-independent liquid crystal-based
refractive index sensor,” J. Soc. Inf. Disp. 29, 305–310 (2021).
247. M. F. Weber, O. B. Jr, S. C. Jr, J. M. Jonza, A. J. Ouderkirk, D. L. Wortman, and
C. A. Stover, “Display with reflective polarizer and randomizing cavity,” U.S.
patent application 6,025,897A (15 February 2000).
248. M. Duelli and A. T. Taylor, “16.3: Novel polarization conversion and integration
system for projection displays,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 34,
766–769 (2003).
249. M. S. Brennesholtz and E. H. Stupp, Projection Displays, 2nd ed. (Wiley, 2008).
250. T. Zhan, Y. H. Lee, G. Tan, J. Xiong, K. Yin, F. Gou, J. Zou, N. Zhang, D. Zhao,
J. Yang, S. Liu, and S. T. Wu, “Pancharatnam–berry optical elements for head-up
and near-eye displays [Invited],” J. Opt. Soc. Am. B 36, D52–D65 (2019).
251. T. Du, F. Fan, A. M. W. Tam, J. Sun, V. G. Chigrinov, and H. S. Kwok, “Complex
nanoscale-ordered liquid crystal polymer film for high transmittance holographic
polarizer,” Adv. Mater. 27, 7191–7195 (2015).
252. J. Kim, R. K. Komanduri, K. F. Lawler, D. J. Kekas, and M. J. Escuti, “Efficient
and monolithic polarization conversion system based on a polarization grating,”
Appl. Opt. 51, 4852–4857 (2012).
253. Texas Instruments, “DMD Optical Efficiency for Visible Wavelengths,”
https://www.ti.com/lit/pdf/dlpa083.
254. Y. X. Ren, R. D. Lu, and L. Gong, “Tailoring light with a digital micromirror
device,” Ann. Phys. 527, 447–470 (2015).
255. M. J. Deng, Y. Y. Zhao, Z. X. Liang, J. T. Chen, Y. Zhang, and X. M. Duan,
“Maximizing energy utilization in DMD-based projection lithography,” Opt.
Express 30, 4692–4705 (2022).
256. S. Essaian and J. Khaydarov, “State of the art of compact green lasers for mobile
projectors,” Opt. Rev. 19, 400–404 (2012).
257. R. Stevenson, “Lasers get the green light,” IEEE Spectrum 47, 34–39 (2010).
258. U. Hofmann, J. Janes, and H. J. Quenzer, “High-Q MEMS resonators for laser
beam scanning displays,” Micromachines 3, 509–528 (2012).
259. A. Avramescu, T. Lermer, J. Müller, C. Eichler, G. Bruederl, M. Sabathil, S. Lut-
gen, and U. Strauss, “True green laser diodes at 524 nm with 50 mW continuous
wave output power on c-plane GaN,” Appl. Phys. Express 3, 061003 (2010).
260. H. C. Yu, Z. W. Zheng, Y. Mei, R. B. Xu, J. P. Liu, H. Yang, B. P. Zhang, T.
C. Lu, and H. C. Kuo, “Progress and prospects of GaN-based VCSEL from near
UV to green emission,” Prog. Quantum Electron. 57, 1–19 (2018).
261. L. Jiang, J. Liu, A. Tian, Y. Cheng, Z. Li, L. Zhang, S. Zhang, D. Li, M. Ikeda,
and H. Yang, “GaN-based green laser diodes,” J. Semicond. 37, 111001 (2016).
262. N. Boni, R. Carminati, G. Mendicino, M. Merli, D. Terzi, B. Lazarova, and M.
Fusi, “Piezoelectric MEMS mirrors for the next generation of small form factor
AR glasses,” Proc. SPIE 12013, 1201305 (2022).
263. H. Urey, N. Nestorovic, B. S. Ng, and A. A. Gross, “Optics designs and system
MTF for laser scanning displays,” Proc. SPIE 3689, 238–248 (1999).
264. R. Mukherjee, K. Debattista, T. Bashford-Rogers, B. Waterfield, and A. Chalmers,
“A study on user preference of high dynamic range over low dynamic range
video,” Vision Comput. 32, 825–834 (2016).
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 859
265. Z. Yang, E. L. Hsiang, Y. Qian, and S. T. Wu, “Performance comparison between
mini-LED backlit LCD and OLED display for 15.6-inch notebook computers,”
Appl. Sci. 12, 1239 (2022).
266. Y. Zhao, N. Matsuda, X. Wang, M. Zannoli, and D. Lanman, “High dynamic
range near-eye displays,” Proc. SPIE 11310, 113100X (2020).
267. G. Tan, Y. Huang, M. C. Li, S. L. Lee, and S. T. Wu, “High dynamic range liquid
crystal displays with a mini-LED backlight,” Opt. Express 26, 16572–16584
(2018).
268. E. L. Hsiang, Q. Yang, Z. He, J. Zou, and S. T. Wu, “Halo effect in high-dynamic-
range mini-LED backlit LCDs,” Opt. Express 28, 36822–36837 (2020).
269. M. Xu and H. Hua, “High dynamic range head mounted display based on dual-
layer spatial modulation,” Opt. Express 25, 23320–23333 (2017).
270. Y. C. Huang and J. W. Pan, “High contrast ratio and compact-sized prism for
DLP projection system,” Opt. Express 22, 17016–17029 (2014).
271. G. Damberg, J. Gregson, and W. Heidrich, “High brightness HDR projection
using dynamic freeform lensing,” ACM Trans. Graph. 35, 1–11 (2016).
272. Y. Meuret, “Contrast-improving methods for digital micromirror device projec-
tors,” Opt. Eng. 42, 840–845 (2003).
273. J. W. Pan and H. H. Wang, “High contrast ratio prism design in a mini projector,”
Appl. Opt. 52, 8347–8354 (2013).
274. W. C. Ding and J. W. Pan, “High-contrast ratio and high-efficiency anamorphic
DLP illumination system design with a freeform surface lens,” Appl. Opt. 60,
6255–6263 (2021).
275. H. Urey, “P-82: diffraction limited resolution and maximum contrast for scanning
displays,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 31, 866–869 (2000).
276. O. Petrak, F. Schwarz, L. Pohl, M. Reher, C. Janicke, J. Przytarski, F. Senger,
J. Albers, T. Giese, L. Ratzmann, P. Blicharski, S. Marauska, T. von Wantoch,
and U. Hofmann, “Laser beam scanning based AR-display applying resonant 2D
MEMS mirrors,” Proc. SPIE 11765, 1176503 (2021).
277. F. Fidler, A. Balbekova, L. Noui, S. Anjou, T. Werner, and J. Reitterer, “Laser
beam scanning in XR: benefits and challenges,” Proc. SPIE 11765, 1176502
(2021).
278. K. V. Chellappan, E. Erden, and H. Urey, “Laser-based displays: a review,” Appl.
Opt. 49, F79–F98 (2010).
279. Jade Bird Display, “AmµLEDTM development kits,” https://www.jb-
display.com/development-kits.
280. E. Tang, “Smallest LED light engines for AR,” Proc. SPIE 11932, 119321W
(2022).
281. J. Reitterer, “Trixel 3 – the world’s smallest laser beam scanner for high-volume
consumer augmented reality applications,” Proc. SPIE 11932, 1193209 (2022).
282. G. Haas, “52-1: invited paper: microdisplays for wearable augmented reality —
OLED vs LED based systems (invited),” Dig. Tech. Pap. - Soc. Inf. Disp. Int.
Symp. 50, 713–716 (2019).
283. Z. Chen, S. Yan, and C. Danesh, “MicroLED technologies and applications:
characteristics, fabrication, progress, and challenges,” J. Phys. D: Appl. Phys.
54, 123001 (2021).
284. Y. H. Ju, J. H. Park, J. H. Lee, J. Y. Lee, K. B. Nahm, J. H. Ko, and J. H.
Kim, “Study on the simulation model for the optimization of optical structures of
edge-lit backlight for LCD applications,” J. Opt. Soc. Korea 12, 25–30 (2008).
285. T. Levola and P. Ayras, “Illumination of LCOS micro displays using planar
lightguides,” in 2008 2nd Electronics System-Integration Technology Conference
(2008), pp. 815–818.
860 Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics Review
286. Y. W. Li, C. W. Lin, K. Y. Chen, K. H. Fan-Chiang, H. C. Kuo, and H. C. Tsai,
“18.5L: late-news paper: front-lit LCOS for wearable applications,” Dig. Tech.
Pap. - Soc. Inf. Disp. Int. Symp. 45, 234–236 (2014).
287. K. Y. Chen, Y. W. Li, K. H. Fan-Chiang, H. C. Kuo, and H. C. Tsai, “P-181L:
late-news poster: color sequential front-lit LCOS for wearable displays,” Dig.
Tech. Pap. - Soc. Inf. Disp. Int. Symp. 46, 1737–1740 (2015).
288. J. Busch and G. Scheib, “Front lighting system for liquid crystal display,” U.S.
patent application 5,477,239A (19 December 1995).
289. E. H. Arend, S. Bhargava, H.-C. Cheng, K. R. Curtis, M. H. Schuck III, and B.
J. Sissom, “Waveguide illuminator,” U.S. patent application US17/571,366 (13
June 2019).
290. K. R. Curtis, “Unveilingmagic Leap 2’s advanced AR platform and revolutionary
optics,” Proc. SPIE 11932, 119320P (2022).
En-Lin Hsiang received his BS and MS degrees from National
Chiao Tung University, Hsinchu, Taiwan, in 2014 and 2016, respec-
tively. Currently, he is working toward a PhD at the College of
Optics and Photonics, University of Central Florida, USA. His cur-
rent research focuses on advanced display technologies including
high-dynamic-range flat panel display and light engines for AR/VR
displays.
Zhiyong Yang received his BS degree in Optoelectronic Engineering
from Chongqing University in 2017 and MS degree in Electrical
Engineering and Computer Science from the University of Michigan,
Ann Arbor, in 2019. He is currently working toward a PhD from
the College of Optics and Photonics, University of Central Florida.
His current research interests include liquid-crystal-on-silicon, mini-
LED backlight, OLED display, and µLED display.
Qian Yang received his BS degree in Physics from Nanjing Univer-
sity in 2017 and MS degree in Physics from University of Rochester
in 2019. He is currently working toward a PhD from the College
of Optics and Photonics, University of Central Florida. His current
research interests include liquid crystal SLMs for LiDAR applica-
tions, planar optics for AR/VR displays, and mini-LED and µLED
displays.
Po-Cheng Lai received his BS degree in Electrical Engineering
from Feng Chia University, Taiwan, in 2014, and MS and PhD
degrees in Electrical Engineering from National Cheng Kung Uni-
versity, Taiwan, in 2016 and 2022, respectively. His research focuses
on the pixel circuit design for active-matrix OLEDs and gate driver
circuit design for active-matrix LCDs.
Chih-Lung Lin received his BS and PhD degrees in Electrical
Engineering from National Taiwan University, Taiwan, in 1993
and 1999, respectively. He is currently the chair of the Department
of Electrical and Engineering, National Cheng Kung University,
Tainan, Taiwan. His current research interests include pixel circuit
Review Vol. 14, No. 4 / December 2022 / Advances in Optics and Photonics 861
design for AMOLED, gate driver circuit design for AMLCD, and
flexible display circuits.
Shin-Tson Wu is a Trustee Chair Professor at College of Optics
and Photonics, University of Central Florida. He received his PhD
in Physics from the University of Southern California and BS in
Physics from National Taiwan University. He is an Academician of
Academia Sinica, a Charter fellow of National Academy of Inven-
tors, and a recipient of Optica Edwin H. Land Medal (2022), SPIE
Maria Goeppert-Mayer Award (2022), OSA Esther Hoffman Beller
Medal (2014), SID Slottow-Owaki Pr ize (2011), OSA Joseph Fraun-
hofer Award (2010), SPIE G. G. Stokes Award (2008), and SID Jan Rajchman Prize
(2008). His research interests at UCF focus on augmented reality and virtual reality,
including light engines, optical systems, and display materials.
A preview of this full-text is provided by Optica Publishing Group.
Content available from Advances in Optics and Photonics
This content is subject to copyright. Terms and conditions apply.
