The market for MicroLED displays is heating up, fueled by a raft of innovations in design and manufacturing that can increase yield and reduce prices, making them competitive with LCD and OLED devices.

MicroLED displays are brighter and higher contrast than their predecessors, and they are more efficient. Functional prototypes have been developed for watches, AR glasses, TVs, signage, and automotive displays, which were shown at CES and Display Week. Vendors now are shifting their focus from just getting prototypes out the door to ramping yield and productivity for all process steps, which should significantly reduce costs.

“The stakes are being raised when you have devices like microLEDs, where the biggest question has been, ‘Can you produce a microLED display in a cost-effective manufacturing process?’” said John Ghekiere, vice president for product and technology at ClassOne Technology.

Fig. 1: Blue GaN microLED arrays with 3um pixel pitch use polychromatic quantum dot integration to achieve full color AR displays. Using a sub-40nm CMOS process, maximum brightness of 10M nits consumes <1W. Source: MICLEDI

That’s a big question, because there is no shortage of possible applications for this technology. “One critical opportunity involves the development of GaN on silicon-based technologies,” said David Haynes, managing director of strategic marketing at Lam Research. “These allow the scaling of manufacturing processes to 200mm and 300mm wafers and create access to the well-established silicon foundry infrastructure. Foundries are already developing standard CMOS-compatible GaN on silicon solutions for power and RF, so fabricating microLED is a logical next step.”

Still, the magnitude of this shift is non-trivial. “The principal challenge is meeting the stringent yield targets for uniformity (color and brightness) and defective pixels at an acceptable manufacturing throughput and cost,” said Ajit Pananjpe, vice president and CTO of Veeco. “Efforts are focused on improving the yield at all key steps, including epi growth and mass transfer, to minimize the need for redundancy or repair. These factors drive display costs.”

Leading tech companies like Apple, Samsung, and Konka (formerly Toshiba) are forging ahead and investing heavily in making microLED modules for AR glasses, smart watches, televisions, digital signage, and automotive displays a reality. This is expected to happen between 2024 and 2027, and is likely to lead to the usual precipitous price drop in display technology as the technology matures. As of today, the least expensive of Samsung’s The Wall TV, available in 89-inch, 101-inch and 110-inch versions, has a starting price of $80,000 (product roll-out is delayed). It is fabricated using PlayNitride’s 30 x 60µm microLEDs on passive matrix thin film transistor (LTPS) drivers. Making microLEDs cost effective is all about bringing that $80,000 price down to a consumer-friendly level of around $3,000.

At the other end of the size spectrum is Vuzik’s AR Shield glasses for enterprise applications (surgical, maintenance, etc.), with a 0.13-inch green display (640 x 480 pixel), fabricated by JB Display with 2 displays with waveguides per pair of glasses, which achieve a brightness needed for sunlit conditions of 4M nits.

Massive investments
The outstanding performance potential for microLEDs is driving large investments for research and development. These directly-emissive displays promise higher pixel density at lower power consumption, nanosecond response times (critical for augmented reality), and wider viewing angle than is possible with organic LED (OLED) displays or LED-backlit LCDs.

The high precision needed to make microLEDs requires a wholesale move to a semiconductor manufacturing mindset including a class 100 cleanroom (<100 particles/ft3, ≥0.5µm), a switch to 200mm sapphire wafers, and scaling of critical epi wafer tools. Upgrading to EUV scanners may be needed for leading-edge microdisplays, along with low damage etching, electroplating, CMP, and wafer cleaning tools.

In the coming three years, Yole projects $3 billion is likely to be spent on microLED manufacturing infrastructure for TVs, smart watches and AR glasses. Players include Apple, Samsung, Epistar, PlayNitride, AUO, Sanan, JB Displays, and many others.

Eric Virey of Yole Developpement detailed some of the announced microLED fabs plans including:

• amsOsram is spending $850M to expand microLED production in Austria and Malaysia;
• Aledia announced $165M fab to make GaN-on-Si nanowire microLEDs;
• AUO is increasing its investment in PlayNitride and Epistar;
• Konka, TCL-CSOT, and Visionex announced a $100M investment;
• MICLEDI and GlobalFoundries are gearing up for microLED production, and
• San’an Optoelectronics announced $2B fab for mini/microLEDs.

The Apple smart watch is likely to be the first mass-produced microLED product, according to Virey, and Epistar is its preferred supplier. “If everything goes according to plan, we will see Apple’s microLED smart watch in 2024.”

Meanwhile, novel device structures are delivering impressive performance. Aledia engineered a novel 3D architecture by forming millions of submicron diameter light-emitting GaN nanowires that project up from the silicon surface to improve luminous area efficiency. Aledia will fabricate the nanowires, called SmartPixels, and employ a silicon foundry for subsequent processing.

In addition to the above investments, there have been significant acquisitions in microLEDs, including Google’s purchase of startup Raxium, and Snap’s acquisition of Compound Photonics.

LEDs go mini, then micro
There are three flavors of microLEDs based on product size:

• Large displays for TV and signage;
• Medium-sized displays for tablets, smart phones, and in-car displays, and
• Microdisplays for XR (AR/VR/MR) glasses and smart watches.

MicroLEDs are miniature versions of the LEDs we use today in streetlights, household bulbs and commercial buildings. The way the industry fabricates white light is by coating blue LEDs with a yellow phosphor that interacts with the blue photons to create white photons. MicroLEDs generally are specified as being <50µm on a side, while LEDs feature 200µm or larger dimensions, and the term miniLED covers everything in between. However, some companies say that microLED refers to any LED without a package, regardless of exact dimensions.

MiniLEDs today use blue LEDs and quantum dots to provide illumination in laptops, tablets, TVs, and indoor/outdoor signage. LEDs emit light (photons) when current is applied at appropriate voltage, and electrons and holes recombine in the active region of the device — the quantum wells. The light’s brightness is a function of applied current, but the emitted color (wavelength) is determined by the difference in energy bands of the semiconductor materials, with AlInGaP-on-GaAs making red light, InGaN-on-sapphire making green and blue.

LED color consistency is reflected in wavelength consistency. For LEDs, binning separates products with better controlled chromaticity from lesser performing devices. Binning is needed to achieve consistent wavelengths to produce lights that match one another. Unfortunately, with microLEDs, binning is not possible due to lack of a package that protects it during handling. As a result, in-spec microLEDs must be segregated for use in the target display, which is tied to transfer and in-line inspection operations.

First-generation microLED displays are expected to use sub-pixels of native red, green, and blue microLEDs, although some companies are pursuing use of blue LEDs with quantum dots to create blue and red sub-pixels. This down-converting approach simplifies the complicated pick-and-place transfer of red, green and blue microLEDs, but at the expense of luminosity loss.

Picking the right process
The fabrication scheme for microLED displays depends on its requirements. For instance, microdisplays used in watches or AR applications require extremely high pixel density of 1,000-10,000 dpi. In these cases, hybrid bonding brings together the GaN/sapphire wafer onto a silicon CMOS driver wafer. Backgrinding then is used to remove the sapphire wafer.

However, for lower pixel density applications, such as TVs or digital signage, the miniLED chips must be spread out onto a display backplane. That approach often utilizes temporary bonding on a carrier wafer to enable redistribution from tight die-to-die packing on the wafer to separation by pixel pitch.

Partnerships between chipmakers and microLED fabricators also enable faster integration of processes. “MICLEDI’s microLED solution, combined with GlobalFoundry’s 22FDX platform, addresses the demanding needs of future AR glasses by providing ultrahigh resolution displays and advanced imaging that make stunning visual detail and color possible,” said Ed Kaste, vice president of industrial and multi-markets at GlobalFoundries, in a statement. “Demand for AR and VR products will soar as users experience more immersive augmented reality.”

MICLEDI, an imec spinoff, is the first company to build microLED arrays for AR on a 300mm CMOS platform. The wafer-to-wafer hybrid bonding step reconstitutes RGB microLEDs with the CMOS backplane. For high-efficiency waveguide integration, the company uses a Fresnel lens atop the LEDs to provide beam shaping. Designing light, attractive AR glasses depends on manufacturing smaller-sized microLEDs at a cost that can be absorbed comfortably into the product’s price.

Another example is the collaborative work among companies taking place at Fraunhofer Institute for Electronic Nano Systems (ENAS). “Our plating tools are already in pilot production with several manufacturers, but success goes back to the integration scheme, said ClassOne’s Ghekiere. “Can we make it work in an optimal way to make our customer’s products price competitive while giving them technological advantages? Our ability to work with Fraunhofer to fill some of those interstitial spaces between the unit processes is really the driver — the plating works, the CMP works, the wafer alignment works. Now can we enable integration, because you’re going to have teams that are competing here — some of the biggest, most valuable companies in the world.”

Perfecting color emissions
MicroLEDs are direct emitters, so they are used as bare die. This is different from LEDs and miniLEDs, which are assembled in a surface-mount package, wire bonded in place, and then encapsulated in epoxy or silicone.

While many firms are pursuing native red, green and blue chips assembled from different wafers, Porotech — a spinoff of Cambridge University — has devised a way to produce all three colors (red, green, blue) on a single wafer. “You can have a pixel that can give you any color at any time, which greatly simplifies and reduces the cost of assembly operations,” said TongTong Zhu, CEO of Porotech. The company is seeking industry partners to aid in developing, scaling, and commercializing the technology. It works by creating a porous layer beneath the wafer surface, which increases the amount of indium that can be doped into the lattice. Subsequently, the quantum wells are grown using MOCVD without introducing any new defectivity, according to Zhu. Porotech’s first partner is IQE, an epi wafer supplier that will act as a foundry for PoroGaN GaN-on-sapphire wafers.

“We are enabling different colors, but also color purity. For instance, our green at 555nm has less than 5nm color shift over different drive conditions,” said Zhu. The technology enables significant improvements in native red microLEDs, which are notoriously inefficient. The color varies more dramatically as a function of temperature than green or blue microLEDs.

How am I driving?
For its smart watch, Apple is going with a silicon CMOS based microdriver,” said Yole’s Virey. “Samsung is using TFT on glass, except it is very complex with 24 mask levels, so yield is pretty low and cost is high. Traditional TFTs for LEDs require 12 mask levels. We are seeing more papers that are talking about silicon drivers, but it may not be the best option for all applications.”

“For AR glasses, power consumption is directly related to driving the microLED display. Low power is critically important since it defines battery size, battery life, and heat dissipation requirements,” said Soeren Steudel, co-founder and CTO of MICLEDI Microdisplays. “Clever driver and control techniques can reduce consumed power, as well as duty cycle management and digital pulse-width-management techniques.”

Process details
State-of-the-art fab tooling is needed to process microdisplays with digital lenses for AR glasses. “Only light with a certain emission angle can be used, which is ± 20° for waveguides. We need to optimize on a pixel-by-pixel level to ensure as much light as possible is emitted within the APEX angle. This significantly impacts system efficiency, especially wall-plug efficiency. A larger pixel correlates with higher lens efficiency.

“High brightness and high resolution are needed for AR glasses, where resolution is directly related to the size of the pixel,” said Steudel. He noted that for full HD (1920 x 1280 pixel) projection, the maximum pixel pitch is 3µm or smaller. “Silicon foundries employing the most advanced 300mm wafer steppers are able to consistently achieve very accurate overlay (<20nm) in an organic high index material. Implementing highly reproducible multi-layer digital lenses at pitches <3µm requires extremely fine features (CD <100nm).” To meet these tight tolerances on a reproducible basis requires advanced photolithography, etch, and CMP tools, according to Steudel. “Additional challenges are very tight specs in terms of etch slopes, layer uniformity, and thickness control.”

Microdisplays often use wafer-to-wafer hybrid bonding, a relatively new fab process, to bond GaN devices to a silicon substrate, which can then undergo further processing on 200mm or 300mm tools. Within the steps for hybrid bonding, processing of flat wafers is essential. “CMP is critically important. MicroLED wafers built in 300mm lines must be ground and flattened to incredibly tight tolerances to be bonded to the ASIC backplane wafer before dicing into finished microLED display modules,” Steudel added.

Unfortunately, as the size of microLEDs scales below 10 µm, the external quantum efficiency drops. “Reduction in luminous efficiency due to carrier recombination on the sidewalls of the microLED is an issue with scaling,” said Veeco’s Pananjpe. “Etching, cleaning, and passivation steps for the mesa etch that is used to define the microLED have to be optimized to minimize surface recombination of carriers. Often, direct etching is used instead of lift-off patterning for finer dimensional control.”

Fig. 2: External quantum efficiency decreases with shrinking microLED dimensions. Source: Yole Développement

Performance of microLEDs is reflected in metrics such as the internal quantum efficiency (IQE), which reflects how efficiently electron and hole combinations in the device convert to luminance. EQE, or external quantum efficiency, measures the success rate at converting electrons passed through the LED to luminance. “We’ve developed ultralow damage GaN etch processes by combining extremely low-power, steady-state plasma etch steps with atomic layer etch processes in the same module,” said Lam’s Haynes. He added that such processes can reduce plasma damage of the GaN surface while offering high throughput operation.

Fig. 3: To meet the near-zero tolerance for bad pixels, microLED fabs are stepping up their inline metrology, automated optical inspection, and testing protocols. Source: KLA

Process flow
The most fundamental part of microLED fabrication involves growing the epitaxial layer stack using 150mm or larger metalorganic CVD (MOCVD) tools, offered by Aixtron and Veeco. “The epitaxy process step must be very well controlled to minimize yield limiting particles, pits, and scratches,” said John Robinson, senior principal scientist in KLA’s Industry and Customer Collaborations group. “Epitaxial layers require tight thickness, composition, and stress control in order to maintain the required wavelength uniformity. In-line inspection and metrology are critical to epi quality. At microLED chip formation, controlling patterning defects is essential for high yield.”

Engineers quantify luminosity using both electroluminance-based (EL) and photoluminescence-based (PL) testing. Both are non-contact methods that detect defects not caught by optical inspection. EL is the more established electrical inspection method. PL is relatively newer and measures the spectral properties of light, looking for defects in emission including the wavelength and PL intensity, which are measured after epitaxial deposition and during wafer acceptance testing. KLA offers a variety of optical inspection tools. Companies offering PL and EL toolsets include Corbeau Innovation, Hamamatsu, InZiv, Toray, and others.

Fig. 4: The OmniPix 2.0 combines optical inspection, PL and EL with 100nm inspection resolution on wafers up to 300mm. Source: Corbeau Innovation.

After the quantum wells are formed by multilayer epi, stepper patterning and etching forms the device regions. Lithography and etch also define the N and P contact pads, followed by deposition of transparent indium tin oxide (ITO), which spreads the current across the surface of the device. Reactive ion etching then exposes the contact pads. The wafer is flipped onto a flexible film, followed by wafer grinding and CMP to remove the substrate.

“The transition to stepper lithography requires CD and overlay process control. In addition, wafer shape non-uniformity and stress may result from film depositions and require process control to meet specifications for wafer flatness,” said Johnson.

Passivation of the microLED devices has become a critical step, because it can help in reducing sidewall carrier degradation. “ALD is becoming a standard process for passivation, which provides significant improvement in external quantum efficiency,” said Yole’s Virey.

Before components are transferred, accurate wafer level maps indicate which microLEDs are defect free and functional within the target specifications. “The maps are critical input to the transfer process,” said Johnson, adding that inline inspection and metrology are equally critical during backplane design. “Anode/cathode electrical testing is also important to ensure performance leading up to transfer.”

Next, hundreds or even thousands of microLEDs are transferred from the wafer on which they were fabricated onto a carrier (interposer), or directly to the display backplane.

Transfer technologies
Many methods are being explored for mass transfer, including stamp, laser, roll-based, and self-assembly using fluid. A transfer yield of 99.9999 (1 ppm) is needed for volume manufacturing. Among these, stamp-based transfer is the most mature and popular, such as in X-Celeprint’s Micro Transfer Printing process. Indeed Samsung is using stamp-based transfer for The Wall TVs.

Stamps are fabricated using injection-molding of polydimethysiloxane (PDMS), and a smooth layer with with tiny PDMS posts are created by photolithography and etching. Typical stamp size is 37 x 37mm or 49 x 49mm. An adhesive ink grabs the array of microLEDs and they are placed using lasers or other means. Companies such as ASM Pacific Technology, Shibaura, Toray and X-Display offer these systems.

Success using stamp-based transfer recently was reported by the EU-funded MICROPRINCE, which provides a foundry pilot line for heterogeneous integration. “We demonstrated printing or transfer yields of up to 99% with misalignment below 1µm. Devices with sizes as small as 100 x 100 x 5 µm were effectively transferred and stacked on CMOS target dies,” said Sebastian Wicht, MICROPRINCE coordinator and program manager of transfer printing at X-FAB MEMS Foundry.

To speed throughput further, several companies are working on multi-head stamp systems, stamps with larger footprint, or two-step transfer flows, according to Virey.

Fig. 5: The stamp- and laser-based transfer methods of microLEDs to backplane offer comparable throughput. Source: Yole Développement

Laser-based transfer tools are offered by are offered by 3D Micromac, ASMPT, K&S, Shin Etsu, Toray, and others. Device separation works by ablating a microscopic layer of GaN, which forms an expanding nitrogen layer to facilitate lift-off. At the field of view of 8 x 2mm2, for instance, a laser pulse can transfer groups of microLEDs with high accuracy (+-1.5um). High in selectivity, laser processes require further development to meet the throughput needs for displays.

A third alternative for microLED transfer uses self-assembly of microLEDs in a fluid medium, where pre-tested microLEDs and an oscillating motion encourages the devices to settle in small wells. ELux Display, a spin-off of Sharp, is fine-tuning the process. “We’ve seen recent interest in the self-assembly method, so I would not count it out,” said Virey.

Following transfer, SMT bonding processes, using solder deposition and reflow, bond the chips to the backplane. Here, automated optical inspection and PL and EL are tested. When defective microLEDs are detected, lasers or other means are used to remove and replace the die, called repair. At final display level, the panel maker performs AOI, EL and PL again.

Conclusion
Though microLEDs displays are still years from commercialization, the biggest names in technology are bringing together expertise from the semiconductor, optoelectronics and display worlds to make it happen. Innovative, out-of-the-box approaches such as the single GaN substrate for all colors from Porotech, to nanowire illumination from Aledia, encourage the industry to think big when it comes to microLEDs.