MicroLEDs: The Next Revolution In Displays?

Technology offers improved brightness, colors, and lower power, but they’re expensive and difficult to manufacture.


Flat-panel display technology is exploding on several fronts as more screens are required for more devices. But one type of display is generating an enormous amount of buzz in the market—microLEDs.

Dozens of companies are working on micro-light emitting diodes (microLEDs), a technology that promises to provide better and brighter displays than current solutions in the market. Apple, Facebook, Samsung and TSMC are just a few of the companies developing microLEDs. Additionally, TV vendors, display makers, startups and equipment vendors are also jumping on the bandwagon.

In R&D for several years, microLEDs are used to make displays for TVs, watches and head-mounted systems like the Google Glass. But microLEDs are difficult to make and expensive. That’s why mainstream commercial displays using microLEDs are not available today and won’t appear for several years.

A microLED is basically an LED (light-emitting diode), which converts electrical energy into light. Traditional LEDs are used for backlights in LCD displays, billboards, consumer electronic items and lighting. MicroLEDs are not to be confused with so-called miniLEDs, which are basically scaled down versions of today’s LEDs.

MicroLEDs are microscopic versions of an LED without a package, and a multitude of them need to be incorporated into a display. One microLED measures less than 100μm (less than 50μm are common), which can be 1/100 the size of a conventional LED. MicroLEDs are self-emissive and don’t require a backlight. In theory, a display using microLEDs provides more color and higher brightness with lower power than today’s displays.

Fig. 1: SEM micrograph of 3-µm size/ 5-µm pitch microLED array fabricated with direct bonding approach. Inset is an optical photograph of microLED switched on. Source: Leti

MicroLEDs are difficult to implement. For example, to develop just one HDTV, the system requires 6 million individual microLEDs. So in a fab, 6 million microLEDs must be manufactured and then transferred onto a backplane in the TV without an error. Making microdisplays using microLEDs is also daunting.

“MicroLEDs are truly the ultimate display. It’s just very hard and expensive to make,” said Max McDaniel, vice president and chief marketing officer for the Display And Flexible Technology Group at Applied Materials. “You have LED displays like you see on the side of the highways. For those, every pixel is one LED. They are on the millimeter scale. Those are called LED displays. MicroLED is where you shrink them down to the scale of tens of microns. You place one in each pixel. It’s so much smaller and harder to do. It’s harder to physically put them where you want them to be. It’s also harder to make the LEDs themselves so that they perform well.”

A multitude of companies are trying to solve these and other problems, but the industry requires some new breakthroughs.

Display mania
In total, the worldwide display market is expected to grow from $150 billion in 2019 to $228 billion by 2028, according to Touch Display Research.

But the display market is also in the midst of oversupply and falling prices. In 2019, display equipment spending is projected to fall 38%, to $13.4 billion, due to “fab delays resulting from weakening market conditions, funding challenges or other issues,” according to Display Supply Chain Consultants (DSCC).

Today, the display market is dominated by two technologies—liquid crystal displays (LCDs) and organic-light emitting diodes (OLEDs). An LCD consists of thin-film transistors (TFTs), which determine the resolution of the display. Meanwhile, based on organic materials like carbon and hydrogen, OLEDs enable self-emissive displays, which are more efficient than LCDs.

Today, the majority of cell phones and TVs use LCD displays. “In the mass consumer market today, it’s high-resolution LCD phones and it’s 55-inch and smaller LCD TVs,” Applied’s McDaniel said in a presentation at the recent Display Week Symposium. “You also have products that are making the transition toward OLEDs.”

LCDs are cheap products built in giant fabs. Today, the largest LCD glass size being built is based on a Gen 10.5 technology. The term “Gen” or generation denotes the glass size. “LCD displays are being made in these Gen 10.5 fabs, which is essentially a 3- by 3-meter piece of glass. It’s a garage door size piece of glass. You can make six 75-inch TVs on one piece of glass or eight 65-inch TVs,” McDaniel said in an interview.

OLEDs are making inroads in the TV market, but they are still expensive. OLEDs are gaining steam in smartphones. In 2018, 27% of all cell phones used OLED displays, but that figure is expected to reach 54% by 2023, according to DSCC.

“OLED technology is attractive. It’s attractive for consumers because of the image quality and form factor,” Applied’s McDaniel said. “It also provides the panel makers a roadmap toward more advanced foldable or rollable devices.”

Other display technologies are also in the works, including miniLEDs and microLEDs. Both technologies are LEDs, which use inorganic materials such as gallium-nitride (GaN).

Conventional LEDs are efficient products with high-peak luminance properties. LED lifetimes exceed 50,000 hours with a luminance measured in the tens of millions of nits, according to Ioannis Kymissis, a professor at Columbia University. “Nits represent the relative intensity of light normalized for human perception,” Kymissis said during a presentation at Display Week.

LEDs come in different configurations, such as monochrome and multi-color. An RGB (red, green, blue) LED, one popular type, consists of the primary colors in the gambit. So, they can create a number of different colors.

A miniLED is a smaller version of a traditional LED and ranges in size from 100μm and above. Like LEDs, miniLEDs are targeted for backlights in displays.

The big buzz revolves around microLEDs. “These are very small. The physics are the same. It’s an LED. It’s a PN diode. When you put current in, you will have light,” explained Francois Templier, research director at Leti.

Like LEDs, a microLED can be configured with only one color (i.e. red LED or red microLED). MicroLEDs also can be configured with the three main primary colors—red, green, blue (RGB). For this, each color in an LED—red, green, and blue—is called a sub-pixel. “You can make a monochrome display. You can make a display with only blue LEDs or only green LEDs,” Templier said. “When we want to make a color display, we need to have three basic colors, which are blue, green and red. This means each pixel should have these three sub-pixels using these three colors.”

A multitude of microLEDs are required to make a display. Nonetheless, microLEDs have higher resolutions and a greater luminance than OLEDs. MicroLEDs have a maximin pixel per inch (PPI) of 5,000 with 105 nits, compared to 3,500PPI and ≤2 x 103 nits for OLEDs, according to LG Display.

MicroLEDs can be incorporated in two product types—microdisplays as well as mid- to large-sized displays. A mid- to large-size display includes smartphones, watches and TVs.

Microdisplays include products that resemble the Google Glass, as well as other augmented reality (AR) glasses. A microdisplay using microLEDs might consist of 100,000 pixels at a size of 10mm x 10mm.

“A lot of companies consider that AR glasses will be your future cell phone,” Leti’s Templier said. “For the microdisplay, there is one main reason for microLEDs. We want higher brightness for AR. The requirement for the brightness is 100x more than existing technology. You can have microdisplay using OLED. They typically provide 1,000 nits. For AR, you need 100,000.”

TVs are a different story. “The main reason is image quality. With microLEDs, you can have better color than OLEDs or LCDs,” he said.

Several companies already have demonstrated microLED TVs. In 2012, Sony demonstrated the Crystal LED display, a 55-inch prototype microLED TV. It incorporated 6 million microLEDs.

Last year, Samsung demonstrated “The Wall,” a 146-inch microLED TV. Earlier this year, Samsung demonstrated a new 75-inch display as well as a 219-inch version of The Wall using microLEDs.

Many others have also demonstrated the technology with some even racing to develop commercial products. “Samsung is trying to bring a microLED TV to the market this year. Hisense and TCL also demonstrated one as well,” said Jennifer Colegrove, chief executive of Touch Display Research, in a presentation.

At the recent Display Week conference, there were several papers on microLEDs. Among them were:

  • AU Optronics devised a 12.1-inch 169-ppi full-color microLED display using LTPS-TFT backplane.
  • Kyocera has developed a small-size microLED display by LTPS high integration technology with the target more than 200-ppi, which is on the same level as achieved by LCD and OLED.
  • Sharp reported on a novel microLED display bonded onto a silicon driver, which it calls “Silicon Display.” A 0.38-inch full-color display with a resolution of 1,053-ppi was demonstrated.

The initial products are expected to be expensive amid a multitude of challenges with the technology. “MicroLED displays are still in the development phase and no consumer products are available yet,” said Eric Virey, an analyst at Yole Developpement, in a paper.

“Unlike OLED, inorganic LEDs can’t be deposited and processed over very large areas. LEDs are grown on 4- to 8-inch wafers and the art of making microLED displays therefore consists in singulating individual emitters and transferring and assembling them onto a backplane substrate,” Virey said. “For most consumer displays such as TV or smartphones, microLED with die size ranging from 3 to 10μm are required to ensure cost compatibility with the applications. For an 8K display, close to 100 million of those must be assembled without a single error with a 1 to 2μm placement accuracy at a throughput exceeding 100 million units per hour. Transfer and assembly are therefore often seen as the single largest technical challenges to overcome to enable microLED manufacturing.”

That’s not the only problem. “For example, while the external quantum efficiency (EQE) of traditional LED can reach 70% or more, the EQE of small microLED (size <5μm) was until fairly recently limited to 1% to 5%. At those levels, microLED can’t deliver on the key promise of better efficiency than OLED. Fortunately, dramatic progress has been reported by various groups over the last 2 years,” Virey said.

Making microLEDs
In the fab, meanwhile, there are some similarities and differences between making microLEDs and traditional LEDs. “MicroLED and semiconductor LED technologies such as miniLED have a similar LED manufacturing process, but the process subsequent to LED manufacturing is very different,” said Mukund Raghunathan, product marketing manager at KLA. “To achieve the size shrink of sub-100μm such as 20μm or lower, the fabs may need different process equipment and a much cleaner cleanroom environment. At 20μm, the ability to tolerate micron-level defects is much lower compared to a miniLED, as the microLED transfer and repair process is very costly.”

There are various ways to make a display using microLEDs. The process flow depends on the display type. In simple terms, the first step is to make an assortment of microLEDs on an epitaxial substrate. The devices are diced, tested and then transferred to a backplane using mass-transfer techniques.

The first step is to make the LEDs themselves. A traditional LED is made using a metal-organic chemical vapor deposition (MOCVD) system. In this system, thin layers of GaN materials are epitaxially deposited on a wafer.

MOCVD is also used for microLEDs. In one example of this process, the Hong Kong University of Science and Technology and others demonstrated a flow where an n-GaN layer is grown on top of a sapphire substrate, followed by a multiple quantum well (MQW) layer and a p-GaN layer. Another way is to deposit the layers on 200mm silicon substrates.

Each method is challenging. “Achieving wavelength uniformity and low defect density to reduce production costs are key factors to successfully apply MOCVD in microLED technology,” said Somit Joshi, vice president of marketing for the MOCVD division at Veeco. “The biggest challenge is generating high quality epi consistently across a large wafer population to meet the single bin requirements for wavelength and brightness uniformity. Since microLEDs require a very tight uniformity across a large transfer area, the requirements are much more stringent than LEDs that are individually packaged. Less than 10 dead pixels are allowed in a display to fulfill the general standard of the display industry. Thus, the yield of LED epitaxy must be very high to reduce the possibility of dead pixels.”

There are other challenges. “Sorting and binning are the methods to enhance wavelength uniformity for conventional LEDs,” Joshi said. “But microLEDs are too small to be sorted and binned. Therefore, the uniformity of LED epitaxy is even more critical. The requirement for epitaxy in conventional LEDs is around 8 to 10nm. In comparison, the general requirement of color uniformity for displays is to reach 1 to 2nm across the display depending on the type of display. It is impractical to achieve a wavelength uniformity of 1 to 2nm across the wafer. With the appropriate transfer technology, the uniformity requirement of microLED can be relaxed to 3 to 5nm across the wafer. This 3 to 5nm uniformity requirement can be met using advanced MOCVD tools.”

Once the layers are grown, the p-GaN and MQW layers are etched to insulate the pixels, according to the Hong Kong University of Science and Technology. Then, the last step is to build connecting pads.

From there, the goal is to take each microLED and transfer it to a TFT backplane or another surface. For this, there are various approaches, such as monolithic and pick-and-place.

For tiny microdisplays, the industry uses the monolithic approach. “What we call monolithic is when we fabricate the display on one substrate,” Leti’s Templier said. “In monolithic, you start with an IC driver. It’s on a CMOS wafer. You grow the LED on top of it. Then, you pattern the LED. You fabricate at the wafer scale. You make several LEDs and then singulate them in the end.”

In contrast, the pick-and-place approach is used for mid- to large-sized displays. In one example, the microLEDs are fabricated using three different so-called epiwafers. The blue microLEDs are made on one epiwafer, while the red microLEDs are made on another epiwafer. Another wafer is for the green microLEDs.

Then, each microLED is diced and transferred onto a TFT backplane using a high-speed pick-and-place system. Some systems can pick-and-place 10,000 LEDs at once, which helps speeds up the process.

Fig. 2: General approach for fabricating microLED displays. Source: Leti

There are other approaches. For example, using what it calls microtube technology, Leti has fabricated a blue and a green prototype display with 40 x 40 pixels at 210μm pitch on a passive matrix.

“First, we process red, green and blue microLEDs with a 25μm side size on different epitaxial materials. MicroLEDs are integrated with N and P contact metallic pads on top, and singulated at a pitch of 210μm,” said Jeannet Bernard and others at Leti, in a paper. “In parallel, we fabricate an interconnection passive matrix with microtubes on the top side with the same pitch size. Then we transfer the microLEDs by flipping the substrate and hybridizing pads on microtubes. The transfer is finalized by removing the epitaxial substrate from the microLEDs. We repeat this transfer step for the 2 remaining colors and thus we obtain a RGB display with a pixel pitch of 210μm.”

Regardless of the approach, the industry faces some challenges. “Typical microLED displays combine conventional LEDs based on sapphire substrates with thin-film transistor (TFT) logic, or in some cases with CMOS,” said Martin Eibelhuber, deputy head of business development at EV Group. “The key challenge is to integrate the LED wafer with CMOS while providing high mechanical stability together with optimal electrical performance to achieve full functionality. To achieve this, a combination of very high precision alignment and high integration of the process flow is required to enable very small pixel dimensions.”

There are other challenges. “These approaches are not yet cost-effective for the consumer market,” KLA’s Raghunathan said. “For instance, while the mass pick-and-place approach enables selective replacement of a defective pixel, which includes dead or dim pixels, it comes at a cost of additional mass transfer steps to replace the defective pixel. Likewise, in a monolithic process, repair and replacement of a defective pixel increases the cost because it requires additional wafer-level processing–first to remove the defective pixels and then to substitute them with non-defective pixels.”

So is microLED technology ready for prime time? Not yet, but it may be in the future.

“Cost reduction is critical for commercialization of microLED applications,” said Steve Hiebert, senior director of marketing at KLA. “This cannot be achieved until the yield at each stage of microLED production is improved. The sources of the yield issues are incoming substrate, epi processes, microLED fabrication processes, and transfer processes, although there will be some uniqueness depending on whether the manufacturing method is a mass pick-and-place process or a monolithic fabrication process,”

Clearly, microLEDs are promising, but there are too many challenges to make it practical today. So OEMs will use LCDs and OLEDs. But if the industry can make microLEDs work, it could turn the market upside down.


Sylvain Muckenhirn says:

Gd morning Mark,
Great summary of microLED status. Thank you.
As you mentioned, I think the LED industry is facing a shift from a technology running with a small amount of manufacturing control and at a low production cost (general lightning where final test sorting can be average and repair costs are not a killer) to a technology requiring more manufacturing control on intermediate steps (active layers deposition) and a rock solid wafer test before die transfer to avoid high repair costs.
Active layer deposition control requires buried defect detection and counting, and optical signature (emission wavelength, emisssion intensity, composition) in the nano-meter range resolution.
Final wafer test after singulation can then focus on optical signature (emission wavelength, emission intensity).
All of the above (especially the later as it is a 100% test sampling) requires speed and as such likely exclude contacting method.
As such, quantitative cathodoluminescence (CL) seems to provide an answer. Already proven on active layer deposition and running in production environment, it is evaluated on singulated wafer final test. Non contact excitation, direct characterization of individual die optical signature, and with proper implementation, speed.
Now, I may be partial to the problem, as i am helping a Swiss company achieving these goals, but in answer to our colleagues in Veeco and KLA, we can say that CL offers a wafer level solution which can inspect sub-micron level defects in a non-destructive way, allows a fast full wafer overview scan in full wafer brush scanning mode, and allows to review defective zones with nm resolution in step-and-repeat scanning mode, as well as singulated final wafer testing.
If the transfer induces defects, displays themselves can be inspected by CL technique. However, while it would identify defective dies, it does not solve the high repair cost. For this side of the problem, redundancy may be a safeguard.
Thank you,
Best regards

Leave a Reply

(Note: This name will be displayed publicly)