Laser Processing of Micro-LED
The speed of innovation and technology development for advanced display products isbreathtaking. While large-scale investment in OLED display production is ongoing in Korea andChina, there is already a new technology that may challenge LCD and OLED displays in somesegments. Specifically, displays comprised of micro-LEDs (μLED) are stepping up to thischallenge.Micro-LEDs based on inorganic III-V semiconductors (such as GaN) hold great promise fordisplays that outperform existing technology in terms of efficiency, brightness, pixel density,lifetime and operating range. Moving from LED (~200μm) to μLED (~20μm) demands technicalinnovations; assembling μLEDs into a display raises many technical challenges, and hence,creates opportunities for laser-based processes.Laser processing is at the forefront of technology in the display industry; lasers enable massproduction of the most demanding tasks. They enable the “perfect” cutting of substrates, finepatterning, low-temperature polycrystalline silicon (LTPS) annealing, temporary carrierdelamination (laser lift-off) for flexible displays, and more.Laser processing offers several opportunities for μLED display production: Laser Lift-Off (LLO) to separate the finished μLED from the sapphire growth wafer Laser Induced Forward Transfer (LIFT) to move the μLED from a donor to the substrate Excimer Laser Annealing (ELA) to fabricate a LTPS-TFT backplane Laser cutting at different levels of aggregation Laser repair of μLEDs to address yield issues and defect ratesCoherent pioneered the LLO and LIFT laser applications. This paper describes them in detail.
Laser Lift-Off
Bulk GaN substrate wafer manufacturing is difficult, and not applicable to commercial LEDproduction. Instead, GaN layers are heteroepitaxially grown on dissimilar substrate materialssuch as sapphire, silicon carbide (SiC) or silicon. The vast majority of LED production todayutilizes sapphire wafer as the growth substrate for the MOCVD due to its small lattice mismatchand relatively low cost. However, as a final carrier material it severely hampers the performanceof the GaN LED. The low thermal and electrical conductivity of sapphire as a device substraterestrict the extractable luminous flux of an LED for two reasons: 1. In a sapphire based LEDdesign, efficient dissipation of the heat generated by the LED drive current is strongly inhibiteddue to current crowding; 2. The contacts must all be connected to the front-side of the LEDresulting in unfavorable light emission characteristics during operation. Hence, a widelyaccepted route in HB-LED development and manufacturing is integration of GaN layers withdissimilar host substrates through wafer-bonding, and subsequent non-contact sapphiredelamination via LLO. For μLEDs it is easy to see that the sapphire must be removed to end upwith thin devices that will comprise flexible displays. The following schematic shows the basicprocess
The μLEDs are processed on the sapphire growth substrate to the level that a temporary orfinal carrier substrate is attached. In the subsequent LLO process, the LED wafer is exposed tohigh intensity UV laser pulses directed through the sapphire substrate, which is transparent at awavelength of 248 nm. The interfacial GaN layer absorbs the UV laser photons, heats up toabout 900°C and undergoes thermal decomposition. The thermal decomposition of GaN duringthe LLO process involves the formation of nitrogen gas and metallic gallium according to thefollowing chemical reaction:
2GaN(s) → 2Ga(l) + N2(g) (1)
The affected zone is minimized and controlled by the fluence of the laser beam; typically at thatinterface, around 10 nm of GaN is ablated and changes to liquid gallium and nitrogen gas. Thesapphire wafer is then easily removable with nearly zero force exerted on the devices. Thegallium may be washed off by water or diluted HCl acid to leave a clean surface on the device.
One important consideration for this process is the choice of the laser wavelength, it must passthrough the sapphire and then have minimal penetration into the device structure. For theGaN, InGaN materials, the 248 nm wavelength of excimer laser (5 eV) is a perfect choice; itexceeds the bandgap of GaN (3.3 eV) but experiences no absorption in sapphire. Some materialcombinations, such AlN with larger bandgap, may be processed using a 193 nm excimer laser.
Using short laser pulses minimizes heat diffusion and reduces the stress on the device layer.The typical excimer laser pulse length of 10-20 ns is perfectly suited to this need. It suppressesthermal diffusion and minimizes heat load on the device. LLO is in essence a one-shot processwhich sets a high demand on beam uniformity and laser stability. With the excellent pulse topulse stability of the excimer laser (< 1% rms), nearly perfect process control is achieved byproviding a large process window. The beam on the wafer must provide the same fluence overits whole cross section – a flat-top beam is required. Furthermore, the exact shape of the beamhitting the wafer in the LLO process must be optimized within the context of the overall processstrategy and the wafer size.
As shown in figure 4, the LLO process can be used to separate the entire wafer or only smallerfields of interest. To remove the entire sapphire wafer from the device layer, two differentprocess strategies are applied, namely LineBeam Scan or Raster Scan. Like most excimer laser applications, LLO also uses a stationary beam in combination with a moving substrate that isprecisely positioned by an x,y,z stage. For the best trade-off between throughput, quality andcost, Coherent offers different system configurations to ideally match the customers’ needs.
For high throughput, the laser beam is formed into a line beam that can cover the completesapphire wafer (2”, 4” or 6”) in a single scan. In our UVblade LLO systems, the line beam isshaped into a uniform top-hat profile with a uniformity of better than 2% sigma to ensureuniform interaction in the process. This setup, in combination with the output power (e.g.100 W, 248 nm), delivers the highest throughput. Moreover, using a beam that covers the entirewafer leads to a uniform process: Intrinsic stress in the film stack that may come from CTEmismatch is uniformly released, which further reduces the impact on the devices. For thesereasons, the line beam scan is the preferred embodiment for full wafer LLO with a requirementfor high throughput. UVblade systems cover today’s typical wafer sizes, and up to 155 mm linelength is installed in pilot production today with a roadmap to 8” and more.
Figure 5. Beam profile of the 155mm LineBeam of the UVblade @ 248 nm with short axis (SA) and long axis (LA). Notethe two orders of magnitude difference in the axis scales.
As shown in figure 5, the UVblade LLO system generates a top-hat beam in the short axis (~0.5mm) and the long axis (155 mm). This ensures that all of the process area is exposed to thesame optimum fluence. UVblade systems avoid the energy overshoot or unwanted heat loadthat is typical for processing with Gaussian beams.
Another optimized strategy for this process is to use a smaller size beam and raster scan thatacross the wafer. In this case, special consideration is given to the shape and size of the beamin order to match with the laser and the device wafer. A typical beam shape as used in ourUVblade system is e.g. 26 mm x 0.5 mm, which covers a 2” wafer in only two scans and takesfour scans for the commonly used 4” wafer. With a laser power of 30 W at 248 nm, thesesystems already reach a high throughput since the laser processing takes only 10 seconds per4” wafer. With higher power and a larger beam size e.g. 52 mm, higher throughput is achieved.High uniformity in the beam ensures a large process window and a uniform process. The rasterscan approach demands the controlled stitching of the individual shots in the scan direction aswell as the stitching between the scans. Obviously, a gap in the stitching area, or too much of anoverlap, should be minimized.
Figure 7 shows a measured beam profile for the 248 nm beam line of the UVblade. The linelength (LA) is 52 mm and the width in the short axis (SA) is about 0.5 mm. The use of high gradeUV optical materials and beam homogenization in both axes gives high uniformity of <2% in theflat-top beam area. Special attention has been spent on the exact shape of the beam and thesteepness of the profile in order to minimize the impact to the sensitive film structure andsupport seamless stitching without artefacts.
In figure 8, a possible strategy for stitching of the fields is sketched. The beam may be shapedlike a chevron in order to minimize the stress on the film structure and get optimum resultsover larger wafer areas.
Laser Induced Forward Transfer
The assembly of a high resolution display from many million μLED chips presents its ownchallenges. Lasers have been studied for this application for many years. Here we describe theconcept of using an excimer laser beam to induce the separation of selected μLEDs from acarrier and then transfer them to a receiving substrate. As described for the LED LLO process,the 248 nm wavelength is perfectly suited for precise ablation of GaN. The process generatessome nitrogen gas, which expands and induces mechanical force on the μLED structure thatpushes the chip from the carrier to the receiving substrate. Using the large beam cross sectionenabled by the excimer laser as shown in figure 6 (52 mm x 0.6 mm), in combination with amask and projection optics, allows transfers of up to 1000 dies in parallel with a single lasershot.
In a variation of the process, the μLED may also be preassembled on a temporary carrier waferor tape by using polymer based adhesives. One characteristic of these adhesives is strongabsorption in the UV. Irradiation with a wavelength of 248 nm or 308 nm will induce photochemical decomposition of the adhesive, which releases the μLED chip and also induces forcesto push the die to the receiving substrate. The energy density that is required to initiate therelease from a polymer tape or adhesive can be 5-20 times lower than the required fluence to
separate the inorganic III-V semiconductors. This means that very high process speeds can beachieved with moderate laser powers.In summary, excimer lasers have excellent potential for emerging laser applications in the μLEDspace. Their unique characteristics such as short UV wavelength, and short pulse length,combined with high energy and power, are well suited for the III-V material systems commonlyused in LED manufacturing. The 248 nm excimer in particular far exceeds the performance of266 nm or 213 nm solid-state lasers. Given this, the excimer enables significantly differentiatedprocess strategies that are highly productive, and therefore cost effective.
扫一扫在手机上阅读本文章
