Beyond Amorphous-Silicon: New Developments in High-Mobility Backplanes

New advances in laser-crystallization technology are enabling larger liquid-crystal and OLED displays.

by Ulrich Hausmann and David Knowles

CONSUMER DEMAND for mobile devices, tablets, and televisions is driving the need for new, more advanced displays. In fact, in devices such as the iPhone and the Samsung Galaxy phone, the display has become one of the key differentiators. These advanced displays are characterized by high pixel densities or by the use of organic light-emitting-diode (OLED) materials. They depend on thin-film-transistor (TFT) backplanes that are typically fabricated from low-temperature polysilicon (LTPS) instead of the more common amorphous silicon. Polysilicon provides much higher electron mobility, but creating it has been one of the more challenging manufacturing steps in the display process. The laser-annealing equipment required for LTPS has garnered a reputation for high cost and variable yields.

Recently, there have been a number of significant advances in LTPS backplane technology. The past few years have seen greater competition among laser-annealing equipment suppliers, leading to a number of innovations. There are also several alternative approaches under development to replace polysilicon. This article will review the progress of both laser annealing and alternatives, which together promise to greatly reduce the manufacturing complexity and speed the adoption of advanced displays. Related key trends and driving factors to be discussed include ultra-high-resolution LCDs, increasing LCD frame rates and display sizes, and the explosive growth of the tablet market.

Ultra-High Resolution, Higher Frame Rates, and Larger Sizes for LCDs

As everyone knows, in 2010 the huge success of Apple's iPhone 4 fundamentally altered the smartphone market by capturing an unprecedented market share. A key feature of the iPhone 4 is its 3.5-in. Retina display with a high resolution (960 x 640 pixels or 326 pixels/in.) that shows very sharp clear images and provides a comfortable viewing experience with clarity rivaling that of print media. LTPS technology, with its very high mobility of approximately 100 cm2/V-sec, enables such performance by reducing the transistor size in the display's active area. This compares very favorably over amorphous-silicon technology with a mobility on the order of 1 cm2/V-sec. This allows for a larger aperture ratio, which is very important for display brightness. The large aperture and low-bus-line loading structure of LTPS also helps to minimize power consumption, which is critical for smartphone displays.

Attaining higher frame rates for larger displays and for driving high-speed OLED displays will require the same high-mobility transistors and low-resistance bus-line technologies required for smartphones. LCD- and OLED-TV development is also demanding faster switching speeds and smaller TFTs because increasing the frame rate in LCD televisions is the key to improving motion and minimizing blur, as well as to supporting new 3-D media applications. These market objectives have pushed frame rates from 120 Hz to 240 and 480 Hz. Driving LCDs at these higher frame rates becomes more difficult due to the inherent switching speed of the liquid-crystal material. OLED technology is considered a strong display candidate for 3-D TV because of its fast response time, which reduces image cross-talk during 3-D switching. This enables a much more comfortable viewing experience.

Explosive Growth and Competition in the Tablet Market

Growth in the tablet market took off in 2010, with all indications showing signs of accelerating growth in 2012 and beyond. The pixel density of current-generation tablets is about 130–150 pixel/in., a range in which amorphous-silicon transistors are considered a suitable array backplane technology because of low cost, maturity of the process, and large production capacity. However, because low power consumption is a critical requirement for tablets, and competition is already fierce, new display technologies are likely to be important factors in next-generation tablets. The trend in tablet displays is projected to move toward higher resolution, small border size, and low power consumption, so high-performance, high-mobility transistors will be key enabling technologies.

Key Technology Shift #1: OLEDs Have Moved into Mass Production

A key technology development of last year was the breakthrough of OLED into mass production, which gave display makers a viable production alternative that offered superior color gamut, high contrast ratio, a wide viewing angle, and fast response times. OLEDs were an important factor in the success of the Samsung Galaxy line.

OLED displays offer a number of advantages compared to LCDs, both for manufacturers and consumers. One major advantage is the ability of OLEDs to provide richer, vibrant colors without the need for a backlight, liquid-crystal material, or polarizer films, thereby simplifying the design and reducing manufacturing costs (see Fig. 1). OLEDs' speed advantage is particularly helpful in implementing 3-D functionality.



Fig. 1: The TFT-LCD stack at left requires more elements (for example, a backlight) than the OLED stack at right. Source: Cymer, Inc.


OLEDs can provide nearly perfect image quality, but have a few drawbacks. Compared to LCDs, OLED displays have to date had lower resolution and a larger border size. Furthermore, because OLEDs are current-driven devices, the display brightness is very sensitive to the driving transistor characteristics. Non-uniformity in the display, which is known in the industry by the Japanese word "mura," has been a challenge in moving OLEDs to volume production. Mura has been linked to many process steps, including laser crystallization, non-uniformity of CVD film, doping concentration, activation annealing, photolithography, and etching. Great progress has been made in all of these areas, and there are now several display makers that offer OLED products.

Key Technology Shift #2: LTPS Has Become the Industry Standard

As discussed, amorphous-silicon has a very limited electron mobility (< 0.5 cm2/V-sec), which makes it unsuitable for high-mobility applications. In addition, amorphous silicon tends to degrade under the high-current loads required by OLEDs, leading to lifetime issues and image sticking. It has therefore mostly been abandoned for high-volume OLED production.

LTPS is now the most mature and highest performing candidate for advanced display manufacturing and is the standard approach for mass production of both high-resolution LCD and OLED displays. LTPS is created by irradiating amorphous silicon with a laser pulse, melting it in a very short time period (<100 nsec). After the pulse, the silicon film forms many small crystals as it solidifies.

Alternative Technology Approaches: Metal-Oxide TFT or Solid-Phase Crystallization

Although LTPS has become the industry standard, there are several other transistor technologies that have been developed as alternatives to laser crystallization. The two most promising are metal-oxide transistors and solid-phase crystallization.

Metal-oxide TFT development has been an active research area over the last several years. The most mature material for metal-oxide transistor active layers is indium-gallium-zinc-oxide (IGZO). It has a mobility of 5–15 cm2/V-sec, significantly higher than amorphous silicon. A 70-in. 240-Hz 3-D TV using metal-oxide TFTs was exhibited by Samsung at SID's Display Week in 2011. Metal-oxide TFTs are also a candidate for medium-sized displays, such as tablets and laptop monitors. The metal-oxide TFTs have a smaller transistor area than amorphous-silicon TFTs, which should increase the aperture ratio and reduce the power consumption. Overall, metal-oxide TFTs offer performance and capital cost in between amorphous and polysilicon.

One challenge facing metal-oxide TFTs as they enter mass production for advanced LCDs is that the TFTs are sensitive to light. This can lead to transistor threshold-voltage shifts when the active layer is under light exposure. A protective layer is needed to cover the metal-oxide transistor to prevent this problem, requiring an additional photomask process. Several display makers are currently planning the pilot production of LCD devices using metal-oxide TFTs, so the practical limitations of this technology should soon become clear.

Metal-oxide TFTs may be faced with larger challenges in extending to OLED displays. OLED TFTs require higher current levels, which can degrade the amorphous metal-oxide material over time. This can lead to a shift in the threshold-voltage levels, and also to image sticking if there is differential aging of pixels due to static images. The proposed solution is to use compensation circuits, similar to LTPS TFTs but, as stated previously, the transistor size must be larger due to the lower electron mobility. This will reduce the aperture ratio and brightness for the most common bottom-emission OLEDs. Metal-oxide TFTs are also n-type transistors, in contrast to the p-type transistors of polysilicon. When used to drive OLED pixels, n-type transistors can cause faster degradation of OLED brightness and image sticking. One solution is to implement a reversed-stack OLED structure, but this can lead to low emission efficiency (< 60% of normal).

Solid-phase crystallization (SPC) is an older approach that uses rapid heating of the glass substrate in a furnace to convert the amorphous-silicon coating to polysilicon without melting. The furnaces typically produce high temperatures (>600°C) for several minutes, which can lead to glass deformation. The SPC process has been relatively slow. One approach to speed the process is to add trace amounts of metal (such as nickel) to speed the process. Despite the potential cost advantages compared to that of laser crystallization, SPC is rarely used in mass production due to difficulties with TFT leakage currents and threshold-voltage shift. The weakness of SPC is its lower mobility compared to LTPS. It requires large compensation circuits to overcome the hysteresis effects, which, in turn, lead to significant limitations for high-resolution displays. SPC is also considered difficult to extend to Gen 6 and Gen 8 glass sizes due to glass softening and sag at high temperatures.

Overcoming Traditional Challenges in Laser Crystallization

With metal-oxide TFTs still in development, and SPC not widely adopted, LTPS continues to be the dominant method to produce high-performance transistors. However, there have been significant challenges in scaling LTPS to volume production.

Successful laser crystallization requires precise control of the process to assure uniform crystallization, high throughput, and low operational costs. The most widely used process, excimer-laser annealing (ELA), uses a high-power, pulsed excimer laser to melt a thin line of silicon. A stage moves the silicon-coated glass substrate under the beam, while the laser is operated at a high pulse rate, processing the glass substrate with a series of pulses. To ensure uniformity of the crystallization process, precise and consistent beam control is required. Such control is vital to control stage timing and positioning, as well as laser power, beam uniformity, and focus. While it is widely used in mass production today, ELA has historically delivered relatively low throughput. The beam length was limited, requiring multiple passes to process a single glass sheet. ELA also has had a relatively small process window, which can lead to polysilicon non-uniformities if conditions wander away from the process center. As discussed in the following sections, today's fourth-generation systems show significant progress on both throughput and process robustness, driven by improvements in laser and optical technology.

Higher Laser Power and Beam Stretching Provide Greater Throughput

The excimer laser is the heart of the ELA system, so system productivity is determined by the amount of power the laser can provide. In 2008, the conventional power of a commercial ELA system was limited to 300 W; in comparison, the newest systems now operate with up to 900 W of laser power. Cymer/ TCZ achieved this 3x gain in laser power by introducing a MOPA laser, first developed for semiconductor photolithography. (The development of the ELA system with MOPA technology was completed and the first system launched in 2009. The MOPA architecture provides a more stable process window and reduces the running cost at the same time.) A MOPA excimer laser consists of a master oscillator (MO) chamber and a power amplifier (PA) chamber. The MO is responsible for establishing optimal optical parameters, while the power amplifier (PA) maximizes pulse energy (and therefore output power). The net result is an optical architecture that generates much higher power levels than a single chamber system, with excellent optical performance parameters and stability.

Beam length is the second critical factor for increasing the throughput of ELA systems. All ELA systems use a highly asymmetrical beam: narrow in one dimension (5–400 μm, depending on the equipment supplier) and long in the other. The longer the beam length, the fewer passes are needed to process large substrate sizes. In 2008, the state of the art was 465 mm. Today, ELA systems are available with beam lengths up to 750 mm, which allows a Gen 5.5 substrate (1300 x 1500 mm) to be processed in only two passes. It should be noted that beam length and power go hand in hand. Since the process energy density remains the same, a longer beam requires higher laser pulse power. Additionally, the repetition rate of the laser was increased to further augment productivity.

Plans are already in the works for ELA systems supporting up to Gen 8. The challenge for making a practical Gen 8 crystallization system is to increase the throughput. There are several concepts in development, including combining multiple lasers and optics to expose the substrate in two passes (see Fig. 2). The stage size increases for Gen 8, and there are challenges in handling such large substrates, but these have all been solved in current Gen 8 amorphous-silicon fabs. There is no inherent limitation in scaling the ELA technology to Gen 8, and Gen 8 systems should be available in the near future for OLED-TV products.



Fig. 2: Top-view illustrations of a transistor located on polysilicon that show (a) small step size (1.5 μm) and (b) large step size (~10 μm). Source: Cymer, Inc.


The Challenge of Providing Uniform Polysilicon for OLEDs

The biggest challenge for laser-crystallization systems today is to improve the uniformity of the polysilicon. This is made even more necessary by the unique requirements of OLEDs: in an LCD, the pixel is voltage-controlled, while OLED pixels are current-driven. TFTs that are required to deliver continuous current are much more sensitive to the underlying electrical properties of the polysilicon. This places tighter requirements on the uniformity of the polysilicon in order to ensure a uniform display. Lack of uniformity in the underlying polysilicon translates to TFT variation, which is then seen as visible mura. One solution for overcoming mura is to include a compensation circuit for each pixel that corrects for pixel-to-pixel TFT variation. Compensation increases the TFT threshold-voltage margin and improves production yield. A key drawback of compensation is the difficulty of making high-resolution displays (greater than 300 pixel/in.) due to the physical space in the pixel needed for the circuit. The most promising approach for increasing resolution and reducing border size is to minimize the non-uniformity of the array backplane during laser crystallization, so as to use fewer transistors in compensation.

Laser-crystallization processes are carefully designed to avoid scan mura, which can be created by shot-to-shot variations of the laser. Scan mura can be seen as small variations in the polysilicon crystal structure and surface roughness. As previously discussed, display makers have created complex compensation circuits in each pixel to correct for scan mura, but the fundamental goal is to improve the laser-beam uniformity to the point that such compensation can be greatly simplified or entirely removed. In pursuit of that goal, Cymer/TCZ has developed thin-beam ELA (TB-ELA) technology that uses a small scan pitch (1–4 μm) to produce better transistor uniformity than conventional ELA (which typically uses a 10–20-μm scan pitch). Figure 2(a) shows a top view of a transistor located on an LTPS formed by using a step size of 1.5 μm. The transistor channel covers an area crystallized by a series of laser pulses, so that the TFT electrical properties are determined by the average of the pulses. Figure 2(b) shows the same view for the case of a larger step size (10–20 μm), where the TFT channel occupies the area crystallized by only one or two laser pulses. The larger step size results in less averaging, leading to increased TFT variation due to pulse-to-pulse variations.

Future Improvements in Laser Crystallization for the OLED-TV Market

OLED TV is considered by many to offer the best performance capabilities for next-generation TV systems. It is also expected that OLED process yields will continue to improve and that OLED material cost will continue to fall. In addition, OLEDs have fewer components (such as backlights and optical films). OLED TV should quickly become cost competitive with LCD TVs.

However, as TV glass sizes continue to increase, scanning the entire substrate with a laser could become quite expensive. Several ELA equipment suppliers have developed a stripe mode to reduce the costs of processing larger substrate. Figure 3 illustrates the concept of stripe mode, in which the laser selectively scans the glass, processing only the locations that will be occupied by transistors. As an example, for a 55-in. OLED-TV product with a 150-μm irradiation area and a 630-μm pixel length, the running cost with stripe mode could be reduced by up to 75%.



Fig. 3: A laser scan using the stripe mode selectively anneals the silicon only in the transistor area. Source: Cymer, Inc.


Display manufacturers are facing a dynamic marketplace that is full of opportunity, but also full of challenges in terms of improving performance and function while increasing production volumes and reducing costs. A future migration to OLEDs will improve display brightness, speed, and pixel density, but also place new demands on manufacturers.

Recent market trends demand a fast ramp-up of TFT technologies that can replace the low electron mobility of amorphous silicon with materials of higher electrical efficiency. Several options are available or under development, including metal-oxide TFTs, SPC transistors, and LTPS.

To cope with this changing landscape, display manufacturers need production platforms that can handle the full spectrum of advanced LCD and OLED technologies. The process must deliver high throughput and be scalable to handle larger display sizes. In addition, display makers need platform stability, process consistency, and high reliability.

Today, laser-crystallization systems are providing very-high-quality polysilicon, and this process has become the current manufacturing standard for advanced LCD and OLED displays. Recent improvements in laser-crystallization technology have improved the yield and reliability of the LTPS process. •


David Knowles is the Vice-President and General Manager of the TCZ Display Products Group at Cymer, Inc. Ulrich Hausmann is the Director of Marketing and Applications and can be reached at