Silicon foundry fabricates microdisplay with liquid crystal on silicon
05/01/2001
Wei Shaiu Chen, UMC, Hsinchu City, Taiwan
Robert Melcher, Three-Five Systems, Tempe, Arizona
overview
Modifications to a standard silicon foundry process flow are being used to manufacture liquid crystal on silicon microdisplays. Key improvements are in the areas of planarization and reflectivity. CMP and thin-film processes have been developed for the technology.
 Three-Five's microdisplays are available in a variety of resolutions from SVGA to WUXGA. |
Liquid crystal on silicon (LCoS) technology enables higher performance in a variety of leading-edge displays. The high brightness and resolution provide advantages for projection TV and multimedia projectors. The smaller pixel size for microdisplays enables their use in handheld devices such as video cameras and web-enabled cell phones. (See "Today's applications for LCoS microdisplay technology" on p. 84 for more information on the current use of microdisplays.)
The marketplace has been gaining interest in LCoS technology because of its flexibility, scalability, and performance. The collaboration between foundry UMC and microdisplay provider Three-Five Systems on their joint manufacturing process serves as an example of how the production of both the silicon backplane and the LCoS imager is critical to maximizing performance in microdisplays. This article will examine the special manufacturing processes developed to produce an LCoS imager and the silicon backplane that is the imager's foundation.
Structure and function of the imager
LCoS microdisplays consist of a silicon backplane, a cover glass, and a layer of liquid crystal in between. The silicon backplane contains circuitry, called an "active matrix array," that includes driver circuitry for the liquid crystal. The silicon backplane is similar in design to a memory chip. Fabricated on the surface of the silicon chip is an array of pixels with mirrored surfaces, at a pitch of around 12µm. A typical LCoS silicon backplane is designed with about 1.3 million individual mirrors. Each one is separately driven by a transistor integrated in the silicon chip that can apply a voltage to the mirror, which also acts as an electrode.
The cover glass has a transparent conductor, typically indium tin oxide (ITO), on the inside of the glass. Between the sandwich of the silicon backplane and the glass is the layer of liquid crystal.
The voltage from the active matrix in the silicon backplane is applied to individual mirrors, producing an electric field across the liquid crystal to the common electrode located on the glass. The physical orientation of the liquid crystal molecules changes with this applied voltage, affecting the polarization of the light passing through them. When the molecules are vertical with respect to the array plane, incident light polarization is not affected and light passes through the liquid crystal unaffected. When the liquid crystal molecules are horizontal with the plane, incident light polarization is altered. At intermediate rotations, partial light passes through. The amount of light transmitted is dependent on the degree of rotated polarization. In this way, a microdisplay acts as an array of independently alterable polarizers, adjusting the angle of rotation of incident light, and generating a gray-scale image.
In projector operation, light from an arc lamp is focused onto the device. The light from the arc lamp is polarized, passes through the liquid crystal, and reflects off the mirrors that are fabricated on the silicon chip. The light is reflected out through the liquid crystal, changing its polarization. From there, the light goes into a projection lens that magnifies the image onto the screen. Near-to-eye applications typically use colored LEDs rather than an arc lamp.
 Figure 1. Liquid crystal on silicon process flow. |
The generation of color is different for the two types of microdisplays (projection displays and near-to-eye (NTE) devices). The projection display uses a "three-valve" approach to color. Three imagers are used to modulate red, green, and blue light. Optics then combine the three colors into a single image that the eye perceives as a full-color image.
The NTE displays use a color-sequential approach. The color-sequential image is formed by illuminating the display with a rapid sequence of red, green, and blue flashes. Each of these separate color images lasts for only a few milliseconds, and the eye integrates all three color images into one full color image.
Manufacturing the LCoS imager
The manufacturing of Three-Five's microdisplay imager is a collaboration with the silicon foundry UMC. The chip design from Three-Five is sent to UMC for fabrication and application of the mirrors. Once returned to Three-Five, the backplane undergoes liquid crystal processing, device testing, and packaging. The microdisplays are manufactured in wafer batch mode. The ITO-coated glass counterplate is processed on an automated passive matrix LCD line (Fig. 1).
First, the wafers are cleaned and a polyimide coating that helps to align the liquid crystal on the imager is applied. Polyimide is applied, baked, and then rubbed in a specific direction to align the orientation of the liquid crystal molecules. The cover glass is also cleaned, treated with ITO, and coated with polyimide that is also thermally cured and then rubbed in a specific direction. Epoxy perimeter seals are placed on the glass to define the perimeter of a single imager, and spacers define the gap between the glass plate and the silicon plate to be assembled later.
The 8-in. silicon wafer and 8-in. glass plate are then assembled and separated as individual imager devices. In a vacuum, each imager is filled with liquid crystal in the space between the glass and backplane. The imager cell is closed and placed into a package consisting of a frame, mechanical support, aperture, and electrical connection. The imager and package then go through a series of quality tests before shipping.
For projection applications, the imagers are shipped to customers who develop and build projection engines consisting of an optical system combined with the imagers. Proprietary ASICs are provided to drive the imagers with the appropriate data. The projection engine involves illuminating the devices; polarizing the light; separating the light into red, green, and blue components; conducting polarization analysis for the reflective light; recombining the red, green, and blue light; and projecting it onto the screen. The projection engine is then built into the end-user product such as a television or multimedia projector.
NTE applications use the imager in a different way. The device is illuminated using a triad of light-emitting diodes for red, green, and blue. The image is then projected and imaged onto the user's retina directly from the microdisplay. The user perceives a virtual image that approximates a 14-in. screen about 3 ft away. Reflectivity is less important in this application compared to large screen projectors because the amount of light required to go directly into the user's eye is very small. The objective is to leverage LCoS to design a cost-effective, small device, and the LCoS technology enables almost arbitrarily small devices without major changes to the fundamental architecture of the imager.
The projector and NTE devices require two different imager designs. For projector applications, three distinct imagers are designed to handle red, green, or blue, and they are used simultaneously in the projector. NTE display systems only require one imager because they generate a color sequential image. The main differences between these two types of imagers are the speed of the drive electronics and the liquid crystal response, which is governed primarily by the thickness of the liquid crystal layer. The NTE's color sequential approach requires a very thin layer of liquid crystal, about 1µm, so that the light can be processed more rapidly to maintain the sequential image. The projection system, however, uses a thicker layer of liquid crystal, about 3.5µm, which is a compromise of liquid crystal material availability, response time requirements, manufacturability, and other factors.
LCoS silicon backplane process enhancements
The silicon foundry has made enhancements to its conventional mixed-mode CMOS process for fabrication of the LCoS silicon backplane. The process is capable of supporting extremely small pixel sizes to enable the high resolutions found in Extended Graphics Array (XGA) microdisplays (1024 x 768) for lightweight and portable front projection applications; Super XGA microdisplays (1280 x 1024) for monitors, high-definition television, and high-resolution projectors; and wide ultra-XGA (WUXGA) microdisplays (1920 x 1200) for very advanced applications.
Recent LCoS fabrication process enhancements have enabled increases in reflectivity, contrast, resolution, and overall image uniformity. These enhancements are based on improving die flatness and pixel reflectance to improve light valve efficiency.
 Figure 2. Across-die flatness of ILD planarization methods measured at the final wafer. |
The LCoS fabrication process is compatible with UMC's standard CMOS process, with the addition of a multilayer reflectivity-enhancing stack. UMC has successfully integrated the LCoS backplane process into its 0.35µm CMOS logic technology. The modifications made to fabrication that produce a high-quality light include: 1) a blocking layer to prevent the illumination from disturbing the silicon circuitry; 2) use of CMP to achieve the extremely high surface planarity needed to achieve high contrast; 3) mirror layer reflectance enhancement; 4) gap fill; and 5) passivation.
IMD planarity improvements
The planarization of the silicon is critical to the LCoS light valve's function. The mirrors are typically 12µm, and the surface must be optically smooth in order to achieve the high reflectivity and high contrast needed for projection devices. Any irregularity in the flatness of the silicon surface could cause the reflected image to be distorted. An uneven die surface could cause the thickness of the liquid crystal layer between the silicon backplane and the cover glass to become warped as well, further decreasing reflectivity and resolution. In addition, die flatness is a key factor for the microdisplay assembly process because an uneven die will disturb cell gap uniformity.
For the conventional CMOS process, die flatness is an expected challenge. Although the CMP process helps improve planarity, it is hard to achieve global planarization. Additional steps have been established in the LCoS silicon backplane fabrication process to leverage optimal CMP conditions and achieve the most effective die flatness.
 Figure 3. Blanket mirror reflectance value. The reflectivity is relative to the Al film. |
ILD planarization methodology. In the conventional 0.5µm/0.35µm process, UMC uses borophosphosilicate glass (BPSG) reflow for interlevel dielectric (ILD) planarization. For the LCoS process, however, if BPSG reflow is used at the ILD layer, it is hard to achieve the best potential global flatness even when applying the CMP process at the intermetal dielectric (IMD) layer. This difficulty is due to the tendency of CMP to follow the topography of the previous step. Therefore, the planarity of the ILD layer had to be improved first. It was found that BPSG reflow and spin-on-glass (SOG) etch-back only improve die flatness by 33%. Using CMP, die flatness can be improved by 50%. When the two methods are combined, the die flatness can be improved 75% over conditions in which no planarity treatment was used (Fig. 2).
 Figure 4. Reflectance spectra for different passivation structures (Al surface, thin oxide only, and oxide-nitride-oxide-nitride). |
IMD CMP condition. To achieve superior planarity for the IMD layer, different CMP pressures and pad conditions were evaluated to determine the best CMP process. It was found that a hard pad can improve die flatness by about 40%, and that lower CMP pressure provides an additional 35% improvement. When both conditions are integrated with the ILD planarization approach, the result is a final die flatness <825Å. The typical process not optimized for LCoS manufacturing results in a flatness of 2500Å.
Mirror layer reflectance enhancement
As the display resolution increases to XGA or SXGA levels, the pixel pitch becomes a serious problem. As discussed earlier, when display panel sizes are enlarged, the resolution and brightness decrease. A mirror layer reflectance enhancement process was developed to improve resolution and brightness for LCoS projection applications.
 Figure 5. Schematic diagram of process steps for post spacer setup, including a) gap fill, b) spacer film deposition, and c) spacer etch. |
In the light valve fabrication process, the key to enhancing mirror reflectance is to reduce the surface roughness. The roughness was reduced by using low-temperature aluminum (cold Al) deposition and skipping the antireflective coating (ARC) layer. The mirror layer is fabricated using 2750Å Al/TiN/Ti to maximize the reflectance. The reflectance for the best Al deposition condition can reach a 92% improvement at wavelengths of 400-700nm (Fig. 3).
 Figure 6. The top view of a completed post spacer location a) before and b) after the post fabrication. |
In addition, thinner metal achieves slightly higher reflectance due to the smaller grain size. UMC has found that if metal film thickness is reduced from 4000Å to 2000Å, the reflectance will increase 1%. The pixel space also influences the pixel fill factor and subsequently the reflectance. For the same pitch, minimum pixel space will have a maximum fill factor, and reflectance is increased as the fill factor becomes higher. Using this advantage, the process reduces the pixel gap from 0.9µm to 0.4µm to improve global reflectance by 80%.
Pixel gap fill and passivation
In order to maintain high reflectance, additional methods are used to attain global planarization. After the mirror layer is formed, the metal gap is filled and capped with a passivation layer. High-density plasma (HDP), an ionized gas, is used for the gap fill, and partial CMP and etch-back are used for the planarization to maximize pixel flatness. The partial CMP step will leave a residue of oxide to protect the metal surface. Tests have determined that different dry etch times produce varied gap depths, but even 50% over-etch increases the gap depth only to 300Å, because the etching rate in the gap is lower than in the blanket area.
After the gap fill, a passivation layer is deposited to protect the mirror surface. The passivation layer consists of four separate structures: oxide only, oxide-nitride (ON), oxide-nitride-oxide (ONO), and oxide-nitride-oxide-nitride (ONON). Thin-film optics are used to optimize the film thickness for a higher reflectance. Three kinds of structures (Al, oxide only, and ONON) were used to verify the simulation results, and the reflectance spectra matched the simulations closely (Fig. 4).
Multilayer spacer
A new technology that has not yet been implemented in production is the multilayer spacer. This addresses the problem of warped glass, which can distort the image and cause the microdisplay to lose resolution. To keep the glass from bending, a brace above the top metal supports the glass that will be attached later. The support is a multilayer ONO spacer on the mirror layer measuring 4µm by 4µm with a height of 1.5µm (Figs. 5 and 6). This spacer also provides a significant improvement to the yield by limiting defects in the liquid crystal and the attached glass.
Conclusion
LCoS technology will have a significant impact on the future of microdisplays, enabling users to enjoy higher-resolution projection devices and more compact, convenient, and portable displays. The LCoS process is the only microdisplay manufacturing process that can be used effectively in such a wide variety of display applications.
 Microdisplays are manufactured in wafer format until the liquid crystal fill process. |
"During the past decade, the dramatic growth of the multimedia projector market has driven the sales of microdisplays to more than $1 billion annually," according to Charles W. McLaughlin of McLaughlin Consulting Group, a market research firm based in Menlo Park, CA. "During the next five years, demand for microdisplay-based rear-projection, large-format, digital televisions and desktop monitors will become the high-growth segment of the microdisplay projection market. Sales of personal display microdisplays used in viewfinders and headsets and embedded in internet appliances will grow to $1.5 billion by 2005, becoming a substantial segment of the $5.6 billion total microdisplay market."
McLaughlin's projections show that the microdisplay market is just beginning to emerge. At this stage, advancement of technology is critical. Better planarization and higher reflectance are required to make LCoS technology a mainstream display choice, and the resources of a silicon foundry are enabling this progress.
Wei Shaiu Chen received his masters in electrical engineering from Taiwan Industrial Technical University in 1996, and joined UMC as a process integration engineer shortly after. He is currently a principal engineer at UMC, focusing on development of the 0.18µm logic process.
Robert Melcher received his BS in physics and mathematics, and his MS and PhD in solid state physics. Since 1999, he has served as the CTO of Three-Five Systems. Prior to that, he led an international team from IBM Research that helped to invent, develop, and market large area projection displays based on reflective LCoS microdisplays.
For more information, contact Deanna Krause, Three-Five Systems, 1600 North Desert Drive, Tempe, AZ 85281; ph 602/389-8800, e-mail [email protected].
Today's applications for LCoS microdisplay technology
A microdisplay is a thumbnail-sized device that displays a high-resolution image, an array of pixels typically 12µm x 12µm. The key advantages of microdisplays over other display technologies are that they are low-cost, compact, and lightweight, while rich in information content.
LCoS technology can scale a microdisplay device to almost any size. The display itself typically measures <1 in. diagonally, but it can offer resolutions from 1/4VGA (78,000 pixels) to WUXGA (over 2 million pixels). A higher number of pixels results in a higher-resolution image. As the number of pixels in an image is increased, the size of each pixel becomes smaller and the image clearer. Other display technologies do not offer such a wide range of resolutions.
In addition to greater pixel density, LCoS has the added advantage of increasing the brightness of an image, which also increases resolution. With large screen TVs and many types of multimedia projectors, resolution and brightness are linked. As the screen becomes brighter, the image loses resolution and vice versa. Consequently, today's projection TVs must be watched in a dark room. With an LCoS television, however, brightness and resolution can be separated and therefore increased in the same device without negatively impacting each other. An LCoS projection TV can provide a bright image even in standard room light.
Due to the advantages of this flexible technology, LCoS can be used in a variety of devices such as viewfinders, wearable displays, computer monitors, large screen televisions, and multimedia projectors.
Near-to-eye microdisplay applications. NTE display systems can use microdisplay technology for portable or "wearable" PCs, as well as web-enabled cell phones. The NTE applications use a single LCoS display with a magnifier that is brought near to the eye, allowing the user to view a magnified virtual image. Because portability is the key feature of these products, the low-cost, compact LCoS display is a suitable approach.
A good example of NTE technology that is facilitated by LCoS is a web-enabled wireless device. Today, a user cannot view a web page on a traditional cellular phone. The average cell phone has about 7000 pixels on the screen. In contrast, the smallest microdisplay has 70,000 pixels, and a typical LCoS display has 500,000 pixels. With this technology, a user could view a high-resolution web page or video content through a handheld device without having to reformat the data.
Projection microdisplay applications. Projection display systems, such as computer monitors, large screen television systems, and multimedia projectors, benefit from LCoS microdisplay technology with higher color depth, improved contrast, and higher resolution at an affordable system cost. LCoS technology's capability of maintaining increased brightness and resolution is a key feature for supporting these applications.