by Murali Jayapala, Ph.D.
Senior Researcher, IMEC
The possibility to fabricate micromirror arrays has sparked a revolution in display systems. Today, devices using micromirror arrays are already indispensable in many products, such as large-scale projection engines or portable projectors. But the promise of micromirror technology is much broader -- it is the enabling force behind a range of research paths in spectroscopy, lithography, volumetric displays and optical networking, just to name a few.
In this article we highlight some of the recent advances in micromirror arrays, made possible by using silicon-germanium (SiGe) MEMS technology. We also present two sample applications enabled by this technology and currently under development: micromirror-based zoom lenses and high-definition holographic displays.
Monolithically integrated MEMS allow massive micromirror arrays
MEMS, such as micromirror arrays, are driven by signal conditioning circuits. Today, about half of the current MEMS market uses a hybrid approach, developing MEMS and controlling ICs separately. Since this approach is modular, it has a shorter development time compared to the monolithic approach. It also allows for independent optimization of IC and MEMS technologies. However, the advantages of monolithic integration outweigh the disadvantages for those systems where performance and miniaturization is of key importance for the application, or when many (think of millions) interconnections are needed between MEMS and CMOS.
With SiGe MEMS technology, it is possible to integrate MEMS, e.g. micromirrors, on top of CMOS using SiGe as structural material, in a MEMS-last process. SiGe is as reliable as silicon, but it can be processed at much lower temperatures, i.e. below 450°C. This allows one to process layers on top of a finished CMOS chip without damaging it. MEMS made from SiGe are extremely reliable and they can be driven very precisely. They show little creep, especially compared to materials such as aluminum. Moreover, and contrary to other monolithic approaches such as MEMS-first or MEMS-interleaved, this approach allows for a high degree of modularity, and it can be implemented in low-cost and state-of-the-art CMOS foundries.
In our R&D fab, we added SiGe processes to a CMOS flow. We used a multilayer process that combines chemical vapor deposition (CVD) and plasma-enhanced chemical vapor deposition (PECVD) to produce high-quality films at relatively low temperature below 450°C. We reached a low resistivity of 1.45 mΩcm, a tensile residual stress of 35MPa and a very low strain gradient of 3.6x10-6 μm-1 for 10 μm thick layers. The contact between Al and SiGe was found to be ohmic, as required for CMOS integration. In our platform, we used state-of-the art equipment for the SiGe deposits and for deep reactive ion etching of SiGe, to ensure a high within-wafer and wafer-to-wafer uniformity and a high process yield.
With the poly-SiGe MEMS platform described above, we succeeded in fabricating a reliable CMOS-integrated 10 cm² 11 MPixel SiGe-based micromirror array. This array is to be used as spatial light modulator (e.g. for video projection, mask writers, optical mask-less lithography). Its 11 million 8μmx8μm-pitch mirrors can each be individually addressed by an analog voltage to enable an accurate tilt angle. The pixel density of this array is almost double compared to the state-of-the-art. For this array, we demonstrated a stable average cupping below 7 nm, root mean square (RMS) roughness below 1 nm and long lifetime (>1012 cycles, no creep). The mirror array was processed on top of standard 0.18 μm analog-CMOS wafers, featuring 6 interconnect levels.
Micromirror-based optical zoom lenses
Mobile phones become more powerful with each generation, and the features on the integrated cameras have kept a similar pace: higher megapixels, extra features in camera photo editing, photo styles and even video.
But one feature that is still lacking in mobile phones is a full optical zoom. Today’s digital zoom systems still have a significantly lower image quality than would be achieved with real optical zoom lenses. Digital zooming is based on the principle that an image is taken at a focal length which is fixed for that camera. The resulting image is then cropped and scaled with software to achieve the appropriate magnification. This works for relatively short zoom ranges, but for larger zooms the image quality seriously degrades.
Conventional, optical zooms use variable focal length lenses. These allow for a much better quality, but they require complex, macroscopically protruding optical elements. So they are not compatible with a small, fixed form factor, required by e.g. small mobile devices.
Based on micromirror arrays, we are developing a zoom lens system that is at the same time compact and fixed, and that works with as a lens with variable focal length. These “smart lenses” overcome the form factor limitations by providing optical zooming without macroscopic movements of optical elements. Our zoom system for mobile phones will have a 3.5x zoom and a compact form factor of 7x5x20 mm. T he zooming is achieved without changing the form factor.
The basis of our smart lens is a planar micromirror array, with the mirrors arranged in concentric circles, i.e. a polar grid array. It is possible to set up the array so that it behaves as a variable focal length lens. This is illustrated in the figure below: by controlling the tilt angles of individual mirrors in the array, parallel rays can be focused onto a focal point and a specific focal distance. So by changing the tilt angles of micromirrors in the array in accordance with this relation, the focal lens is changed. For instance, given a polar grid array with a diameter of 7 mm, with mirrors with a maximum tilt angle of 2 degrees, the focal length can be varied from 50 mm to infinity.
Schema of a lens with variable focal length, made with a planar polar grid micromirror array.
In 2010, using an optical simulation tool, we’ve successfully modeled the required micromirror arrays (see figure below). With this model, we’ve simulated micromirror-based lenses with variable focal-length. In addition, we’ve drawn up the specifications for micromirror arrays optimized for mobile phones and professional cameras. Last, we’ve also tested micromirror arrays with various hinge designs, looking for an optimal technology for the smart lenses. In 2011, we’ll fabricate micromirror prototypes and build a system demonstrator for zooming.
High-quality, real-time holographic displays
The holy grail of visualization technology is undoubtedly holography, promising a 3D experience that is much more natural than what is possible with today’s limited 3D stereoscopic projection. But the realization of such a high-definition computer-generated holography (CGH) system remains an open challenge.
The quality of the visualization is to a major extent determined by the physical properties of the display device: high-definition holographic display requires hundreds of millions of individually controlled sub-micron diffractive pixels. Moreover, CGH is extremely computational-intensive, which complicates making real-time applications.
Holography, as a physical phenomenon, is based on light diffraction and interference. For it to work properly, with usable viewing angles, the light-diffracting elements of the holographic display must be sized close to or below the wavelength of the light, i.e. they have to be sub-micron. And to achieve high-quality visualization, we need a massive amount of these light-diffractive elements. This is because there is no one-to-one match between the voxels (3D pixels) representing the 3D scene and the diffractive pixels on the display; the diffractive pixels are programmed to render a so-called holographic fringe that recreates the encoded 3D scene as a whole. This means that each diffractive pixel contributes to a multitude of 3D voxels and the visual information encoding the 3D scene is distributed (more or less evenly) over the whole array of the diffractive pixels, i.e. the holographic display.
Today, CGH visualization systems achieve only a relatively low visualization quality and a narrow viewing angle. The display units are typically based on LCD, LCOS, or micromirror technologies with scaling limits at around 2-4 µm, which limits the projection angle to less than 10 degrees. Furthermore, the displays with the highest resolution have at most up to 8 million pixels, which is largely insufficient for visualization of high-definition 3D scenes.
IMEC’s diffractive pixels (top left) forming a holographic test chip (bottom left) that encodes a 3D scene: Focus on “A” cube (top right) and “B” cube (bottom right) of the same scene, i.e. human eyes may naturally focus on the whole scene.
As a solution, we are developing a High-Definition Holographic Display (HoloDis) based on SiGe NEMS (nano-electromechanical system). In a first step, we have manufactured test structures ranging from a pixel size of 250 nm to 2,400 nm. We use these to experimentally verify the theoretical assumptions for computer-generated holography (see figure above). With the test structures, we also simulate the diffractive optical behavior and efficiency of such diffractive nanodevices. In addition, we’ve started designing a prototype SiGe NEMS holographic display and we’ve conducted an architecture study for the monolithic integration of the diffractive devices and their driver IC.
Our next step will be to manufacture a first prototype HoloDis system with 1 million sub-micron diffractive devices on a chip. This will be the first prototype demonstrating holographic video display with a diffraction angle of approximately 30 degree and a projection angle of 15 degree. In the longer term, we aim at realizing a HoloDis system with more than 900 million diffractive elements with a pixel pitch of less than 500nm (see figure below). Such HoloDis system is the central building block for true high-quality 3D visualization systems with projection angles above 30 degrees.
Visualization R&D with SiGe MEMS technology
With our SiGe MEMS technology, we have already fabricated large arrays with micromirrors of down to 8 µm in width. We are now further developing the platform to produce even smaller micro- and nano-device arrays (below 1 µm). These will be needed to make breakthrough, high-quality 3D holographic displays. We are also exploring other ways of forming the micromirror surfaces to provide a high reflectivity and fill factor. This is with the aim to produce high quality micromirror arrays for smart lens applications. In addition, we further explore the potential of our MEMS technology for other vision applications, for example diffractive projection systems or hyperspectral imaging systems.
Dr. Murali Jayapala is a Senior Researcher at IMEC, Belgium. In 2005, he received his Ph.D. in Applied Sciences (Computer Engineering) from the Katholieke Universiteit Leuven (KULeuven), Belgium. In 1999, he obtained his M.E. in Systems Science and Automation from Indian Institute of Science (IISc), India. Dr. Murali Jayapala has co-authored several articles in international conferences and journals in the field of low power embedded systems design. He has served in program committees of international conferences and workshops and he is also a member of IEEE.
Copyright 2011 MEMS Investor Journal, Inc.