Detecting failure modes in today’s MEMS
by Noushin Dowlatshahi and Bob Chanapan, Veeco Instruments Inc.
Microelectricalmechanical systems (MEMS) are taking a whole new role in our day-to-day life, and are much more widely used than ever before, due to a wide range of benefits including their low mass, fast mechanical response, low power consumption, and potential for lowering end costs. Meanwhile, test and quality control of such devices has become more critical to facilitate the insertion of this technology into critical applications. Therefore, it is not surprising that precision metrology has a huge role in this steady advancement of MEMS technology.
Unlike traditional semiconductor devices, today’s MEMS devices require characterization in both their static state and under actuation. Parameters of interest include shape, dimensions, surface roughness, sidewall angles, film thickness, residual stress, feature volumes, response times, thermal properties, resonance frequencies, stiction, environmental compatibility, and more. The greatest difficulty in MEMS metrology is the nearly endless diversity in form and function of the devices. Each type of device has different complexity in the manufacturing process, a different geometry (from tens of microns to tens of millimeters in width), and vastly different performance requirements or designed responses to external stimuli. Thus, while curvature, surface roughness, and switching time may be critical for micromirror arrays, co-planarity of the capacitor parts and linewidths may be the most crucial parameters for an accelerometer. Initial or long-term failure of a device may also be driven by an array of factors, such as geometric errors in production, contamination, improper removal of sacrificial layers, stiction, environmental attack, fatigue, electrostatic clamping, fusing, delamination, or electrical damage.
Figure 1: Wyko NT dynamic measurement reveals a hidden defect in a micro-mirror.
During research and development, there is also a need to map the behavior of devices and materials placed under external stresses (e.g., temperature, pressure, or corrosive agents) in order to fully understand the time-course evolution of these processes. Early detection of failure modes not only enables improvement at the next fabrication cycle and increased yields, but it also identifies whether the problem is isolated or systematic, directly impacting field failure quality control.
Figure 2: Wyko NT profiler image of a MEMS fabricated microphone.
MEMS applications in the domain of radio-frequency (RF) circuits serve as a perfect example. For certain capacitive MEMS switches, it is crucial to examine the stiction between the metal layer (top electrode) and the dielectric layer covering the bottom electrode. The charge build-up in the dielectric material can result in what would typically be called a “failure mode.” The actuation voltage is directly affecting this charge build-up. Researchers need to understand how an increase or decrease in the actuation voltage can impact the lifetime of typical MEMS. This actuation voltage is typically related to the device geometry, mechanical and material properties, and residual stresses in the devices. When voltage is applied to a typical capacitive RF MEMS switch, the electrostatic charges cause a distributed electrostatic force, leading to deformation of the micro-structure. Later, due to the storage of elastic energy, the structure tends to return to its original state. Accurate metrology can play a significant role in characterizing such a switch by defining and evaluating the membrane’s deformation and analyzing thermal impact, as well as the material and mechanical properties of the metal.
Another widely used type of MEMS device that exhibits particular metrology needs is the micromirror array, which is finding use in adaptive optics for both space and consumer electronics, as well as in projectors, televisions, and digital cameras. These mirror devices must be 100% reliable and accurate. Because of variations in device geometry that result from the micro-fabrication process, the performance from one device to the next can vary, making it difficult to reliably ensure accuracy and repeatability of the actuator positions. These arrays typically consist of simple cantilever beams, torsion beams, tethered (piston-style) beams, circular membranes, and oval membranes. One technique of fabricating the micromirrors is to use low-temperature adhesive wafer bonding to deposit a thin layer of monocrystalline silicon device to a CMOS wafer. The digital micromirror device (DMD) developed by Texas Instruments, which incorporates more than 500,000 individually addressable micromirrors, has made great strides in both performance and reliability of such devices. Digital light processing technology based on DMD has been used in such diverse products as projection displays with film-like projected images and photographic-quality printers. Reliability testing of the DMD has demonstrated greater than 100,000 operating hours and more than 1 trillion mirror cycles.
It is crucial for micromirror array manufacturers to examine the uniformity in how the micromirrors tilt in an array, the time it takes them to tilt, and how repeatable this behavior is. This is in addition to the traditional parameters of mirror curvature/deflection angle roughness and surface topography. Moreover, as the devices are packaged in hermetically sealed packages, the encapsulation process can further affect the operational performance. Environmental factors including temperature and humidity also impact on such devices, creating the need of a fourth-dimensional capable, non-contact metrology solution. Inspecting the device in a meaningful manner can monitor and minimize fabrication errors and positively impact the micromirrors components within an array.
Manufacturers have relied on a range of methods to identify structural failure modes in MEMS: optical microscopy, stylus profilometrey, infrared (IR) thermal imaging, focused ion beams (FIB), scanning electron microscopy (SEM), digital holography (DH), atomic force microscopy (AFM), and transmission electron microscopy (TEM). Each of these techniques has pros and cons, and selecting the proper solution depends on a number of factors, from device type to stage of manufacturing.
SEMs generally offer the highest lateral resolution available for static MEMS imaging. These systems operate by focusing a beam of electrons on a small area and detecting the electrons that are scattered from the surface. Measurement time is typically between 5???10s after the sample has been loaded into the system. Lateral resolutions down to the nanometer level are possible, and depth-of-focus of such systems can be quite large. SEMs provide an invaluable tool for evaluating sidewall angles and roughness of very steep parts, and are often the only systems capable of these measurements in such applications as microfluidics and capacitive accelerometers. The primary limitations of the technique are that samples must be placed in vacuum, and the penetration depth of electrons is very small, limiting effective analysis on packaged parts. Also, parts must often be cross-sectioned in order to achieve the proper orientation for best measurement, which can be cost prohibitive. Lastly, edges may produce artifacts if they become charged, leading to inaccurate data.
Laser Doppler vibrometers are frequently used for rapid characterization of out-of-plane MEMS motion. These systems employ a beam of modulated laser light that is reflected from the moving test piece. The returned light is Doppler-shifted from the original modulated beam, and the two beams are then compared. A spectrum analyzer or other electronics can be used to determine the magnitude of the shift, and thus the velocity of the moving part. Displacement can also be measured through integration of this signal. These systems are very sensitive to motion, with sub-nanometer resolution in the direction perpendicular to the test beam. More sophisticated systems enable the measurement point to be scanned across the surface, to build up a complete picture of the MEMS out-of-plane motion, though phase information between each of the monitoring points is lost due to the time lag between measurements. The lasers on these systems can also be coupled to other optical systems directly or via fiber options, such as stroboscopic bright-field microscopes, so one can get the in-plane-motion through a separate series of measurements with that equipment. However, all systems are sensitive to the roughness of the surface, requiring enough light to be scattered back into the detector for adequate signal-to-noise. Therefore, very smooth or rough surfaces can degrade accuracy.
Digital holography (DH) has been employed for many years to measure vibrating devices. In this technique, a digital camera is used to record a hologram produced by interfering a high-quality reference beam with a beam reflected from the sample under test. As the test object is later deformed, the modified object beam is compared with the original digitally recorded hologram, and the deformation can be quantified using standard phase-shifting or other interferometric techniques. These systems can therefore achieve nanometer-level measurement of out-of-plane motion of devices. Using high-speed cameras, deformations of up to several thousand hertz can be measured. The systems can also be combined with stroboscopic methods to measure motions up to several megahertz, similar to the strobed systems mentioned previously. Like the other optical methods, digital holographic systems can measure parts through a transparent cover glass. However, as the source employed is a laser with long coherence length, they can suffer speckle problems from diffuse MEMS surfaces and may also have problems with transparent films on a surface.
Atomic force microscopes (AFM) employ nanoscale tips attached to the end of a cantilever that is brought very close to the test surface. Measurement of the tip deflection allows full 3-D characterization of the surface, and van der Waals forces between the tip and sample can be detected. AFMs utilize a variety of measurement modes, including “contact mode” (where the tip lightly touches the sample) and “tapping mode” (where high-frequency tip oscillations are used to minimize contact with the sample while maintaining the high signal-to-noise ratio). AFMs have the highest lateral and vertical resolutions available of any three-dimensional metrology instrumentation, with sub-nanometer features visible in all three dimensions. However, the field of view for a scan is typically only about 120µm2, and vertical limits are about 10µm in height. In addition, AFM measurements are relatively slow. Due to the scan speed and contact nature of the scan, these systems are incapable of measuring MEMS while under actuation.
Optical profiling (white light interferometry) has long served as a standard technique for measuring surface topography of MEMS and optical MEMS devices. In a white-light optical microscope, the illumination source is traditionally a tungsten-halogen bulb, coupled into an optical system with several interferometric microscope objectives. Light from the source is split in the objective, with part of the beam traveling to the device under test and the other part to a high-quality reference surface. The light recombines to form a pattern of interference “fringes.” The reference surface is then translated relative to the test surface, and multiple frames of data are obtained during the translation. The resulting series of fringe patterns are analyzed to calculate the surface profiler of the device. Optical profilers provide rapid, 3-D, non-contact sample surface characterization. Consequently, this method has proven quite useful for MEMS production processes. However, this method has traditionally also been inadequate for determining a device’s dynamic ability to perform, such as actuation, deformation, rotation, etc.
CalTech overcame this primary limitation of optical profiling by combining it with stroboscopic illumination for dynamic measurement of MEMS devices in motion. Stroboscopic illumination effectively “freezes” the motion of MEMS structures, allowing nanometer-resolution measurement of their surface shape. By varying the amplitude, phase, and frequency of the drive signal, multiple measurements can be taken to describe the device’s full range of actuation/deformation. Moving MEMS structures can then be analyzed for flatness variation, tilt, lateral translation, linearity of motion, stiction, and other key device parameters.
Today’s top-of-the-line optical profilers incorporate this technology along with other advances to provide a best measurement solution for MEMS researchers and manufacturers. Veeco’s ninth-generation Wyko NT profilers incorporate LED illumination for higher light levels and In-Motion capability for measuring devices under actuation from 0???2.4mHz with angstrom-level resolution. Objective modules that can visualize parts through packaging are also commercially available.
Thus, it is now possible to perform non-destructive, rapid, extremely precise MEMS characterization through packaging and under operation on a single measuring tool.
Noushin Dowlatshahi is marketing product engineer at Veeco Instruments Inc. Bob Chanapan is marketing product manager at Veeco Instruments Inc.