Future Challenges for MEMS Failure Analysis

Transcription

1 Future Challenges for MEMS Failure Analysis Jeremy A. Walraven Sandia National Laboratories. Albuquerque, NM USA Abstract MEMS processes and components are rapidly changing in device design, processing, and, most importantly, application. This paper will discuss the future challenges faced by the MEMS failure analysis as the field of MEMS (fabrication, component design, and applications) grows. Specific areas of concern for the failure analyst will also be discussed. 1. Introduction MEMS research is a relatively young field compared to ICs. MEMS design, fabrication, packaging, and reliability testing are still in their infancy and require constant revision and improvements now and over the next several years. MEMS failure analyisis (in this context) is a younger field than MEMS fabrication and design. Although MEMS have been around for a number of years, with failure analysis support for production, packaging, testing, and field operation, the tools and techniques required to properly diagnose the root cause of failure need to be upgraded and designed specifically for MEMS failure mechanisms. MEMS failure mechanisms can be as unique as the devices themselves. In ICs, considerable efforts are taken in handling and testing to properly characterize and assess device performance and compare the performance to device specifications. One major difference between ICs and MEMS testing is the environmental conditions. In many instances, ICs are tested in various environments ranging from various temperature and humidity conditions to vacuum and inert gas. In MEMS technology, similar handling and testing procedures are implemented, but the device is required to work with a given environment [1]. Varying the test environment can dramatically change device sensitivity and functionality. The added complexity of mechanical motion requires added care in handling and testing. Fortunately, MEMS has the advantage of leveraging IC FA tools and techniques for MEMS analysis. However, as the number of devices and applications grow, the MEMS failure analyst must become more diverse and multi-disciplinary in their knowledge base to properly diagnose the root cause of failure. This has become clearly evident in the failure analysis of thermally versus electrostatically driven actuators, microbiological and microfluidic devices, optical and RF components, and the wide array of sensors available for use. 2. MEMS Technologies and Future Challenges MEMS components can be classified into six distinct categories. These categories of MEMS are based on their application. These categories include: Sensors Actuators RF MEMS Optical MEMS Microfluidic MEMS Bio MEMS Sensors are devices designed to sense changes in and interact with their environments. Devices found in this class include chemical, motion, inertial, thermal, and optical sensors. Actuators are devices designed to provide power or stimulus to other MEMS components devices. In MEMS, actuators can provide power using either an electrostatic or thermal stimulus. RF MEMS are devices used to switch, transmit, filter, and manipulate radio frequency (RF) signals. Typical devices include; metal contact switches, shunt switches, tunable capacitors, antennas, tunable filters, etc. Optical MEMS are devices designed to direct, reflect, filter, and/or amplify light. These components include optical switches and reflectors. Microfluidic MEMS are devices designed to interact with fluid-based systems. Devices such as pumps, valves, and channels have been designed and fabricated to transport, eject, and mix small volumes of fluid. ITC INTERNATIONAL TEST CONFERENCE /03 $17.00 Copyright 2003 IEEE

2 Bio MEMS are devices that, like microfluidic MEMS, are designed to interact with biological samples. These devices are designed to interact with proteins, biological cells, medical reagents, etc. and can be used for drug delivery or other in-situ medical analysis. [2] These six areas represent vastly different applications of MEMS devices currently being used or being developed for both commercial and government applications. Given the breadth of devices created using MEMS technology, the failure analyst must not only be cognizant of the fabrication process, functionality, and application of the MEMS component but also the environment the component will be operated in. Many of the challenges in failure analysis are global and will require infrastructure development, but many more are application dependent and are discussed in this paper. 2.1 Sensors Sensors are a very general class of MEMS where the devices are designed and fabricated to sense a change (or multiple changes) in its environment. These devices have been designed and fabricated to sense changes in fluids (contaminants), inertia, force, gas, etc. Given all these potential sensing functions and various designs, many challenges await the failure analyst. Some global issues for sensor failure analysis include parallel testing and analysis, bench test setups to simulate sensing environments, and depackaging/deprocessing of the device. Many tools and techniques are available to assess failed MEMS. Tools leveraged from the IC industry have helped to resolve various failure mechanisms. One major problem is the ability to diagnose the root cause of failure for multiple devices. Very few tools exist that can test and monitor (both electrically and mechanically) hundreds of MEMS components. The challenge here for failure analysis is assessing tens or possibly hundreds of die, each with a possible unique failure mechanism, but more importantly, very little background information on the failure mechanism. One major hurdle for the failure analyst in creating a bench test setup to simulate a sense environment is simulating the chemicals the device is designed to sense. In some instances, devices designed to sense changes in inertia or force can be easily simulated. However, devices designed to sense chemical constituents present problems for locally reproducing the failure. For chemical sensors in particular, these devices are designed to sense contaminants that may be harmful. This makes reproducing the failure mechanism in the laboratory a difficult task. The failure analyst must also be aware of the packaging scheme used in manufacturing the MEMS component. For devices designed to specifically interact and sense an environment, rapid changes in the environment or faulty depackaging can all lead to erroneous results or inducing other failure mechanisms during failure analysis. In MEMS particularly sensor based devices removing the lid and changing the environment of the device can alter its functionality. Decapping or depackaging a MEMS component can introduce a different failure mechanism in the device prior to performing the key failure analysis activities. Problems such as particle contamination, rapid environmental changes, and the potential for damaging the MEMS component all exist and can impede failure analysis efforts. Leak testing or hermiticity testing of devices with very small cavities can also be a problem. Detecting small leaks in a system is important to the failure analyst to help determine any changes in device sensitivity. A schematic of a MEMS component packaged underneath a cap is shown in Fig. 1. MEMS Cap Integrated Circuit Fig. 1. Schematic of a packaged MEMS component underneath a cap. Note the volume of the package, and the volume of the MEMS underneath the cap. 2.2 Actuators Actuators are components designed to provide power to move other MEMS components. Devices such as RF MEMS, optical MEMS, microfluidic, or BioMEMS require some form of actuation to contact another surface, move a micromirror, or move a fluid. MEMS based actuators are either thermally or electrically stimulated. Thermal actuators use heat produced from power dissipated in the device. This causes expansion of the component, providing the displacement necessary for motion. For electrostatic actuators, electric fields are produced and attract other components to create movement. Dominant failure modes observed on both thermal and electrostatic actuators include particle contamination and stiction. In some cases, the actuators can have impacting or rubbing surfaces resulting in wear or the formation of debris. The main concerns for failure analysis of MEMS based actuators are non destructive analysis, failure analysis of anti-stiction coatings and films, and dynamic analysis of multiple components in parallel. Each actuator type has its own unique set of failure mechanisms and problems for failure analysis. 851

3 a b Fig. 3. Polysilicon comb finger contacting a ground plane (arrow) due to ESD damage. [3] Fig. 2. a) Failed thermal actuator leg. b) close-up of a melted leg due to excessive current/heating (arrow). For thermally driven actuators, the typical failure mechanism is thermal degradation. The effects of electrical/thermal cycling of these devices are currently underway. Overstress events have been shown to cause degradation by inducing permanent deformation in the throw mechanism. This permanent damage has been shown to cause adverse effects in the thermal actuator by producing out-of-plane motion instead of linear motion, reduced linear throw, and welding of the thermal actuator to the ground plane. Although these failure mechanisms are relatively easy to diagnose, the challenge for the failure analyst is to determine what temperature the thermal actuator got to when permanent damage occurred. Thermal analysis of moving components is very difficult. Techniques used to produce thermal images can be destructive or impede motion of the thermal actuator. Deposition of fluids on the surface such as FMI and others work well on ICs but will prevent motion of the device. The importance of understanding how hot a thermal actuator gets, and where the localized heating occurs will aid in modeling the thermal behavior and help mitigate thermally based failure mechanisms. A thermal actuator failure due to excessive heating in shown in Fig. 2. In electrostatic actuators, there are typically a large number of comb fingers present create the electric fields necessary to attract the other half of the actuator drive. In multi-layered MEMS components, defects and damage on comb fingers that cause failure may occur. Rapidly identifying which comb finger generates the failure mechanism is important. The challenge for the failure analyst is to rapidly identify the group or single comb finger that induced failure rapidly and non-destructively. Examples of failures in electrostatic actuators are shown in Figs. 3. Actuators with rubbing surfaces can also fail by wear [4]. 2.3 RF MEMS RF MEMS themselves have a wide variety of applications ranging from tunable filters to contact switches. In most cases, the main issue of interest for the failure analyst is the device metallurgy. RF MEMS offer significant improvements over their macroscale counterparts, but in the areas of contact metallurgy, little is known or understood about contact properties and surface interaction at the micron scale. Critical issues such as contact area, asperity size and geometry and the number of asperities in contact are critical in understanding contact mechanics. By understanding the surfaces and their properties, the failure analyst can better understand the failure mechanism(s) associated with contact surfaces. An example of sidewall contact surfaces in an RF MEMS switch using gold metallurgy is shown in Figs. 4a, b, and c. Another challenge for RF MEMS failure analysis is nondestructive analysis of an adhesion or cold-weld failure. Here, two parts of the component are adhered to one another resulting in a short. This failure mechanism can occur in a variety of ways; however, locating the actual sticking point without removing the top or side components is very difficult. In some instances, the components can be removed to allow the surfaces to be examined. The problem with this approach is the immediate contamination of the surfaces when the top or side components are removed, exposing the contact surface. As shown in Fig. 5, a prototype contact switch detailing the actuation pad and contact surface. If the top portion of the switch were removed to resolve a failure 852

4 a Actuation pad Contact switch b c Fig. 5. RF MEMS prototype switch. Note the actuation and contact sites. Fig. 4. a) functional and destroyed contact sites. b) contact damage, c) close up of damage. Images courtesy of E. J. J. Kruglick ( Berkeley Sensors and Actuators). mechanism, the underlying exposed surfaces would be immediately contaminated, resulting in potential erroneous conclusions [5]. A common feature of interest in all contact devices is the surface roughness of the contact area. Knowing the RMS roughness of the contact surfaces of a device is critical to understanding the contact area. RMS roughness can be performed quite well using an AFM; the process is serial and only works on small sample areas. The challenge for the failure analyst is to develop a tool that can perform multiple measurements on single and several devices in parallel at the same time. This would provide the statistical information needed to understand the RMS roughness of a device and correlate these roughnesses from device to device, wafer to wafer, and lot to lot. Another failure mechanism observed in various RF MEMS is charge accumulation in the surrounding dielectric material. This failure mechanism is very detrimental to shunt switches, where charge is deposited into and through dielectric material. The main problem with residual charge in the dielectric is residual charge increasing the voltage required to actuate the device and pass through a signal. This failure mechanism occurs over time and over a repeated number of cycles. Residual or trapped charge can cause failure through increasing voltage required to actuate the device and pass a signal through. By constant voltage increases, the dielectric will breakdown, resulting in catastrophic device failure. The challenge for the failure analyst is to identify this failure mechanism before dielectric breakdown occurs, and determine how much charge is trapped in the dielectric material. By determining how much charge is trapped in the dielectric material, one can model the time required to failure under constant or cyclical response. 2.4 Optical MEMS These MEMS components are typically used in the telecom industry, and are fabricated in large arrays. These components tend to have large surface areas that are tilted, rotated, or generally moved or more degrees. Optical MEMS offer similar challenges to the failure analyst that RF MEMS do. Analysis of RMS roughness of the mirror surface, non-destructive analysis of adhesion or stiction failures underneath the mirror, and charge accumulation resulting in no actuation are major concerns that the failure analyst needs to address. Typical mirror sizes range from 50µm to 1mm in diameter. Obtaining the surface roughness of one mirror using AFM is achievable, but requires time and several locations for analysis. Typical arrays are found in 16x16, and 256x256. Analyzing the surface roughness of all these mirrors is unrealistic using current AFM analysis techniques. The failure analysis challenge is to develop a tool or technique that can allow RMS roughness along the entire array of mirrors in parallel, in several places along those micromirrors. As shown in Fig. 6, a single micromirror with a large surface is shown in the tilted state. Note the diameter of this surface is 1mm. Measuring the RMS roughness along the entire area of the mirror would be difficult using AFM. Random areas could be characterized, but analysis would be very time consuming. 853

5 Another added difficulty as shown in Fig. 6 would be non-destructive failure analysis underneath the mirror surface to identify shorting or stiction failures. As in RF MEMS, failure analysis underneath the micromirror typically requires deprocessing or removal of the mirror. Both of which can compromise the failure mechanism and lead to erroneous failure analysis results. Although leak detection and deprocessing were discussed earlier, they are pertinent to the analysis of these devices. One area that requires addressing is the application of a diagnostic or test fluid. Part of the overall objective for successful functionality of a device is being able to trace the motion of the fluid during pumping. This may be achievable using IR based systems, but fluids are typically transparent and offer very little contrast to diagnose fluid flow. Using a diagnostic fluid with high contrast or a fluid with high contrast contaminants such as gold spheres allows the failure analyst to better understand the mechanism of fluid flow within the device. This test fluid can also be used to test for leaks, cracks, or breaks in a system. An example of a failed microfluidic device is shown in Fig. 7. Here, the test fluid contained inside the electrostatic drop ejector was water. Upon bias, the powered components were attacked via anodic oxidation resulting in severely damaged components unable to squeeze the fluid through the cap. Fig. 6. A 3x3 micromirror array with the center mirror actuated to 15 o tilt position. 2.5 Microfluidic MEMS This class of MEMS offers the most difficulty for the failure analyst. Most of the tools and techniques currently used for failure analysis were leveraged from the IC industry, and were not designed to be used with fluids. Typical analysis using SEM or other vacuum systems requires the device be flushed of fluid prior to analysis. This can compromise the failure mechanism and lead to erroneous results. The challenges for the failure analyst include functional and structural analysis while maintaining device and tool integrity, fluid contamination and compatibility with MEMS and analytical tools, deprocessing, leak detection, and application of a diagnostic fluid for analysis. In this MEMS technology, the failure analyst should have some background information on fluid mechanics, and a strong knowledge of the fluid used in the device. With the development of the environmental SEM, it may be possible to perform high-resolution analysis of microfluidic devices in a partial pressure system without evacuating the device. This will maintain the component and tool integrity while performing critical analysis. Fig. 7. A failed microfluidic drop ejector. The top level polysilicon cap was removed to reveal the anodically damaged poly 0 power trace. Note the damage along the length of the trace. [6] 2.6 BioMEMS BioMEMS suffer similar problems as microfluidic MEMS with the additional presence of biological material. BioMEMS are more difficult to analyze and diagnose due to a required understanding in the device process, operation, and most importantly, operating environment. The current issues for the failure analyst include biocompatibility (compatibility of the biological material under electrical or mechanical stress over time), functional testing, and device deprocessing. Caps, and/or other membranes, are used to contain the fluid. Analysis of the structural components underneath the cap or membrane is often difficult and requires removal of the upper level. Knowledge and insight into the biological material used during operation will also be 854

6 valuable. In this field, the failure analyst must be multidisciplinary and able to relate the analysis and findings and work with engineers and biologists. An example of a BioMEMS device is shown in Fig. 8a and b. The fluid is pumped through the device via three channels. The polysilicon teeth perforate the cells. Medicinal fluid is added to the cells to repair the contents within the cell. a b Fig.8. a) A BioMEMS device with an integrated cell masher designed to perforate a cell. b) A close-up of the cell masher teeth. The arrow represents the direction of fluid flow. 3. Conclusions MEMS components are extremely diverse in their application and function. Failure analysts will have to be equally diverse and/or multidisciplinary in their analysis of these devices. Many tools and techniques developed from the IC industry have been used for MEMS FA, but more MEMS-specific FA toolsets have to be developed for diagnosis of these failure mechanisms. Many of the devices discussed in this paper have global issues associated with failure analysis. Many non destructive techniques must be developed to assess the failure mechanisms. Tools and techniques that can perform these functions on a larger scale will also be required. To achieve this, industry will have to work with academia and government institutions to create the knowledge base required for tool and technique development for global and local defect localization. 4. Acknowledgements The author would like to thank Ingrid de Wolf (IMEC in Belgium), and Brad Waterson (ADI) for their time, efforts, and contributions to this document. The author would also like to thank Jerry Soden (Sandia) for reviewing this manuscript. The author would also like to thank the Microelectronics Development Laboratory (MDL) at Sandia for their processing efforts. Sandia National Laboratories is a multiprogram laboratory operated by the Sandia Corporation, a Lockheed Martin Company, for the National Nuclear Safety Administration for the United States Department of Energy under Contract DE-AC04-94AL For further information about MEMS technology at Sandia, please visit our website at: 5. References [1] R.W. Beegle, R.W. Brocato, and R.W. Grant, "IMEMS Accelerometer Testing - Test Laboratory Development and Usage," Int. Test Conf., pp , [2] J. A. Walraven, Introduction to Applications and Industries of Microelectromechanical Systems, to be published, proc. of ITC. [3] D. M. Tanner, J. A. Walraven, L. W. Irwin, M. T. Dugger, N. F. Smith, W. P. Eaton, W. M. Miller and S. L. Miller, The Effect of Humidity on the Reliability of a Microengine, Proceedings of IRPS, San Diego CA, 1999, pp [4] J. A. Walraven, J. M. Soden, E. I. Cole Jr., D. M. Tanner, and R. E. Anderson, Human body model, machine model, and charge device model ESD testing of surface micromachined microelectromechanical systems (MEMS), EOS/ESD 2001 Symposium, pp. 3A.6.1 3A [5] J. A. Walraven, E. I. Cole Jr., L. R. Sloan, S. Hietala, C. P. Tigges, and C. W. Dyck, Failure Analysis of Radio Frequency (RF) MEMS, Proc. of SPIE, Oct. 2001, vol. 4558pp [6] J. A. Walraven, P. C. Galambos, E. I. Cole Jr., A. A. Pimentel, G. Roller, A. Gooray, Failure analysis of MEMS electrostatic drop ejectors, Proc. 27 th ISTFA, 2001, pp