Failure Analysis Associates
Framingham, MA 01701
Cambridge, MA 02142
Univ of Wisconsin-Madison
Engr Bldg Rm 201 AHL
Madison, WS 53706
Dept of EE&CS
Univ of California-Berkeley
497 Cory Hall
Berkeley, CA 94720-1774
Dept of Mechanical Engr
Univ of California-Berkeley
Berkeley, CA 94720-1740
Lawrence Livermore National Laboratory
Mail Stop L-222
Livermore, CA 94551
Proceedings published as Volume 518
of the Materials Research Society
Symposium Proceedings Series.
* Invited paper
8:30 AM *N1.1
Chairs: Stuart Brown and Henry Guckel
Wednesday Morning, April 15, 1998
PASSIVATION AND SUBSTRATE EFFECTS ON THE YIELD STRENGTHS OF GOLD FILMS. Omar S. Leung
, William D. Nix, Stanford Univ., Dept of Materials Science & Engineering, Stanford, CA.
It is known that the yield strengths of thin films on substrates depend on grain size, passivation, and the constraint of the substrate. Grain boundaries affect yield strength by preventing dislocations from moving within the film while passivations and substrates affect yield strength by preventing dislocations from exiting the surfaces of the film. The effects of passivation and substrate are expected to be related to the film thickness, but it can be difficult to quantify these relations because grain size and film thickness are usually interrelated.
In order to study the effect of passivations on the yield strengths of films on substrates, wafer curvature tests have been conducted on bare and passivated gold films deposited onto silicon wafers. During the heating stage of these tests, the grains diameters grow from about one fourth of the film thickness to about the film thickness. Windows are then etched into these wafers and into unheated wafers in order to measure the yield strengths by bulge testing. By performing bulge tests on free standing gold films with various grain sizes, we are able to measure the grain size contribution to yield strength without passivation or substrate effects. These results are then compared to results of bulge tests on titanium oxide-passivated Au films. This enables us to isolate the effect of the passivation on yield strength. Comparison of the bulge tests with the wafer curvature tests for a given film then allows us to isolate the effect of the substrate on the film strength.
9:00 AM N1.2
ADVANCES IN FATIGUE TESTING OF MEMS MATERIALS. Stuart Brown
, Christopher Muhlstein, William van Arsdell, Failure Analysis Associates, Cambrige, MA; Dennis Freeman, MIT, Dept of Electrical Engineering, Cambridge, MA.
Although microelectromechanical structures have been fabricated from single crystal and polycrystalline silicon for many years, the long-term response of these structures has not been thoroughly investigated. Long term reliability is important as safety critical silicon MEMS are currently being produced with expected operating lifetimes exceeding 10 years. Other commercial applications expose silicon sensors to elevated temperatures, combustion gasses, and moisture. This effort both sizes several years of fatigue testing of silicon and presents recent results on the effect of temperature and water on silicon durability. The basic technique used in this analysis consist of resonant silicon beams with deliberately introduced cracks and stress concentrations. A change in resonant frequency indicates either crack growth or the initiation of cracking. The sensitivity of this technique permits the measurement of crack front increments of less than 1 nanometer and crack tip velocities of less than 10-14
meters per second.
9:15 AM N1.3
RESONANCE METHOD: AN ATTRACTIVE WAY TO EVALUATE MECHANICAL PROPERTIES OF THIN GOLD FILMS. Patrick Attia
, Patrice Hesto, Institut d'Electronique Fondamentale, Universite Paris XI Orsay, Orsay, FRANCE.
Micromechanical test structures have been developed for evaluating mechanical properties of thin gold films. Using micromachining techniques, gold cantilever beams, of different geometry, were fabricated. These beams which are tpically 40 to 60 m long, 1
m thick and 2 m wide were resonated by electrical excitation, in situ
in a SEM. From these measurements, Young's modulus, Q-factor and fatigue propertles were deduced. The measured resonant frequency and the knowledge of the device geometry allowed the Young's modulus of these thin films to be evaluated. The resonance frequency is determinate by the frequency of the applied voltage that produced the maximum vibrational amplitude of the beam. The first undamped transverse resonant frequency of a cantilever beam is: where E, , t and L are respectively the Young's moodulus, density, thickness and length of the cantilever. Measured for cantilever beams with different lengths the frequencies showed a linear dependance against 1/L2
. Using Eq. (1), the Young's modulus was found to be about 90 GPa. This value is in total agreement with both the bulk and published values for such thin films. This method is also capable of evaluating experimentally the oscillating quality factor Q. The cantilevers described here, exhibit a relatively high Q factor (500 - 650). This factor is probably related to internal losses, controlling the intrinsic material Qi-factor. Thc fatigue properties of the materials are inferred from a shift in the resonant frequency as a function of time. We note that the resonan frequency of the beams increased around l.3% for a number of vibrational cycles around 10 billons. Assuming that the dimensions of the beams are not affected by the ageing process, an increase of the resonant frequency could be directly related to an increase in Young's modulus.
9:30 AM N1.4
TRANSVERSAL TYPE PIEZOELECTRIC RESONATOR USING ZnO THIN FILM ON MICRO-FABRICATED ELINVER (Fe-Ni-Cr) ALLOY. Yukio Yoshino
, Murata Mfg. Co., Ltd., R&D Division, Yokohama, Kanagawa, JAPAN.
A transversal type 3.58MHz piezoelectric resonator has been fabricated using ZnO thin film on ELINVER (Fe-Ni-Cr) alloy. The ZnO/ELINVER structure piezoelectric resonator has been designed to have a 150ppm temperature coefficient of frequency (TCF) from -20 degrees to 70 degrees centigrade. The temperature coefficient of ELINVER alloy can be controlled to cancel the TCF of ZnO thin film by heat annealing. ZnO thin film on ELINVER alloy shows good c-axis orientation equivalent to that of ZnO thin film on a glass substrate. However, the c-axis orientation of ZnO thin film is influenced by surface roughness of ELINVER alloy. The wet etching process has been adopted to shape the resonator made of ELINVER alloy. The cross section of the resonator is a tapered structure created using different sizes of photo mask on each side of the ELINVER surface. The tapered cross section of the transversal type resonator greatly improves the frequency characteristics of the resonator. The electric characteristics of the resonator after the improvement include resonation frequency of 3.58MHz which is trimmed by a YAG laser, and resonation resistance of about 50 . The temperature coefficient of frequency is under 150ppm at the temperature range of -20 degrees to 70 degrees centigrade.
10:15 AM N1.5
MULTILAYER MICROELECTROMECHANICAL STRUCTURES FOR MATERIAL PROPERTY CHARACTERIZATION. R.I. Pratt
and G.C. Johnson Department of Mechanical Engineering, University of California, Berkeley, CA.
A new class of micro-electro-mechanical structures (MEMS) is presented. These multilayered structures are based on the lateral resonators described by Tang, et al.  and Pratt, et al., and are designed to characterize material properties of the different materials making up the system. Of particular significance is the ability to characterize many new materials previously untestable by resonant testing techniques.
The basic lateral resonant structures are made of a single material, usually doped polycrystalline silicon. They consist of beams which are anchored to a substrate at one end and support a rigid mass at the other. The rigid mass has a comb shaped region on each side. These combs fit interdigitally into similar comb shaped regions which are also fixed to the substrate. These combs serve to electrostatically drive the structure, as well as to electrically sense the subsequent motion. Since the ability to drive the structure and sense its motion requires that the structure be made of a material that is electrically conductive, the class of materials which is characterizable by resonance techniques has, until now, been somewhat limited. However, if multlayer structures are tested in conjunction with these basic monolithic structures, additional properties, both of the base material and of the subsequent layers, can be determined. Such properties as stiffness and mass density of many new materials can be obtained by examining the behavior of multilayer beams and comparing them to their single layer counterparts. Experimental designs and the associated analytical techniques for obtaining various properties will be presented.
10:30 AM N1.6
THE EFFECTS OF TEXTURE ON THE YOUNG'S MODULUS OF POLYSILICON. Sangwoo Lee
, Dong-il Dan Cho, Seoul National University, Seoul, KOREA.
This paper investigates the effects of texture, annealing condition, and phosphor doping on the Young's modulus of 2um-thick polysilicon microstructures. Linear lateral comb-drive resonators were used to measure Young's modulus. In contrast to other published data, which report almost no dependence of Young's modulus on the texture, our results show the strongest dependence on the texture, and almost no dependence on annealing and doping conditions. This texture dependence can be explained by noting the fact that the Young's modulus of single crystal silicon on a (111) plane is 168.9 GPa and that on a (110) plane varies between 130.2 GPa in <001> direction and 187.5 GPa in <111> direction. Ignoring that effects of grain boundaries, for polysilicon with <110> preferred orientation, a value between 130.2 GPa and 187.5 GPa is expected. Considering crystallography, this value should be lower than 158.9 GPa, and the value should be isotropic. For polysilicon with mixed <111> and <110> preferred orientations, Young's modulus is expected to lie between that value and 168.9 GPa, depending on the ratio of those orientations. For polysilicon structures deposited at 585ºC, annealed and doped at various conditions, the measured ratio of <111> grain volume to <110> grain volume was 49.72.9. The measured Young's modulus was 167.34.9 GPa. For polysilicon deposited at 605ºC, the grain volume ratio was 8.72.8, and the experimental Young's modulus was 160.51.3 GPa. For polysilicon deposited at 625ºC, the ratio was 6.82.7, so the Young's modulus is expected to be similar to that of the 605ºC polysilicon. However, the measured values was 153.71.8 GPa, which is slightly lower. The measured resistivity of the 625ºC polysilicon was 23.5 higher than that for the 605ºC polysilicon prepared at the same doping conditions, implying that a larger number of grain boundaries are present. A large number of grain boundaries in general results in a larger resistance and a smaller Young's modulus.
10:45 AM N1.7
MECHANICAL PROPERTY MEASUREMENT OF 0.5-m CMOS MICROSTRUCTURES. Michael S.-C. Lu
, Xu Zhu, Carnegie Mellon University, Dept. of Electrical and Computer Engineering, Pittsburgh, PA; Gary K. Fedder, Carnegie Mellon University, Dept. of Electrical and Computer Engineering & the Robotics Institute, Pittsburgh, PA.
Microstructures integrated with CMOS are made from combinations of aluminum, silicon oxide, and silicon nitride thin films by undercutting the silicon substrate. The dual function of metallization and dielectric layers as an isolated electrical interconnect and as a structural material, and the ability to use different combinations of layers to form structures provide more degrees of freedom for designing microelectromechanical devices than homogeneous MEMS materials (e.g., polysilicon). In this paper, we will present the design of metrology test structures made from the fourteen possihle composite structures, and report on measurements of effective Young's modulus, effective residual stress and curling on a dice-to-dice and run-to-run basis. Microstructures are fabricated by post-processing chips made in Hewlett-Packard 3-metal n-well 0.5-m CMOS. The key to the fabrication process is use of the top metal layer in the CMOS process as a mask to define the composite metal/dielectric microstuctures with thickness up to 5.3 m and minimum beam widths and gaps of 1.2 m. Structures are made by anisotropic dry etching of the exposed dielectric layers to form sidewalls, followed by isotropic plasma etching of the silicon substrate to release the structures. A compact test structure layout of 2 mm by 2 mm in size is submitted monthly to gather statistical data on material properties. Test structures include cantilever arrays for material stiffness and stress gradient, bent-beam sensors for effective residual stress, n-well structures for silicon undercut, and a large-mass resonator for failure characterization. Lateral bending in very narrow beams is observed because of the misalignment of metal layers. Different beam compositions have different values of stiffness, curl, and residual stress. For a nominal 3-metal composite beam, the effective Young's modulus is 52 GPa, its radius of curvature is 4.8 mm, and residual stress is 69 MPa. Silicon undercut is dependent on the post-CMOS processing and on hole size, and is about 12 m for open edges.
11:00 AM N1.8
HEATING EFFECTS ON THE YOUNGíS MODULUS OF FILMS SPUTTERED ONTO MICROMACHINED RESONATORS. H. Kahn
, M.A. Huff*, and A.H. Heuer, Department of Materials Science and Engineering, *Department of Electrical Engineering and Applied Physics, Case Western Reserve University, Cleveland, OH.
The use of surface-micromachined comb-driven lateral resonant structures is a well-established technique for measuring the Youngís modulus of structural materials (typically polysilicon) which comprise the resonator. An ac signal drives the structures into resonance at a frequency which is determined by the stiffness of the folded-beam supports. The stiffness depends on the geometry and the Youngís modulus of the material.
In this study, metal films are sputter-deposited onto fully released micromachined polysilicon resonators. The stiffness of the composite beams is set by the Youngís moduli of both materials. Therefore, once the modulus of the polysilicon is initially determined, the measured resonance frequency of the composite structure will reveal the modulus of the sputtered material. In addition, during testing, a dc current is passed through the folded-beam supports which heats the beams, particularly the low resistance sputtered metal. In this manner, the effects of heating on the Youngís modulus are studied. Specifically, a sputtered TiNi shape memory alloy film has been investigated; it exhibits two distinct values of modulus with an intermediate transition region, indicative of the temperature-induced reversible martensitic transformation displayed by shape memory materials.
11:15 AM N1.9
MEASUREMENT OF MECHANICAL PROPERTIES IN SMALL DIMENSIONS BY MICROBEAM DEFLECTION. O. Kraft
, R. Schwaiger, Max-Planck-Institut für Metallforschung, Stuttgart, GERMANY; and W.D. Nix, Stanford University, Stanford, CA.
In this paper, we describe a technique for measuring elastic and plastic properties as well as fracture in small dimensions by microbeam deflection. The test procedure is similar to a macroscopic bending test where an external load is applied to a beam and its bending measured. Silicon and silicon oxide microbeams were made by different growth, lithography and etching techniques. The beams were then deflected with a nanoindenter, while the applied load and the displacement of the beam were recorded. The elastic properties were determined from the initially linear load-displacement behavior. Brittle fracture of the beams was seen as an abrupt drop in the applied load. Beams made from wet-thermal oxide showed a higher strength than ones made from oxide grown by a CVD process. In order to study properties of thin film materials the method can be extended by depositing films onto the beams and measure the behavior of the beam/film composite. For instance, yielding of a metal film is seen as a deviation from the linear-elastic behavior of the beam. In this paper, we present results on metal films such as Al, Cu and Pt. Finally, it is outlined how the technique can also be used to determine fracture toughness of the interface between films or to characterize the fatigue behavior in small dimensions.
11:30 AM N1.10
TORSIONAL FAILURE STUDY OF SINGLE CRYSTAL SILICON BARS USING MEMS MICRO INSTRUMENTS. Taher Saif
, Mechanical and Industrial Engineering, University of Illinois at Urbana-Champaign, Urbana, IL; N. C. MacDonald, Electrical Engineering, Cornell University, Ithaca, NY.
MEMS (MicroElectroMechanical Systems) based micro instruments allow sub-micron materials characterization, and in-situ
observation of failure mechanisms. Furthermore the samples can be fabricated in the same way as the MEMS and micro electronic components are fabricated to reveal the real life mechanisms of failure. We present an experimental study on a single crystal silicon (SCS) bar subjected to pure torsion using MEMS micro instruments. The bar is in the form of a pillar, anchored at one end to the silicon substrate. It is attached to a lever arm at the other end. The pillar has a minimum cross sectional area at its mid height. The cross section coincides with the (100) plane of SCS. Torsion is generated by applying two equal forces on the lever arm on either side of the pillar. Two micro instruments apply the forces. Each consists of an electrostatic actuator and a component that calibrates it. The actuator generates high force ( at 50 V
) and is capable of developing large displacements ( ). Calibration involves determination of the force generated by the actuator at an applied voltage, as well as the linear and higher order spring constants of its springs [MRS 1997 Fall meeting (NN1.3), Boston]. Each microinstrument is thus calibrated independently. With the application of forces by the two micro instruments, a torque is generated which twists the pillar. The angle of twist at different applied voltages are recorded using an angular scale. The corresponding torques are determined from the calibration parameters of the actuators. Torque is applied until the pillar fractures. Two such sample pillars, samples 1 and 2, are tested. There cross sectional areas are 1 and 2.25 . We find that both the pillars behave linearly until failure. The stresses prior to fracture are evaluated based on anisotropic theory of elasticity. Samples 1 and 2 fail at shear stresses of 5.6 and 2.6 GPa
respectively. The fracture surfaces seem to coincide with the (111) plane of SCS.
11:45 AM N1.11
YOUNG'S MODULUS, YIELD STRENGTH AND FRACTURE STRENGTH OF MICROELEMENTS DETERMINED BY TENSILE TESTING. Staffan Greek
, Fredric Ericson, Uppsala University, Dept of Materials Science, Uppsala, SWEDEN.
Some mechanical properties of thin film microelements, e.g. fracture strength, depend on the manufacturing process, the load application as well as on size and shape of the microelements. Hence, the test structures that are used to determine mechanical properties should have dimensions of the same order of magnitude as an application structure, i.e. microelements must be used to accurately characterize MEMS. The fabrication of test structures should not depend on a specific process but must be realized in the same process as an intended application in order to give accurate results. Microelements are easily viewed in an SEM, but to be handled and tested in situ a micromanipulator was developed. Test structures were designed as released beams fixed to the substrate at one end, with a ring at the other. A high-precision testing unit was mounted on the micromanipulator next to the test structures. In the SEM, the testing unit was manoeuvered to grip the ring of the test structure beam and a tensile test of the beam was then executed. From the test data Young's modulus and fracture strenth of polysilicon and single crystalline silicon were evaluated. The fracture surfaces were examined and compared. Young's modulus, yield strenth and fracture strength of microelements made from electroplated nickel and nickel -iron alloy were also measured. Relative measurement of test structures with different beam lengths enabled Young's modulus to be evaluated with an accuracy of +/- 5%.
Chairs: John Gilbert and George Johnson
Wednesday Afternoon, April 15, 1998
1:30 PM *N2.1
ROUND-ROBIN TESTS OF MODULUS AND STRENGTH OF POLYSILICON. Stuart Brown, Failure Analysis Associates, Cambridge, MA; George Johnson, U. C. Berkeley, Dept of Mechanical Engineering, Berkeley, CA; Wolfgang Knauss, Cal Tech, Aeronautics and Applied Mechanics. Pasadena, CA; William Sharpe, Johns Hopkins Univ, Dept of Mechanical Engineering, Baltimore, MD.
Various approaches to measuring the mechanical properties of thin-film polysilicon have been developed over the past several years, and there is now a need to compare these to assess their advantages and disadvantages. The Air Force Office of Scientific Research (Capt Brian Sanders, Program Manager) has recognized this and sponsored a limited initial Cross-Comparison or Round-Robin with four participants.
The geometries and loadings are quite different. Brown uses a curved electrostatic comb to deflect a specimen in in-plane bending. Johnson subjects cantilever beams of different lengths to in-plane bending and also deflects a fixed-fixed beam with an electrostatic comb drive. Knauss and Sharpe pull in-plane tension specimens with electrostatic gripping and measure strain directly on the specimen - Knauss with atomic force microscopy and Sharpe with laser interferometry.
The specimens are manufactured at the Microelectronics Center of North Carolina (MCNC) on centimeter-square dies at adjacent locations on the wafer. They are also released by MCNC in an effort to treat the material as uniformly as possible. The presentation will focus on the results obtained with brief descriptions of the four different techniques and procedures.
2:00 PM N2.2
CURVATURE OF A CANTILEVER BEAM SUBJECTED TO AN EQUI-BIAXIAL BENDING MOMENT. Peter Krulevitch, Lawrence Livermore National Laboratory, MicroTechnology Center, Livermore, CA; George C. Johnson, University of California, Berkeley, Dept. of Mechanical Engineering, Berkeley, CA.
The problem of a cantilever beam subjected to an internal bending moment has a number of practical applications, including the analysis of residual stress in thin deposited films. The simple bending cantilever structure has seen extensive use in the form of thin film micromechanical devices, such as thermal and shape memory film bimorph microactuators, IR detectors, and microfabricated valves. Another relevant problem is the curvature of thin film cantilever beams subjected to gradients in residual stress. In the usual analysis, the transverse bending moment is neglected when the cantilever is sufficiently narrow. In this paper we demonstrate that this assumption is not valid.
When a cantilever beam is subjected to a uniaxial moment M, bending it to an axial radius of curvature r, the transverse cross section undergoes an anticlastic curvature. The transverse radius of curvature, -r/, is inversely proportional to Poisson's ratio . On the other hand, a plate subjected to an equi-biaxial bending moment of magnitude M exhibits the same radius of curvature, r/(1-), in both the transverse and axial directions. An example of this type of loading is a plate-like substrate uniformly coated with a thin film under residual stress. We address the following question in this paper: as the width of the plate (under the bi-axial moment) is reduced such that it has the dimensions of a beam, does the axial curvature retain its 1/(1-) dependence? We show that the biaxial modulus, given by E/(1-), must be used even for narrow beams when subjected to a biaxial bending moment.
2:15 PM N2.3
MESO(INTERMEDIATE)-SCALE ELECTROMECHANICAL SYSTEMS FOR THE MEASUREMENTS AND CONTROL OF SAGGING IN LTCC STRUCTURES. Mario Gongora-Rubio, IPT, Sao Paulo, Brazil; Jorge Santiago-Aviles, EE Dept, University of Pennsylvania, Philadelphia, PA; Luis Sola-Laguna, DuPont Experimental Station, Wilmington, DE.
Sagging of suspended or laminated structures is a common problem in the processing of Low Temperature Cofired Ceramics (LTCC). These glass-ceramic composites are susceptible to plastic deformation under the stress of body forces once the glass transition temperature of the binder is reached during processing. We have designed and fabricated using the conventional methods of LTCC fabrication, Meso-scale structures (ranging in size from 100 micrometers to one centimeter) to quantify and seek control strategies for this problem. We have implemented bridge structures, cantilever beams and membranes to emulate most of the conventional structures encountered during packaging or sensor device fabrication. Among the results obtained are the fact that when an LTCC tape with holes of diameters in excess of 400 micrometers is laminated, the tapes above and below deform in the inside perimeter of the cavity, but for smaller diameters the deformation is negligible. In the case of bridging structures, one can compensate for the potential effect of body forces by screen printing a thick film which form an overlayer with tensile internal stresses upon sintering. This compensates for the body forces deformation and often yields straight bridges. The use of fugitive phase materials which may disapear or flow during firing is another way to support bridging structures. We have explored several of these materials, and results will be presented. Another attempt at sagging control is the use of a sacrificial glass layer for structure support. Experimental results of the use of this technique will be shown.
2:30 PM N2.4
FATIGUE TESTING OF THIN FILM MATERIALS FOR MEMS APPLICATIONS.
Abstract not available.