The tensile test is the accepted standard for determining mechanical properties of materials used in large structures and components. It, along with the compression test, generates uniform states of stress and strain thereby permitting their direct measurement. However, tension test methods applicable to freestanding thin films were virtually non-existent only 15 years ago. While several researchers using different techniques and procedures now generate comparable results, there is as yet no accepted standard for tension testing of thin films. Neugebauer conducted tensile tests of gold films ranging in thickness from 50 nanometers to 1.5 microns in 1960! It is remarkable that no similar tests followed until the new test method of Read and Dally in 1992 – a gap of 30 years. Neugebauer evaporated gold films with a tensile specimen pattern onto rock salt crystals, glued the crystal ends to a unique test machine, and then dissolved the salt away from the tensile region. Read and Dally’s method is quite similar except the substrate is silicon crystal and removed by selective etching. An easier test method in which one end of the patterned tensile specimen remains attached to the silicon substrate was introduced by Tsuchiya in 1997. The free end of the specimen can be gripped by electrostatic attraction to a probe or by various gluing methods. Unique test machines are developed by each researcher.Strain can be measured by overall elongation of the specimen, which in many cases is sufficient since the specimen can be relatively long compared to its end conditions. Direct strain measurement is of course preferred, and various techniques are used. Sharpe introduced an interferometric technique in 1997. Long used optical digital image correlation (DIC) for gold films in 2001, and Chasiotis has used AFM images in the same manner. Currently, various forms of DIC are being used at this size scale, and it certainly appears to be the preferred technique. These DIC methods have advanced rapidly in the past few years and promise to be even more powerful with computer improvements.Finally, MEMS structures can be used to produce tiny test machines incorporating thin tensile specimens in the processing. Saif’s work of 2004 is an excellent example.This paper is a review of the development of thin-film test methods over the past 15 years. It will describe and discuss the merits of each and include a complete bibliography.

The mechanical characterization of thin films is of great importance for their use as structural layers in MEMS or ICs since device functionality and reliability depend strongly on mechanical parameters. Towards a consistent extraction of mechanical properties, this paper reports on key improvements of both bulge and microtensile techniques applied to an extensive set of materials, brittle and ductile, typically used in MEMS applications.Membranes and microtensile specimens made of silicon nitride, silicon oxide, polycrystalline silicon and aluminum have been fabricated and characterized. Deposition methods such as LPCVD, PECVD, thermal oxidation and PVD have been used. Properties extracted from these materials included pre-stress σ₀, elastic modulus E or plane-strain modulus E_ps, Poisson’s ratio ν, Weibull parameters (strength μ and modulus m) and strain hardening coefficients n. Their dependence on deposition conditions will be emphasized.The bulge test uses a setup where up to 80 membranes on a silicon wafer can be characterized in a single series of sequential measurements. This enables the efficient acquisition of large data sets necessary, e.g., for the determination of reliable fracture data. The membrane profiles resulting from pressure loads are measured with an auto-focus sensor integrated with automated position tables for scanning purposes. A peculiarity of these bulge experiments is the use of diaphragms with large side length aspect ratios. This makes it possible to formulate and apply an accurate, analytical model of the diaphragm mechanics including multilayers with individual prestress and mechanical constants, and the support stiffness.The microtensile test exploits a wafer-detachable test structure design where the specimens are produced on a c-Si frame with coplanar fixed and mobile components defined by DRIE. Delicate handling of the tensile samples is thus avoided. The specimens bridge the gap between the frames experiencing truly uniaxial loads by geometrical constraint when displacements are imposed to the moving component. Such test structures allow the characterization of several material samples in a single experiment and thus contribute to the efficient acquisition of statistically relevant data. Force data are read from a load-cell while local strains are obtained from 2D-interferometric measurements, stroboscopic image analysis and calibrated actuators movement. Current work is developing a wafer scale tensile setup complementing the capability of the bulge test setup.More than 500 membranes of some films have been measured: from the brittle materials analyzed, LPCVD nitride films have the highest fracture strength and are highly predictable (m up to 60). In contrast, oxides are weak and have the lowest Weibull moduli (∼5). Elastic, fracture and ductile properties obtained from both techniques agree with each other and literature, however with narrower uncertainties due to the high throughputs of data.

The measurement of the mechanical properties of submicron sized specimens is extremely challenging due to difficulties for manipulating samples, for applying small load and for extracting accurate stresses and strains. We present a novel, versatile concept of micromachines to measure the mechanical response, up to large strains and fracture, of thin films allowing multiple loading configurations and geometries. This new concept relies on MEMS microfabrication techniques. The first idea is to use internal stresses to actuate and impose the load to the tested sample which is directly deposited on a patterned substrate. The second idea is to multiply elementary machines rather than to build on single complex multi-purpose micromachine. The applied stress and strain are directly inferred from a local displacement measurement. The versatility of the concept allows other loading configurations than simple tension such as shear, biaxial, compression, three-point bending and double cantilever opening. About 5000 elementary micromachines have been processed on top of a single silicon wafer, delivering statistically representative information. The mechanical response of pure Al, Al/Si alloys, Ti and silicon nitride films, with thicknesses ranging between 80 and 500 nm, has been addressed. For instance, stress strain curves have been measured for the Al films, with the strength increasing linearly with 1/thickness to reach values larger than 1 GPa for the thinnest samples while keeping more than 8% of uniform elongation. The strength of Al/Si alloys is two times larger than for Al films. The strength of the Ti films reaches more than 2 GPa. The fracture stress of silicon nitride films attains about 4 to 5 GPa, approaching the theoretical cleavage stress. This method based on MEMS fabrication techniques constitute a combinatorial experimental platform for addressing several critical MEMS reliability issues.

This paper presents the results of an experimental study of fatigue in nanostructured Ni thin films produced using electro-deposition techniques. The underlying mechanisms of fatigue are elucidated in stress-life and short/long crack fatigue crack growth experiments. The studies reveal the potential role of pre-existing material defects in crack nucleation. They also show new experimental evidence of the role of microvoid formation and grain growth during short and long fatigue crack growth of nanostructured Ni MEMS thin films. The implications of the observed phenomena are discussed for potential applications of Ni MEMS structures with nanostructured grain sizes.

Micro devices are widely used in applications for radio frequency communications. In many of these applications, they are exposed to cyclic mechanical loading at typical operating frequencies in the GHz regime. In our work, we have used patterned Al thin films on a piezoelectric substrate to study ultra-high-cycle fatigue. The working principal is based on frequency filtering by the to and fro transduction of an electrical signal into an acoustic wave. An alternating applied electric field generates acoustic deformations on the surface of the piezoelectric substrate by the transduction and can thus be used to stress metallizations at extremely high frequencies. At the resonance frequency, large deflection amplitudes and, hence, high stresses may occur. The devices are tested in the GHz-regime, which allows studying cycle numbers up to 10^14. Fatigue effects, such as the formation of voids and extrusions that result in a frequency shift or even in short circuits are observed.In our work, we have concentrated on the question how cyclic stresses at extremely high frequencies activate defect mechanisms, in particular if they are able to drive dislocations in thin metal films leading to the observed irreversible damage.For this, we have conducted discrete dislocation dynamics simulations showing that dislocations are still able to move in the GHz-regime. It is observed in these simulations that even small asymmetries in the loading conditions lead very quickly to an accumulated plastic strain due to the extremely high frequencies. As a result, the role of the microstructure of the metal films such as grain size or texture is discussed.In order to explore further possible failure mechanisms, tested Al thin film devices were analyzed by Focused Ion Beam microscopy (FIB) and Scanning Electron Microscopy (SEM) evaluating quantitatively the observed extrusions and voids. Furthermore FIB micrographs were used to determine the mean grain size, to investigate whether or not there is a significant influence of the grain size on the extrusion and void formation. Our attempt to correlate details of the dislocation behavior with the damage formation at ultra high cycle fatigue in thin metal films is of general importance for the reliability of micro devices operated at radio frequencies. It becomes obvious that even very moderate loading conditions may become detrimental due to the high frequency connected to an extremely fast accumulation of strain.

Testing of materials properties of micro electromechanical systems (MEMS) is required during the development stage of devices, during fabrication and for analysis of failed components. The testing approaches can be classified in techniques for testing on wafer-level, on die-level and on lift-out samples. Testing on wafer-level is a key element in in-line quality management during fabrication. A fast (few seconds) and non destructive test procedure is required. The testing on wafer-level or die-level is used to detect faulty devices or divide devices on a wafer into different quality classes. The possibility of the detection of bad / or good devices on the wafer-level is an essential cost factor, since unnecessary packing and assembly cost can be avoided and the yield during inte-gration of several dies, e.g. in stacking applications, is dramatically increased. Furthermore ma-terials and also geometric parameters can be monitored during fabrication. Due to the complex-ity of MEMS components a simple electrical measurement is not sufficient. Rather mechanical, chemical, optical, electrical excitations are required and either optical or electrical measurement methods are needed. In the papers the focus is given on techniques for mechanical stimulation and evaluation. Techniques for excitation of MEMS structures on the wafer level, methods for detection of static deformation as well as dynamic vibrations and numerical methods for data evaluation will be presented. Tests on die-level are used to fabricate dedicated test structures, which are tested off-line. The off-line test allows also the destructive-testing of devices. An example of such an test struc-ture are micro-chevron-samples, which can be used for strength characterization of wafer-bonded interfaces. In the paper the current status of this test approach will be shown. Lift-off techniques are required if materials properties of parts of devices has to be measured, e.g. in failed components. An new approach for testing of parts of MEMS structures was devel-oped. The test includes the preparation of test samples by focused ion beam technique milling, handling of test samples and subsequent testing with dedicated test equipment. The approach allows the measurement of mechanical properties like tensile or bend strength or fatigue prop-erties of already fabricated devices. Results of strength investigations will be presented.

The miniaturization of functional structures to the micron and nanometer scales is of widespread technological and scientific interest. Examples include nanotubes, nanowires, mechanical switches, robotic tools, and non-volatile memory elements. As dimensions decrease, surface-to-volume ratios increase, and the effects of properties such as adhesion, friction and damping on system performance become increasingly important. Here, we address the interaction between adhesion and surface topography in dry and humid ambients. To measure adhesion, we employ microcantilever structures using a fracture-mechanics based methodology. In dry ambients, adhesion is as low as several microjoules per square meter. Analysis of the surface topography shows that the highest surface asperities control the average surface separation, but that the main contribution to adhesion is from non-contacting areas. In humid ambients, we observe a strong interaction between relative humidity and surface roughness with millijoule per square meter adhesion values. To analyze the results, a constitutive model for an asperity bridged by a liquid condensate to a flat surface is needed. Under equilibrium conditions, it can be shown that the work of adhesion is less than the surface energy created. This apparent paradox is resolved by a thermodynamic control-volume analysis, and is due to heat-induced evaporation of the liquid during the pull-off process. A more general Griffith criterion results. Experimentally, topographic correlations between the upper and lower surfaces, asperity plasticity and disjoining pressure of the ambient vapor are also important factors. When these effects are all taken into account, good agreement between the experimental and calculated results are found. This work will help enable more reliable micro and nanostructures.

MicroElectroMechanical Systems (MEMS) that have contacting surfaces necessitate surface modification to mitigate adhesion, friction and wear. Most MEMS applications avoid rubbing surfaces because of the challenges associated with reliability in such devices. Surface treatment for non-contacting MEMS is focused mainly on obtaining structures that survive handling associated with production and assembly. A lubrication approach to create reliable sliding contacts would enable a significant expansion of the design space.Chemisorbed monolayers have been used to modify the surface energy of oxidized silicon after etching the sacrificial oxide layers, in order to resist water adsorption and capillary adhesion. A variety of other surface treatments for silicon microsystems have also been investigated, including carbides and oxides to improve wear resistance. Recent emphasis in MEMS lubrication has been on systems that allow for damaged areas to be healed by lubricant diffusion over time. This approach offers some promise, although surface diffusion may limit the effectiveness of this approach for applications involving long duration operation.Interfaces that are deeply buried in complex structures present particular challenges for lubrication. Line-of-sight surface treatment approaches are not applicable to these contacts, and chemisorbed monolayers have exhibited limited mechanical durability in repetitive contact. To address these problems, vapor phase lubrication has been developed for silicon MEMS to enable extremely long duration operation with little or no wear. Molecules in the vapor phase adsorb on silicon surfaces and react to form a friction and wear reducing film. Therefore, as long as the molecule is available in the environment, the lubricant film can be replenished. Micromachined silicon tribometers have been used to demonstrate the vapor phase lubrication concept on MEMS devices, and to compare friction and adhesion behavior with and without the wear-reducing film. Operational lifetime is increased from the order of 1E4 cycles with a chemisorbed monolayer alone, to in excess of 1E8 cycles with vapor phase lubrication. Macro-scale pin-on-disk measurements have been used to produce wear areas large enough for surface analysis by ToF-SIMS. These measurements indicate the formation of a high molecular weight product in the wear scar that is hypothesized to be responsible for the dramatic improvement in tribological behavior. Probable reaction pathways leading to the reactant formation will be discussed.

Chemically bound octadecyltrichlorosilane (ODTS) self-assembled monolayers have been successful in protecting microelectricalmechanical systems (MEMS) from stiction related failure in both the processing stage and the initial stage of use (i.e. “single shot” applications). This lubricant, however, is not successful for MEMS devices expected to cycle repetitively for extended time due, primarily, to the inability to replenish the ODTS in situ. Successful lubrication by ODTS is also limited to a narrow range of temperatures and pressures. A new “Bound + Mobile” lubrication scheme, which combines the bound ODTS with a replenishable mobile tricresyl phosphate (TCP) phase, has been proposed as a solution to these deficiencies. This talk will focus on recent atomistic simulations that characterize both the self-diffusion and the incorporation into defects of the mobile phase as well as the adhesive properties and interfacial structure of the newly formed ODTS/TCP/ODTS interface. This work has been supported by the Air Force Office of Scientific Research, MURI grant # FA9550-04-1-0381.

Fracture of brittle materials fabricated for surface micromachined devices hinges upon understanding the stochastic nature of material failure in conjunction with its microstructure and the material volume under stress. Experiments and numerical analyses were conducted to understand the deterministic and probabilistic parameters entering the problem of fracture in two material systems: Polycrystalline silicon and tetrahedral amorphous diamond-like carbon. Fracture investigations were carried out to determine the effect of material inhomogeneity and stress gradients in brittle films with prefabricated mode-I and mixed-mode critical cracks. It was shown that the location of the crack tip determines the macroscopic failure stress and, thus, the effective critical stress intensity factor for both mode-I and mixed-mode cracks. As a result, the stochastic nature of polysilicon strength is strongly affected by variations in the local critical stress intensity factor. This finding limits extrapolations of statistical probability of failure functions to small material domains that are comparable to grain size, or to specimen sizes that do not guarantee statistical homogeneity. Furthermore, it was shown that through-the-thickness stress gradients are critical in the determination of the material fracture toughness, especially for films that are thicker than one micron. These data were corroborated with an investigation of the accuracy of a single two-parameter Weibull cumulative probabili