Applying a methodology for microtensile analysis of thin films
06/01/2002
by Betty Yeung, Bill Lytle, Vijay Sarihan, Motorola, Tempe, Arizona
David T. Read, National Institute of Standards and Technology, Boulder, Colorado
Yifan Guo, Conexant Systems, Newport Beach, CA
Overview
A microtensile method, used to evaluate material properties of thin films, features test specimens designed and produced using common wafer-processing steps. The microtensile test apparatus incorporates advanced imaging techniques to accurately interpret sample behavior. Measurements of nickel and aluminum films were conducted, highlighting the uniqueness of thin films with respect to bulk materials. Overall, this work shows the necessity and feasibility of characterizing thin films.
Accurately determining the constitutive properties of thin films is a daunting challenge. Their mechanical behavior and characteristics — strength, ductility, grain structure, and elongation — differ significantly from those of bulk materials of the same composition.
We have looked at a microtensile test to address this challenge. The methodology, developed at NIST and implemented at Motorola, uses standard wafer fabrication processes to produce damage-free thin films with a controlled geometry (i.e., a test structure), a microloading system that applies controlled tensile loads to test structures, and a computer that automatically captures test data and images to attain accurate material data.
The method has been used, for example, to evaluate copper films to establish material properties and analytical models of low-k and copper interconnects to assess their mechanical behavior [1]. The work presented here shows our approach to studying aluminum (Al) and nickel (Ni) films [2] for characterization and selection of under-bump metallurgy and processes, ultimately obtaining material properties for mechanical modeling.
To accurately determine thin-film properties, it is essential that damage-free test-structure specimens be prepared in a controlled and reproducible manner while maintaining grain structure and geometry of real thin films that emulate IC fabrication. But it is equally important to produce a specimen that enables the properties of the film to be isolated and evaluated.
 Figure 1. Process steps used to create test structures of a given thin-film metal. |
Our process begins with a <100> silicon wafer with 0.12-0.5μm of thermal oxide (Fig. 1a). We use lithography to pattern rectangular openings and streets between individual die through the oxide to bare silicon, on each side of the wafer (Fig. 1b), creating 150 sites on a 100mm wafer. Then the metal of interest is deposited on one side of the wafer (Fig. 1c), ensuring uniform thickness and complete coverage. Lithography patterning and etching creates defined thin-film specimens aligned to the underlying oxide windows (Fig. 1d). The final step removes the silicon under each specimen site, using wet etching, leaving suspended thin-film metal "paddles" supported across an open hole etched through the silicon wafer (Fig. 1e). By doing so, the die streets are etched to allow cleaving of individual die.
 Figure 2. Typical thin-film specimen test structures. |
The resulting thin-film test structures (Fig. 2) are ~2.5mm long with a gauge length (i.e., the section between the tapers of the specimen) of 600μm as defined by the mask design. Widths can range from 20-200μm. We have evaluated thicknesses, determined by film deposition, from 0.7-6μm. We measure specimen geometries prior to testing, including thickness measurements with profilometry.
 Figure 3. Layout of the microtensile test system. |
Test system
Our microtensile test system includes a loading subsystem and computer control and acquisition capability (Fig. 3) and applies small magnitude forces, typically a few gram-force in increments of a milligram-force or less. This enables application of controlled tensile displacements on the thin film under test and operations or other preparations on test structures without inducing premature damage or loading. These are key aspects to proper testing and handling of thin films. The main components of the test system include grips on which the sample is mounted, force and displacement sensors, and piezoelectric (PZT) stacks that act as expandable columns. All the moving components are cantilevered a small distance from the base of the device to eliminate any effects of sliding friction. The test system's capability provides a displacement range of 50μm ±20nm and a load range of 2 N ±1 nN.
In operation, the PZT stacks expand in response to an applied voltage. This displaces the moving crosshead and induces a tensile load and displacement on the specimen. These are first measured as voltages by force and displacement sensors and then converted, through calibration factors, to actual grip displacement and applied force.
 Figure 4. Microtensile loading of a thin-film specimen. |
We perform tests by mounting a test structure into the tester grips and then breaking away the sides of its silicon frame so that its ends are free. The typical mode of loading a specimen is by applying a constant, linear ramp rate of displacement. Other loading profiles, such as a sine wave or force control, can be applied depending on the output of interest. Displacement is induced on the test structure's moving end (Fig. 4) so that a load is applied until the thin film fractures. During all this testing, data are recorded by the tester's computer control system.
Image analysis
We configure our test structures with tapered ends (see Fig. 2) to decrease stress at the gauge section's ends, thus controlling the failure to remain within the gauge. During testing, the microtensile tester records total displacement of the moving grip, which includes displacement within the gauge section and the tapered regions. Thus, this direct displacement measurement from the test data is an "apparent" displacement or an estimation of gauge length displacement. Such approximations call for a means to more accurately determine displacement, strain, or Young's modulus of the tested film material, which necessitates incorporating other strain measurement techniques. Due to our thin-film test structures' small size, it is impractical to use conventional strain gauges or mechanical extensometers.
Two advanced image-analysis techniques, electronic speckle pattern interferometry (ESPI) [3] and digital image correlation (DIC) [4, 5], provide means to perform accurate strain measurements during actual test. Algorithms for the two methods allow for a defined gauge length to be tracked using photomicroscopy. DIC correlates the displacement of an initial image with a deformed white light sample image. ESPI examines changes in speckle intensity caused by laser light.
 Figure 5. Tensile behavior for Ni thin films. |
A practical test
In recent work, we looked at (Fig. 5) the mechanical properties for Al films and Ni on an Al underlayer (see table).
A film's physical properties and its processing are significant factors affecting resulting mechanical properties. The Ni film is deposited by plating, while its Al underlayer is applied by a sputtering process. The stress-strain data indicate that the Ni film is characterized by elastic behavior. The measured tensile strength of this thin film is higher than bulk values. The difference in geometrical magnitude in the film and bulk material provides an explanation of this difference. The bulk material testing involves a relatively large volume through which numerous minute defects exist. As defects would degrade the integrity and strength of a material, the overall strength of the bulk material is decreased. On the other hand, the nature of the thin film is such that its thickness is on the order of the material's grain size. The relative defectivity within the thin film as compared to that within its bulk counterpart is less.
 |
Conclusion
The use of standard thin-film processing techniques for sample preparation combined with a microtensile system and imaging techniques enables evaluation of actual thin-film materials used in IC fabrication. The microtensile method couples the preparation and use of damage-free tensile specimens of thin films, application of small-scale loads at appropriate levels, and refined deformation measurements for thin-film material properties to be determined accurately. The measurement of a material, such as Ni, indicated thin-film properties dissimilar to bulk Ni properties.
These results verify the importance, necessity, and feasibility of distinguishing thin-film materials from their bulk counterparts. This ultimately allows for accurate representation of thin films in a variety of applications, including reliability assessment and material input for predictive engineering through simulation.
Acknowledgments
This document includes contributions from the National Institute of Standards and Technology (NIST), a US government agency.
References
1. A.A. Volinsky, et al., MRS Symposium Q Symposium Proceedings, Fall 2000.
2. M. Grupen-Shemansky, et al., ASME Advances in Electronic Packaging, EEP-Vol. 26-2, Vol. 2, 1999, pp. 1303-1308.
3. D.T. Read, Measurement Science and Technology, 9, 1998, pp. 676-685.
4. W.H. Peters, Optical Engineering, 2 (3), 1982, pp. 427-431.
5. M. Sutton, et al., "Photomechanics," Top. in Applied Physics, 77, 2000, pp. 323-372.
Betty Yeung received her BS from Duke University and is a package design engineer at Motorola Inc., Tempe, AZ 85284; ph 480/413-7734, fax 480/413-4511, e-mail [email protected].
Bill Lytle, a Motorola development engineer in flip chip packaging technology, received his MS from Arizona State University.
Vijay Sarihan, a distinguished member of Motorola's technical staff, received his PhD from Cornell University.
David T. Read, a NIST physicist, received his PhD from U. of Illinois.
Yifan Guo, Thermal Mechanical Reliability Group manager at Conexant Systems, received his PhD from Virginia Polytechnic Institute.