IEC 62047-21: Semiconductor Devices — MEMS — Part 21: Test Methods for MEMS Thin Films

IEC 62047-21, published in 2014 as part of the IEC 62047 series on MEMS (Microelectromechanical Systems) devices, specifies standardized test methods for determining the mechanical properties of thin films used in MEMS fabrication. The standard was developed by IEC Technical Committee 47 (Semiconductor devices) and addresses a critical challenge in MEMS engineering: the mechanical properties of thin-film materials at the micrometer scale can differ significantly from their bulk material counterparts, and these properties directly determine device performance, yield, and long-term reliability. As MEMS technology has expanded from automotive sensors to consumer electronics, biomedical implants, optical networks, and 5G RF front-ends, the need for reliable thin-film characterization has become increasingly important.

The standard covers test methods for determining key mechanical parameters: Young’s modulus (elastic modulus), residual stress, fracture strength, hardness, and creep behavior. These parameters are essential inputs for MEMS design simulation, process optimization, and reliability prediction. The thin-film materials addressed include polysilicon, silicon nitride, silicon dioxide, aluminum, copper, gold, nickel, and various dielectric and piezoelectric thin films commonly used in MEMS fabrication processes. The test methods are designed to be compatible with standard semiconductor fabrication processes and can be implemented using on-chip test structures that are fabricated alongside the device components.

Test Methods for Mechanical Characterization

IEC 62047-21 specifies several complementary test methods, each with specific advantages for measuring different mechanical properties. The tensile test method uses on-chip test structures where thin-film specimens are fabricated with specially designed gripping features. A micro-force actuator applies tensile load while a displacement sensor measures elongation. From the stress-strain curve, Young’s modulus, yield strength, ultimate tensile strength, and fracture strain can be determined. The standard specifies specimen geometry requirements including gauge length (typically 100-500 micrometers), width (10-50 micrometers), and thickness (0.1-10 micrometers). The alignment accuracy between the loading direction and the specimen axis must be within 1 degree to avoid bending artifacts.

The beam bending method provides an alternative approach for measuring Young’s modulus and fracture strength. Both cantilever and fixed-fixed (bridge) beam configurations are specified. In the cantilever method, a known force is applied at the free end using a nanoindenter or atomic force microscope (AFM) probe, and the resulting deflection is measured. The spring constant of the cantilever relates directly to Young’s modulus through the beam geometry. The fixed-fixed beam method uses pressure loading or electrostatic actuation to deflect the beam and extract the mechanical properties from the deflection-voltage relationship. The standard provides detailed formulas for extracting material properties from the measured force-deflection data, accounting for factors such as residual stress, beam geometry imperfections, and substrate compliance.

Legend: E = Young’s modulus, sigma_y = yield strength, sigma_UTS = ultimate tensile strength, epsilon_f = fracture strain, sigma_f = fracture strength, H = hardness.

The resonance method offers the highest accuracy for Young’s modulus determination by measuring the resonant frequency of micro-resonator structures fabricated from the thin film of interest. The standard specifies both electrostatic and piezoelectric actuation methods, with optical or capacitive detection of the vibration amplitude. From the measured resonant frequency and the known geometry, Young’s modulus can be calculated with an accuracy of approximately +/- 2%. This method is particularly attractive because it is non-destructive, can be performed at wafer level, and is compatible with automated test equipment used in semiconductor manufacturing. However, it requires careful design of the resonator structure to isolate the desired vibration mode and avoid coupling with parasitic modes from the substrate and support structures.

Specimen Preparation and Data Analysis Requirements

Reliable thin-film testing begins with proper specimen preparation. IEC 62047-21 specifies detailed requirements for test structure design and fabrication to ensure that measured properties represent the intrinsic material behavior rather than artifacts of the test structure or measurement method. The specimen must be free from process-induced damage including etch residues, ion implantation damage from dry etching, and thermal stress from deposition or annealing processes. For surface micromachined specimens, the release process must remove sacrificial layers without damaging the test structure. The standard specifies acceptable release methods including HF vapor etching for silicon dioxide sacrificial layers and XeF2 dry etching for silicon sacrificial layers.

Data analysis procedures are specified in detail to ensure consistency across different laboratories. For tensile testing, the stress-strain curve must be corrected for system compliance (the elastic deformation of the test apparatus), initial slack in the specimen-gripping system, and thermal drift during the measurement. For beam bending methods, the force-deflection data must be analyzed using large-deflection theory when the deflection exceeds 10% of the beam thickness, since the linear (small-deflection) beam theory becomes inaccurate. The standard provides correction formulas for each case and specifies the acceptable ranges for applying simplified analytical models versus requiring finite element analysis (FEA) for data interpretation.

Engineering Design Insights for MEMS Thin Films

From a MEMS design perspective, understanding and managing thin-film mechanical properties is essential for achieving reliable device performance. Several key insights emerge from the application of IEC 62047-21 test methods. First, residual stress gradients through the film thickness are often more critical than the average stress value. A stress gradient causes cantilever beams to bend upward or downward, affecting device performance in accelerometers, gyroscopes, and micromirrors. The standard recommends measuring curvature on released cantilever arrays of different lengths to extract both the average stress and the stress gradient, using Stoney’s equation modified for continuous films.

Second, fracture strength of thin films follows a Weibull distribution rather than a normal distribution, reflecting the statistics of flaw-controlled brittle fracture at the microscale. The standard recommends reporting Weibull modulus (m) alongside the mean fracture strength, as the Weibull modulus directly indicates the reliability and predictability of the material strength. A high Weibull modulus (m > 10) indicates consistent, predictable strength, while a low modulus (m < 5) indicates high variability and requires larger safety margins in design. For polysilicon, typical Weibull modulus values range from 5 to 12 depending on the deposition conditions and surface roughness.

Third, the fatigue behavior of thin films differs dramatically from bulk materials. In bulk metals, fatigue failure typically occurs after 10^3 to 10^7 cycles. In MEMS thin films, particularly silicon-based materials, fatigue has been observed at extremely high cycle counts (10^9 to 10^12 cycles) under high-frequency actuation, with a mechanism involving progressive oxidation and moisture-assisted crack growth at the surface rather than classical dislocation-mediated fatigue. The standard provides guidelines for accelerated fatigue testing using resonant structures, enabling lifetime prediction for MEMS devices that undergo billions of cycles during their operational life.

Fourth, temperature dependence of mechanical properties is critical for MEMS devices operating over a wide temperature range. Young’s modulus of silicon decreases by approximately 50 ppm/deg C, while the thermal expansion coefficient is 2.6 ppm/deg C. These temperature coefficients, while small, can cause significant performance drift in precision MEMS sensors and must be characterized using the test methods specified in the standard at multiple temperatures (typically -40, 25, 85, and 125 deg C). The temperature-dependent data enables designers to implement compensation algorithms and predict device behavior across the entire operating temperature range.