The Measurement of Cyclic Creep Behavior in Copper Thin Film Using Microtensile Testing (MEMS and Nanotechnology)

ABSTRACT

A micro-tensile testing for studying the cyclic fatigue mechanical properties of freestanding copper thin film with thickness of sub-micrometer application for MEMS was performed to observe its mechanical response under tension-tension fatigue experiments with a variety of mean stress conditions at cyclic loading frequencies up to 20 Hz. Tensile sample loading was applied using a piezoelectric actuator. Loads were measured using a capacitance gap sensor with a mechanical coupling to the sample. The experiments were carried out with feedback to give load control on sputter deposited 300, 500 and 900 nm copper thin films. Loading cycles to failure reached over 10A6 at low mean load with a trend of decreasing cycles to failure with increasing mean load as anticipated. The cyclic fatigue results provided clear evidence for a cyclic creep rate dependent and change in failure mechanism from crack formation to extended plasticity as the mean load is decreased.

Introduction

With the fast development of the microelectronics technology, the system and device design required further developing the complexity and packing density of micro-nano scale devices into the future allowing ever smaller in the range of nanometer scale and more densely packed structures to be fabricated. Continued growth of microsystem technology requires still further miniaturization, with a corresponding need to understand how length scales affect mechanical behavior. As a result, the mechanical properties of sub micron and nano scale thin films have become one of the most important issues.


Moreover, MEMS (Micro Electro Mechanical Systems) has become an important technology. The goal of MEMS is to integrate many types of miniature devices on a single chip and has been used in a variety of applications. To date, fabrication of MEMS devices has generally been performed using conventional integrated circuit fabrication techniques. Thus, materials properties play a role of importance in MEMS as it has performed in the integrated circuit applications. In recent study of MEMS fabrication, design, and testing confirms that devices whose fabrication, design, and operational attributes including environment have been optimized rarely exhibit bulk mechanical failure by fatigue or fracture [1]. Instead, MEMS failure mechanisms may associate degradation of reflecting MEMS thin films including surfaces, grain growth, deformation and time dependent deformation due to creep or fatigue.

Among various materials application for microelectronics and MEMS, copper thin film is one of the most common used thin films materials since its electrical conductivity. In particular, cyclic behavior of copper thin films, could ultimately limit product lifetime in its applications. Example such as micro-machined copper thin film resonators requires a vacuum package to obtain the highly selective frequency response. However, the resonant frequency is strongly dependent upon the device dimensions and characteristics of deposited structural thin films. Thus, a reliable design parameter, such as mechanical properties of copper films is needed in order to overcome the process variation, inherent mechanical stresses as well as cycle fatigues.

Here, the experimental testing technique whose characteristics represent nominal operating conditions is demonstrated. It is designed to use a specimen with electroplated frames, pins align holes, misalignment compensates springs and a load sensor beam connected to a freestanding microbeam. The specimen is then fitted into the microtensile apparatus to carry out a series of micromechanical uniaxial tensile and displacement controlled tension-tension fatigue experiments tests.

EXPERIMENTAL PROCEDURES

The experimental setup consists of a tensile test chip, a piezoelectric actuator and a load sensing capacitor. The original design were reported previously [3,5 ], a number of changes were made both in the system and sample design and fabrication to make it more robust in terms of consistencies of process yields, tests reliability and operation efficiency.

Overview of the sample design and fabrication

The sample design consists of a freestanding beam and a supporting frame connected to the piezoelectric actuator. Figure 1 shows the schematic of the test chip, the details specimen geometry and the dimension. The central part of the testing chip is the sputtered freestanding copper thin film (with a gage section measuring 600 ^m in length and 100 ^m in width, and preferred thickness of 0.3, 0.5, 0.9 ^m) to be tested for its mechanical properties, held at one end by the displacement sensor beam to mount a Polytech PI high resolution capacitor sensor, which allows the measurement of in-plane deflections. Same thing as other small scale tension tests, the gripping of the specimen is due to the adhesion between the frame substrate and the specimen materials. This eliminates the necessity of an extra gripping mechanism for the specimen. The sample was placed flat, with copper thin film up (frame down), on horizontal mounting blocks, or mounts. Grips rise vertically out of the top faces of the mounts, fitting through the sample that extends through the frame at either end of the tensile sample. The mounts are held in position relative to one another and the two mounts could move independently of one another. The other end of the specimen is held by a set of grip which have the ability to reduce the misalign errors.

The schematic of the test chip

Fig. 1- The schematic of the test chip

The details on fabrication of the sample, the loading fixture, as well as an analysis on the issue of alignment were provided in [2~5]. Figure 2 is a modified version of the fabrication sequence and is more robust in terms of process yields. The processing substrates are using 4" (100) silicon wafers. A silicon nitride layer in 1^m thick served as a sacrifice layer coated on top of it. Copper films with thickness of hundreds of nanometer thick were then being sputtered. The copper films were then being patterned in microlithography lift-off process. Total of 32 test samples can be obtained from each wafer. The thickness of each tested thin film is measured by a profilometer after sample fabrication.

The sample fabrication sequence

Fig. 2 -The sample fabrication sequence

Overview of the system

The stage apparatus is custom design equipped with environmental cells and controlled by PC through National Instrument LabVIEW program. This assembly consists of a micromechanical testing system with height-adjustable grips, built-in piezoactuator with position sensor, load cell and temperature sensor. The control electronics include a closed-loop piezoelectric controller, amplifier and waveform generator. Monitored signals are conditioned and then fed into an A/D board which is located in a PC. Data acquisition is performed with LabVIEW software. The system is covered in a well insulated, temperature controlled box and is supported on a vibration isolation table. It consists of two aluminum blocks rigidly bolted to the vibration isolation table. The left block mounts a displacement controlled piezoelectric actuator. The right aluminum block mounts with height adjustable pin via an x-y-z stage to assure well alignment of the sample and also supports the capacitor load cell. The control electronics include a closed-loop piezoelectric controller with 0.1 um resolution and a waveform generator. The maximum displacement provided by the piezoelectric actuator has travel range over 50 um.

Loading was applied through the retraction of piezoelectric actuator to pull the test chip through the pinhole. Loads were measured by a capacitor load cell with a resolution of less than 0.1 mN. The displacement is transmitted to the displacement sensor beam and the strain can be calculated from the difference of displacement in load sensor readout.

Tensile force is applied on the specimen by adding a displacement on one end of the chip while the other end is held fixed. The displacement is transmitted to the displacement sensor beam by the specimen and causes a deflection on the displacement sensor beam. The force F on the film specimen is defined as F = kd, where k is the spring constant of the combination of force sensor beam and linked capacitor senor, and d is the beam deflection, it can also be measured from the capacitive displacement senor. Figure 3 shows the calibration figures of displacement versus voltage measurement of capacitive senor and thus the spring constant of the load versus senor measurement can be calibrated. For the stress and strain calculations of the freestanding thin film, the dimension of each test sample is individually measured. The thickness of each tested thin film is also measured by a profilometer. Strain is determined from that measure the distance between the left marker A and the right marker B with a resolution of ±10.0 nm.

the calibration figures of displacement versus voltage measurement of capacitive senor

Fig. 3- the calibration figures of displacement versus voltage measurement of capacitive senor

Simultaneously, external microscope with CCD camera is equipped to acquire in situ image for study during testing. Tension-tension type fatigue tests were utilized in this study due to the inability of thin film tensile structures to support compressive loads. Initial experimental parameters were determined based on results from monotonic tensile testing. Two parameters, displacement amplitude (A) and mean displacement (d) were varied in a systematic manner. Figure 4 and 5 schematically shows how the triangular waveform sent to the piezoelectric actuator was varied in the experiments. Figure 4 and Figure 5 show experiments where constant displacement amplitude, Ao was maintained while the mean displacement was varied from do to d1. One experimental constraint to note is that in order to maintain either tension-tension or tension-zero experiments, the displacement amplitude may not be more than two times the mean displacement.

schematic of fatigue experiments I

Fig. 4 schematic of fatigue experiments I

 schematic of fatigue experiments II

Fig 5 schematic of fatigue experiments II

RESULTS

Tension-tension fatigue testing results

The prior studies for fine grain size metal film fatigue [10] suggested that a free-standing tensile-type test, among other benefits, offers results which may be directly interpreted and affords ease of fatigue analysis. However, both qualities are important at a size scale where neither material properties nor testing methods are well established. Tension-tension type cyclic tests were utilized in this study due to the inability of thin film tensile structures to support compressive loads. It is clear that fatigue lifetime is dependent on both fluctuated load amplitude and overall mean stress. A detail study on mean stress versus loading cycles, fatigue cycles versus sample elongation of 500 nm and 900 nm thick copper thin film were perform to investigate their tensile fatigue behavior.

Initial experimental parameters were determined based on results from monotonic tensile testing to determine the fraction of its maximum stress. The mean stress (D) was varied with constant load amplitude (A) in a systematic test. In each tests, total of 15 cyclic tests set were performed from the mean stress of 20 percent maximum stress to 100 percent maximum stress with every 30 MPa extensions. Samples were run at a mean displacement corresponding to each load, including displacement amplitude corresponding to ± 20 MPa experiment, the maximum amplitude for this particular mean displacement (A/2 ^d). A summary of the results on a loading cycles vs stress ratio curve of 500 nm and 900 nm copper films can be seen in Figure 6, 7.

loading cycles vs stress ratio curve of 500 nm copper films

Fig. 6 – loading cycles vs stress ratio curve of 500 nm copper films

loading cycles vs stress ratio curve of 900 nm copper films

Fig. 7 – loading cycles vs stress ratio curve of 900 nm copper films

In bulk copper materials, under constant stress amplitude, 1×102 ~103 cycles without failure is often defined as a fatigue limit. In this study, the fatigue tests on copper thin films with thickness of 500 nm and 900 nm, neither of the samples failed within 1×103 cycles. Although testing conditions are different, we consider both of our tested 500 nm and 900 nm samples to be past the fatigue limit. In particular, when the experimental mean stress was increased to 80 percent of maximum yielding stress, the fatigue life time can still reach over 1×104 cycles as shown in Figure 6, 7. This is contrary to the behavior of bulk copper materials where the low fatigue cycle leads to sample failure in such loading condition. Moreover, it was observed that the fatigue to failure cycles can exceed 1×105 cycles without fracture in the 20 percent maximum stress amplitude on both thickness of the tests samples. The results indicate a trend of increasing cycles to failure with decreasing thickness of copper down to submicrometer scale. However, the trend on fatigue cycles to failure between 500 nm and 900 nm was not clearly reflecting to the reduction of its thickness since the grain sizes were similar in both films.

Cyclic creep testing results

Figure 8, Figure 9 summarized the sample elongation rate versus fatigue life for fixed stress amplitudes of 20MPa and at various mean load (stress) tests as indicated of 500nm and 900 nm copper films respectively. These sample elongation values were obtained from the value of DC offset of piezo feedback as described.

In Figure 8, for the mean load value of 26.75 mN (estimated to be 535 MPa), the cyclic elongation of sample reaches 7.8um. In the contrast, for the mean load value of 35 mN (estimated to be 700 MPa), the cyclic elongation of sample is less than half micron. As well as Figure 9, for the mean load value of 12.6 mN (estimated to be 140 MPa), the cyclic elongation of sample reaches 6.8um. In the contrast, for the mean load value of 53.55 mN (estimated to be 595 MPa), the cyclic elongation of sample is less than half micron. For the data shown in Figures 8 and 9, cyclic plastic flow decreases with mean stress increases, indicating the sensitivity of cyclic creep to imposed mean stress and a trend in decreasing plasticity with increasing mean load is well described.

the sample elongation rate versus fatigue life for fixed stress amplitudes at various mean load (stress) tests of 500nm copper films

Fig. 8 – the sample elongation rate versus fatigue life for fixed stress amplitudes at various mean load (stress) tests of 500nm copper films

the sample elongation rate versus fatigue life for fixed stress amplitudes at various mean load (stress) tests of 900nm copper films

Fig. 9 – the sample elongation rate versus fatigue life for fixed stress amplitudes at various mean load (stress) tests of 900nm copper films

CONCLUSIONS

A novel experimental sample design and apparatus for uniaxially tensile testing microtensile of thin films under monotonic loading/unloading and tension-tension fatigue conditions has been designed and fabricated. Furthermore, it is capable of testing over a large range of displacement rates up to 20 m/s as well as a range of frequencies up to 25 Hz. The experiments are carried out with feedback to give load control on sputter deposited 500 and 900 nm Cu thin films. Loading cycles to failure reached as high as 10A5 at low mean load with a trend of decreasing cycles to failure with increasing mean load as anticipated. The cyclic creep results provided clear evidence for a creep rate dependent and change in failure mechanism from crack formation to extended plasticity as the mean load is decreased.

Next post:

Previous post: