A nano-tensile tester for creep studies (MEMS and Nanotechnology)


Free-standing metallic thin films are increasingly used as structural components in MEMS. In commercial devices, long-term reliability is essential, which requires determining time-dependent mechanical properties of these films. The uniaxial tensile test is a preferred method due to uncomplicated determination of the stress and strain state. However, at the MEMS-scale this method is not straightforward: specimen handling and loading, force and deformation measurement need careful consideration. Here we discuss the challenges of the application and measurement of nano-Newton forces, nanometer deformations and micro-radians rotation alignment ensuring negligible bending in on-chip tensile test structures during long periods. We then present a novel tensile-testing instrument with in-situ capabilities in SEM and Optical Profilometry. The design solutions to measure these small forces and deformations whilst ensuring a uniaxial stress state will be presented.


Mechanical testing for material behavior characterization has brought much understanding into the mechanics of materials at the macro scale. Nowadays, however, miniature devices with dimensions at the sub-micrometer scale, such as MEMS, are processed routinely, which has revealed unexpectedly new mechanical micro-mechanisms. This has spurred research into new mechanical characterization techniques to understand the physical fundamentals at the (sub)-micron scale, e.g. nano-indentation [1], FIB-enabled in-situ micro-tensile testing [2], fully integrated and dedicated tensile test MEMS [3]. One important outcome of this research is that testing at the nano-scale is far from trivial [4;5]! To address this issue, a novel nano-tensile methodology is presented here for which all fundamental aspect of tensile testing have been reconsidered in its design.

A suitable testing methodology faces a number of challenges. First of all, such a methodology needs to be sensitive enough to measure the nano-Newton forces and nanometer deformations involved at this scale. Well-defined loading conditions are preferred to facilitate interpretation of the deformation state, thus favoring the uniaxial tensile test. Boundary conditions should also be carefully controlled to minimize undesired influences, such as surface roughness or friction effects, while challenges of specimen handling, loading and alignment need to be addressed as well. Furthermore, easy specimen variation is required to enable systematic studies of the influences of, e.g., mechanical size-effects. Finally, in-situ SEM testing capability is necessary to unravel the physical origin underlying (the often complex) microscopic deformation mechanics [6].

Design of methodology

The design of the methodology has the uniaxial tensile specimen at its heart. The specimen fabrication and variation determine the requirements for the loading method, force and deformation measurement. After establishing these requirements, suitable solutions to obtain them are discussed followed by their respective implementations in the design of the uniaxial tensile tester.

Specimen fabrication

A common method [7-9] to create tensile specimens is the micro-fabrication of dog bone shaped specimens on a substrate, with one end and the gauge section free-standing e.g., by under etching. Considering specimen handling and fabrication, it is highly preferred to test on-chip structures instead of separate ^m-sized structures. Moreover, applying the same micro-fabrication procedure as done for the actual device guarantees the relevance of obtained results. Therefore, dog bone shaped tensile specimens are designed and fabricated on silicon chips, see Figure 1. The cross section dimensions of the gauge section are varied to probe size-effects from sub-micron into the micro range, because MEMS devices are designed this range. The specimen length is made as long as possible to facilitate the elongation measurement. Based on these dimensions and the desire to perform creep measurements at tensile stresses of 1-100 MPa, the force range is determined. Finally, a chip of 10×10 mm2 is filled with ~60 specimens. In short, this approach takes advantage of the precision and ease of reproduction of microfabrication, and the ease of geometrical variation within a chip design.

Chip layout and tensile specimen design

Figure 1 Chip layout and tensile specimen design

Load application

The loading of the specimens has to deal with applying desired forces to the specimen and ensuring a homogeneous uniaxial tensile stress is attained. The application of the forces is simplified to mounting a chip and gripping the free end of the tensile specimen. Several approaches to gripping have been reported: electro-static clamping, application of adhesives and mechanically locking. These methods have their pros and cons, but a choice is only possible after considering the effect of incorrect gripping, because this can lead to misalignment resulting in undesired bending stresses in the specimen. Some forms of misalignment can occur (see Figure 2): i) force and specimen’s longitudinal axis are parallel, but not co-linear, ii) force and longitudinal axis are at an angle. Based on a straight forward analysis of statics and elastic beam theory, the initial ratio of maximum bending stress to desired tensile stress can be estimated. The bending stress from non-co-linearity can be assumed negligible, if the point of force application is in-line with the specimen’s longitudinal axis. Furthermore, actuation of translations should be feasible at this scale with 5-10 nm precision, which is less than 1% of the specimen width and thickness. The bending stresses resulting from rotational misalignment can be minimized by reducing the ration l/t or by minimizing the misalignment angles. As it is desirable to increase l (for the elongation measurement), the angles need to be minimized. Allowing for cbend/otensiie < 5%, then already 9<10-4 rad for l/t=500 ^m /0.5 ^m! Therefore a precise mechanism is required to rotate the specimen and/or the gripper.

To minimize the unwanted bending stresses, gripping is done mechanically by creating a hole in the pad at the free end of the specimen and a pin at the end of a stiff beam, the so-called gripper, see Figure 2. By using bulk silicon micro-machining of an SOI-wafer (200 ^m handling layer, 50 ^m device layer), the gripper can effectively be fabricated with sub-micron precision from the ^m-scale at the specimen end to the mm-scale to facilitate instrument integration. The bending is minimized in two ways. First, the hole in the pad has a sharp feature that aligns the contact point of gripper and pad to the specimen’s axis.

Second, the longitudinal axes of specimen and gripper can be translated in three directions and rotated about two axes with respect to each other with high precision to further minimize misalignment. Coarse translations (mm-range, ^m resolution) are applied through manual thumbscrews, whilst fine translations are applied through a commercial 3-axis piezo stage (MadCityLabs, Nano-T225M: xy-parallel kinematics, z-axis separate), having in-plane and out-of-plane resolution of <1 over a range of respectively 200 ^m x 200 ^m and 50 ^m and pitch, yaw, roll of ~10-6 rad. The rotations are carefully set through precision mechanisms based on elastic hinges with angular resolution of <10-4 rad.

Schematic illustration of (a) ideal uniaxial tensile test for the tensile specimen fixed at one end, (b) unwanted bending moment due to non-co-linearity of specimen's axis and force and (c) due to angular misalignment between specimen's axis and force

Figure 2 Schematic illustration of (a) ideal uniaxial tensile test for the tensile specimen fixed at one end, (b) unwanted bending moment due to non-co-linearity of specimen’s axis and force and (c) due to angular misalignment between specimen’s axis and force

The manipulation of the rotations needs to be measured to ascertain the bending stress contribution. Whilst in-plane rotations can be measured straightforwardly by top view imaging of the specimen with an (electron-) optical microscope and simple image processing, the out-of-plane rotation is not as straightforward. As an optical profilometer will only allow for measuring the misalignment to <10-3 rad, an electrical mechanism is included. Two electrical contact pads are placed on the substrate adjacent and parallel to the specimen’s axis, spaced at a distance S=1 mm. Then a second pin is placed on the gripper at distance S from the loading pin. By grounding the pins and measuring the electrical currents through each contact pad, contact between the gripper and substrate can be monitored when the gripper approaches the substrate. In the case of no misalignment both contacts will be made simultaneously. If there is misalignment, this can be detected with 5zpiezo / S < 5-10-5 rad. With this method of gripping and aligning of load to the specimen, it is estimated the bending stress contribution will be <10% of the desired tensile stress.

 Load transfer using a 'gripper': pin-in-hole contact and alignment of axes minimizes moments and bending stresses

Figure 3 Load transfer using a ‘gripper’: pin-in-hole contact and alignment of axes minimizes moments and bending stresses

Load measurement

The previous solutions yield a method of precisely loading the specimen. As indicated in Figure 1, the range of loads spans 5 decades, whilst 2 decades are additionally desired for nN-precision! Nano-indenters are currently the only instruments with such capabilities. Practically speaking it is not necessary to have nN-resolution for mN forces. Therefore, it is chosen to create three exchangeable load cells, each spanning part of the range. The load cells are simple parallel leaf spring mechanisms that are fabricated from a Ti-Al-V alloy through electric-discharge machining. The gripper is mounted at the end of the parallel leaf spring. The stiffness of these springs combined with the deflection measurement determines the measurable force range and resolution (see Table 1). By using a highly precise capacitive sensor (Lion Precision C5-D probe + CPL190 Driver), the leaf spring deflection u is measured. Calibration of the load cell is then achieved by tilting the load cell with a known angle and measuring the deflection caused by a component of the gravitational force exerted on the gripper of known mass. Finally, using accurate electronic readout ensures the long term stability of the load measurement.

Table 1 Desired load cell characteristics





k [N/m]

Capacitive displacement sensor

250 ^m

5 nm @100 Hz

0.1% of full range


Force range 1:

0.01 pN…10 ^N

<0.01 ^N

<0.05 ^N


Force range 2:

1 mN…1 mN

<0.1 ^N

0.5 ^N


Force range 3:

100 pN…100 mN

<10 ^N

50 ^N


Deformation measurement

The measurement of deformation is feasible with microscopical tools like SEM or even optical microscopy, if dimensions across which deformations are to be measured are large enough. The foremost component of deformation measurement is the uniaxial strain. By placing markers along the length of the gauge, small enough to have negligible influence on the stress state, and on the substrate, the displacement of these markers can be measured yielding the uniaxial strain. With digital image correlation (DIC) this can be resolved to ~0.1 pixel for SEM images and ~0.01 pixel for light microscopy images [10] which yields strain resolutions of ~10-4 respectively ~10-5 for an image of 1000 pixel length. Furthermore, by employing SEM to obtain hi-resolution images at the micro-scale, not only in-situ qualitative observations of the deformation mechanisms are obtained, but also quantitative 2D deformation fields in combination with DIC.

Positioning and actuation

The last aspects of the design of the methodology are the chip mount, positioning and load actuation. The chip is mounted on a small platform through mechanically clamping. The platform is also equipped with a resistive heater and thermometer to heat the chip up to 150 °C. The design of the nano-tensile stage then integrates the chip platform, the load cell with gripper, and positioning and alignment mechanisms, see Figure 4. Positioning of the gripper and load cell is achieved by a coarse manual xyz-stage, and a fine electronic xyz-piezo stage. The x-actuator of the piezo stage also serves as the load actuator (200 ^m stroke, <10 nm resolution). Alignment to minimize bending stresses as previously specified is done for 4 rotations: 1) 0y and 2) 0z of the specimen, 3) 0z between x-piezo actuator and load cell’s direction of deflection and 4) 0y of the load cell and xyz-stage to set the load cell horizontal with respect to gravity. Based on these considerations, the alignment and test procedure is as follows:

1. Under an optical profilometer the load cell is leveled so it is in the center of its range.

2. Then the gripper is translated to the specimen with the manual manipulators followed by in-plane alignment by optically measuring the angle between the long side of the gripper and the beam and applying simple image processing.

3. The out-of-plane alignment can then be done by first using optical profiling to achieve alignment within 10-3 rad and then the electrical contacting mechanism to align within 10-4 rad.

4. After translating the gripper away from the specimen only using the piezo stage, the setup can be transferred to the SEM, where if necessary the load cell leveling is done again (required if SEM and profilometer are not exactly parallel).

5. Then in the SEM, the misalignment is checked before the gripper is hooked into the specimen with the piezo stage.

6. Finally the creep test can be performed by loading with the piezo stage, measuring the force with the load cell and recording the deformation through SEM imaging.

This design succeeds in creating a compact setup that fits in a scanning electron microscope. By translating the gripper for specimen loading, the imaging of the specimen’s deformation is greatly facilitated. Finally, the use of accurate electronics and a feedback loop to conduct force controlled creep experiments combined with the mechanical precision design will enable performing stable and precise long term in-situ measurements.

 Schematic overview of the realized tensile tester, indicating the various actuated degrees of freedom, mounting of load cell and placement of chip with tensile specimens

Figure 4 Schematic overview of the realized tensile tester, indicating the various actuated degrees of freedom, mounting of load cell and placement of chip with tensile specimens


In the authors’ opinion, the here-presented nano-tensile methodology is the first technique that meets all of the requirements simultaneously: force resolution of 10 nN, strain resolution of 10-5, minimization of bending stresses to 10% of the desired uniaxial stress, easy specimen variation and handling and a compact and stable setup capable of in-situ SEM creep measurements. The strength of the methodology will be demonstrated through highly precise measurements of uniaxial stress-strain curves of on-chip ^m-sized free-standing Al-(1wt%)Cu beams (used in RF-MEMS applications).

Next post:

Previous post: