As progress is made in the performance of semiconductor devices, changes in materials and processing methods are regularly required. The trend towards miniaturizing these devices integrally involves components at or close to the nanoscale.
The mechanical properties of materials at this scale can be enormously different from bulk material, which bring forth new concerns during processing as well as in field use. A reliable testing method is crucial for assessing the structural stability of these devices.
This article looks at the use of a
Hysitron that is used to examine the failure mechanisms involved in FIB-milled back-end-of-line (BEOL) microbeam samples. ® PI 85L SEM PicoIndenter ®
Figure 1. Hysitron PI 85L SEM PicoIndenter.
BEOL Microbeam Sample Preparation
In the microelectronics sector, the properties of individual device components (such as thin films on Si wafers) are repeatedly studied to evaluate their suitability for application in devices. While this can be a valuable first step, it is an insufficient representation of the real behavior of those components once added into full structures. In-situ testing not only enables the evaluation of total interconnect stacks, the failure mechanisms can also be monitored directly with an electron microscope. This article looks at the BEOL structures containing seven metal layers that were studied in-situ.
Figure 2. Schematic of sample preparation using FIB-milling.
To test these samples in the SEM, three-point microbeam samples were prepared using a focused ion beam (FIB) system, as illustrated in Figure 2. The microbeams were milled from the cleaved edge of the BEOL sample using a three-step milling procedure to reduce ion implantation. The sample was first milled top-down, establishing the beam length and coarse width. The beam was then released from the substrate below by tilting the sample 90° and undercutting the sample until the material under the beam was removed. To finish, the sample was milled top-down a second time, defining the final width of the beam. Four beams with two different widths and two different lengths were prepared, as shown in Table 1. The specific geometry of these samples also lends itself specifically well to finite element modeling (FEM).
(µm) Loading Rate
(µN/s) Critical Load
(mN) Critical Displacement
Table 1. Summary of the dimensions, loading rates and critical loads and displacements for each beam. As expected, for beams of a fixed length, the critical load is linearly dependent on width. Similarly, for beams of a fixed width the critical load is inversely proportional to length.
The sample (containing all four beams) was mounted on an instrument-compatible microscopy stub using conductive adhesive, and mechanically secured in the
Hysitron PI 85L. In the SEM, the beams were analyzed in load-controlled mode using a wedge-shaped conductive-diamond tip. Applied loading rates spanned from 200 to 750 µN/s (see Table 1), and video was acquired during every test for synchronization to the recorded mechanical data.
Figure 3 shows the load-displacement curve and corresponding scanning electron micrographs for Beam 1. In the region of the curve marked a→b, the beam is bending. As a vital force is reached (b and c), this is followed by crack initiation and growth in an adhesive and cohesive manner toward the edge of the beam (d→e), and then unloading (e→f). This type of fracture pattern was noticed in beams 1, 2 and 3. However, for beam 4 (illustrated in Figure 4) a different failure mechanism was noticed—a single crack occurred at the bottom of the beam, opposite the point of loading.
Figure 3. Beam 1 before and after mechanical testing, and the corresponding load vs. displacement curve. Interfacial delamination is observed between the Cu layers and the brittle dielectric. Real-time video of the test is shown below.
Figure 4. Beam 4 before testing and at critical load. In this beam, a crack opened directly opposite the point of loading. Real-time video of the test is shown below.
Finite Element Modeling
The uniform geometries of the microbeam samples are especially well suited to complementary finite element modeling of the bending tests. A 2D FEM analysis of beam 4 was conducted, which undoubtedly supports fracture initiation from areas with highest values of maximum principal stress. In contrast, the fracture patterns in beams 1-3 were found to correlate best to von Mises stress distribution. This indicates two distinct and competing fracture mechanisms, one which is triggered only by standard stresses and the other which is triggered by shear stress. The fundamental mechanism is thought to be established by several different parameters, including the distribution of copper within the brittle dielectric (which dictates the ability to follow a ductile or brittle fracture pattern), the loading rate and the precise beam geometry.
Quantitative in-situ mechanical testing in the electron microscope allowed for direct observation of adhesive and cohesive crack propagation in FIB-milled beam samples of BEOL structures. The mechanisms accountable for failure in each beam were possibly determined by several factors, as supported by FEM analysis and by direct observation in the electron microscope.
This in-situ method can be used to describe the nature of the failure mechanisms within each device component, as well as for full structures; structures with an ever growing degree of internal complexity. Moreover, the combination of this experimental method and modeling makes it possible to identify the weak interfaces in a structure, outside what can be predicted by the measurement of individual layers. This offers vital information for device development and manufacturing that had formerly been difficult to obtain.
Data courtesy of K. Vanstreels, IMEC, Leuven, Belgium.
For more information see Vanstreels, et. al., Appl. Phys. Lett. 105, 213102 (2014).
This information has been sourced, reviewed and adapted from materials provided by Hysitron Incorporated.
For more information on this source, please visit