What is the role of imec in semiconductor technology, and what are the unique assets it offers to partner companies?
Imec is recognized worldwide as a research and development (R&D) hub undertaking R&D work in collaboration with various companies. Part of this work is in the semiconductor industry, where we collaborate with our partners to make computer chips smaller and more efficient in terms of performance and energy use.
Our site at Leuven in Belgium boasts a range of unique assets, including 12,000 m2 of clean room space with all the equipment required to process wafers, analyze different processes, and explore the deposition of different films.
More than 5,500 skilled people from 95 different nationalities work at our site, so it is a global organization. We are already trusted partners of a range of companies and startup projects, and we work in close collaboration with the university here.
Our work at imec involves trying to bridge the gap between smart application technology and advanced semiconductor technology. As an organization, we have a good degree of knowledge in R&D and advanced semiconductor technology, and we use this to try to solve problems in different sectors.
We also try to drive innovation and develop new technologies while working towards a more sustainable society. We want to work on projects that improve health and well-being, are good for the climate, and drive innovation in different industries.
My department works with mechanical and thermal modeling and characterization. This work involves examining materials’ mechanical properties, such as performing package warpage tests, determining stress-strain distributions, investigating fracture mechanics, and characterizing a material’s properties.
We also examine materials’ thermal and electrical properties, for instance, by using stress sensors to investigate stress in ferroelectric memories. The connection between mechanical and thermal properties is also important to us; for example, we may try to discover materials’ coefficients of thermal expansion (CTEs) or the temperature-dependent properties of polymers.
We try to connect all this data, leveraging this in finite element simulations to see how these materials behave mechanically, electrically, thermally, and at different scales, from the nanometer scale to the centimeter scale.
We can use these simulations in combination with experimental approaches, for example, optimizing bonding parameters to ensure the cost-effective 3D stacking of dies to improve this process flow. We also perform package-level reliability testing to highlight problems and determine how to address them.
In cases where Microchips exhibit local heating, we develop cooling solutions through experiments and simulations to determine where the heat is located and how it can be minimized.
Image Credit: raigvi/Shutterstock.com
How does indentation-based fracture testing work, and why is it critical to evaluate thin film adhesion?
Our goal is to develop finite element simulations of real devices to optimize them and see how they perform using a range of experimental test methods. These methods include four-point bending or indentation-based fracture to test adhesion and nano-indentation or shear microprobing to assess elastic-plastic properties.
Once we have performed these tests, we perform post-test analysis and failure analysis and gather enough statistical information to enable simulations as part of an iterative improvement process.
The first part of this process is finite element modeling, which involves breaking something down into its composite elements and assigning properties to each element to simulate how its respective material will behave under specific test conditions.
This process requires a lot of information. For example, we need the geometry of each element, its dimensions, material properties like Young’s modulus and CTE, and nonlinear material properties like creep or plasticity. We also need loading conditions for mechanical simulations. and reliability models.
This process has many inputs, but the most important consideration is ensuring realistic input for these models so that we can gain a realistic estimation of the behavior of the structure or element being tested.
Nano-indentation is key to this testing, and at imec we use various equipment to perform this. We have two Bruker systems: the TI Premier and the TI950 TriboIndenter. We also use a special stage dual-axis goniometer, which allows us to refine the alignment of the probe and sample. A 300 mm magnetic chuck and a heating stage are also available.
Nano-indentation typically uses a diamond probe, though this can differ when working with some materials. The probe is pressed inside the material, with recordings taken of the imprints of the indents and the generation of a forced displacement curve.
The probe must be calibrated before use, meaning we must know the material’s contact depth vs. contact area. The right equations, such as the Oliver-Farr model for bulk materials or the IDF model for thin films, can provide this.
We also make use of the Bruker PI 89 PicoIndenter. This interesting tool features an indenter system inside an SEM, allowing us to visualize fractures and perform shear tests.
What are the main differences between four-point bending testing and wedge indentation testing for thin films, and when is each method preferable?
Four-point bending is a standard and popular adhesion testing method. This method involves preparing samples with silicon substrate before gluing the film stack to another silicon substrate film stack with epoxy. Beams are prepared with a notch in the center before applying four-point bending.
The forced displacement curve of these tests reveals a stable plateau for crack extension, which is then transferred to the critical energy release state.
The indentation-based testing of adhesion is complex, and there are many ways to determine the critical energy release rate. Another popular approach is wedge indentation. Wedge indentation typically involves a wedge of a certain angle and length, for example, a 90-degree angle, but this can also be wider.
The wedge is pressed inside the film, and the delamination is recorded. Equations that consider film thickness, material properties, and the ratio between the plastic indentation volume and the film volume above the facial crack allow us to estimate the interface’s adhesion energy.
This approach has some problems, such as tilting between the sample and the probe. Depending on how the wedge probe is aligned with the sample or how the sample is mounted on top of the stage, there may be an asymmetric opening that is not ideal for analysis.
Applying this approach to different film thicknesses, for instance, a low-k material like organic silicon glass, will result in different openings in line with different thicknesses. Performing the relevant equations may reveal good agreement between these thicknesses, although there might be a small variation between results due to process conditions.
Relevant process parameters include the porosity or additional treatments applied to the materials during preparation. For example, hydrofluoric acid (HF) treatment could be performed on a low-k film.
Looking at the resulting forced displacement curves and correlating these to treated and open areas can be used to investigate how adhesion energy is altered by the HF treatment. The ability to capture changes during processing is key to assessing process stability.
The radius of the probe is also a key process parameter. Probes may be subject to a rounding effect after fabrication, but equations to determine wedge indentation always assume a perfect probe. A more rounded probe tip with a larger tip radius will result in larger cracks, larger stress areas, and more buckling. In this case, the equation would need to be adjusted.
There are several notable differences between four-point bending and wedge indentation. For example, we prepared a low-k 200 nm thin film deposited on a metal barrier on top of a silicon substrate. Then, we compared the results from the four-point bending with those from the wedge indentation.
Some combinations of measurements showed very good agreement, while some saw a difference in results between wedge indentation and four-point bending. This is because different techniques use different opening modes, indentation-based testing, and four-point bending, which may transfer different amounts of energy during testing. Differences can occur, but this can depend on the specific stack of the material.
Three-sided pyramids can also be used for indentation-based testing, for example, a Berkovich indenter, which has a wider angle to fracture inside materials, or a cube-corner probe, which can be used for more acute indentation.
How do different porosity levels in materials impact the critical energy release rate during wedge indentation testing?
An interesting consideration when working with fracture mechanics and fractures in dielectric materials is that cube-corner indentation exhibits a jump in critical force at around 120 nanometers deep and around 100 μN of force. We observed this at room temperature in this case.
When we performed the same experiment after heating up the sample, we noticed that the critical force vs. temperature correlation shifted upwards. We thought this was because of thermal stress, so we did some calculations to determine that the low-k material’s CTE depends on the substrate’s CTE.
We looked at the properties of the low-k material, including Poisson’s ratio and Young’s modulus, and then two coefficients: A0, which shows how critical force changes with temperature, and A1, which shows the relationship between critical force and stress.
We measured the A0 for a low-k material and then used curvature measurements to assess how stress in that film changes with temperature. We correlated those values for different geometries of the probe to generalize this model. This allowed us to develop an equation that allowed us to easily determine the CTE.
We evaluated this using temperature changes, first cooling, then heating, or heating, then cooling.
Dielectric materials are porous, meaning they are very sensitive to humidity. If we heated and then cooled the materials, we saw that Young’s modulus and hardness were stable, even at low temperatures. We also observed potential moisture uptake that can affect hardness and modulus measurements.
In terms of critical force, heating and cooling led to a nice linear behavior, making it easy to acquire the CTE from a material.
How do different metal densities in backend-of-line structures influence crack propagation during shear testing?
Different commercial chips have different architectures and metal densities. Indentation measurements can test the integrity of these materials, and cross-sectional imaging is also useful in estimating the integrity of these areas. A bump shear test can also be performed on top of the chip to determine when a backend-of-line failure occurs.
It is important to know when backend-of-line fracture events happen and how many bump events prompt these to assess chip quality. After shearing off a bump or inducing some cracking, we noticed that the cracks were always located between the Z group. These top layers are larger and exhibit a higher metal density than the softer layers below.
This is because there is a mismatch in the total Young’s modulus of these layers compared to the bottom layers, and the crack is always pushed towards that interface.
When we tested this via shear testing, we found that in areas of low metal density, almost 80 % of the failures recorded were backend-of-line failures. In areas of high metal density, for example, this reduced to only 3 % fails, which means these structures are more robust.
This testing allows us to rank different materials and determine which area performed the best. There may also be different metal densities where there are connections between different levels (via).
We can also perform cube corner indentation testing on top of these structures, looking for radial cracks like we would for thin films. The dimensions of these structures are very small—just 8 or 9 nm. When we press with a large probe relative to the structure, we measure the average behavior of the complete stack rather than a local measurement on one line.
We performed this test on different technologies, including copper low-k, ruthenium low-k, or ruthenium with air gaps. We observed differences in fracture or critical forces when using cube corner indentation. We also did this for different densities, plotting the fracture force as a function of the metal density.
What are the benefits of using in-situ picoindentation techniques for visualizing cracks and fractures in backend-of-line structures?
In situ experiments are another useful tool for evaluating cracks and fractures in backend-of-line structures. To do this, we prepare different dedicated beams, either cantilevers or double-clamped beams, before creating a flip-top-down cross-section and an undercut and then cleaning.
This approach results in a mix of low and high metal densities. The dimensions of the beams are important, and these should be measured. We can rank and compare different materials when plotting beam stiffness vs. failure force and density.
For instance, the low metal density has a low failure force and a lower beam stiffness than the higher metal density. The important thing is being able to see the location of the crack. The drawback of this approach is the need to do many different tests, and it is hard to acquire enough data to gain useful statistics because preparation takes more time for these tests.
It is important to perform dedicated tests to fully benchmark the thermo-mechanical integrity of thin-film advanced interconnects. This includes fundamental research and tests on commercially available test chips, which are closer to reality regarding their behavior.
This can be done using indentation-based fracture and adhesion testing, shear microprobing, in situ picoindentation, and many other types of testing still in development. These experimental techniques should always be supported by simulations, which can be used to further optimize the process.
Watch the full webinar
About the Speaker
Kris Vanstreels received his MS degree in Physics from KU Leuven in 2001 and a PhD degree in Physics from UHasselt in 2007. From 2001 to 2007, he was working as a researcher at the Institute of Materials Research of UHasselt, where he was involved in the physical characterization of different materials systems and the development of high-resolution in-situ measurement techniques. In November 2007, he joined the Reliability and Modeling group of IMEC in Belgium as a researcher. His main research activities are focused on mechanical characterization and reliability of back-end-of-line stacks and 3D stacked-ICs.
This information has been sourced, reviewed and adapted from materials provided by Bruker Nano Surfaces and Metrology.
For more information on this source, please visit Bruker Nano Surfaces and Metrology.
Disclaimer: The views expressed here are those of the interviewee and do not necessarily represent the views of AZoM.com Limited (T/A) AZoNetwork, the owner and operator of this website. This disclaimer forms part of the Terms and Conditions of use of this website.