Deformation and Failure Mechanisms of Bulk Metallic Glasses: Gauss Centre for Supercomputing e.V.

Project Overview

Bulk metallic glasses (BMGs) are known to have remarkable mechanical properties, such as high tensile strength, elasticity, and yield strength, which surpass those of many crystalline and polycrystalline metals. These properties make BMGs highly promising candidates for applications requiring materials that can withstand high and complex mechanical stress. However, BMGs have drawbacks; they show strain softening, resulting in localized deformation in the form of shear transformation zones that later lead to the formation of shear bands. This strain-softening characteristic limits their broader application potential, as it can lead to surface defects and, ultimately, fracture.

The primary aim of this project was to understand the relationship between the thermal history of BMGs and their subsequent mechanical properties, particularly how they deform under nanoindentation. The investigation focused on understanding the atomic-scale mechanisms that govern the link between the liquid fragility and the formation of shear bands at the glassy state. We also addressed the question how the performance of BMGs in friction applications can be improved.

The Challenge We Faced

When it comes to these bulk metallic glasses, the story is quite complex. Their mechanical behavior depends greatly on how quickly they were cooled from a hot, molten state into a glass, which to physicists means into a disordered solid. If they are cooled quickly, they are more ductile, but if cooled slowly, they become stronger yet more prone to cracking. This is because, during cooling using a continuously varying cooling rate, any structural changes at the fragile to strong transition temperature are smeared out. However, the exact nature of this transition from fragile to strong behavior and how it affects the formation of shear bands and other deformation mechanisms is not fully understood.

Through this project, we aimed to explore these questions by simulating the nanoindentation process, which involves pressing a rigid indenter into the surface of BMG samples to study how they respond to stress.

Why Supercomputing Was Required

You might wonder why we needed such a powerful computer for this. The answer is that when you are trying to simulate how millions of atoms move and interact in such a complex process, regular computers just do not cut it. For this project, we had to simulate up to 23 million atoms to see how these materials responded to being pressed down. That is an enormous amount of data and calculations! This is where the JUWELS supercomputer came in, helping us handle this enormous task.

Scientific Work Accomplished and Results Obtained

The main big task was to simulate the process of pressing into the BMG samples using nanoindentation. We focused on a particular type of BMG made from Zr_0.6Cu_0.3Al_0.1, which is close to composition in commercial use. We used an indenter with a radius of 100 nanometers to press into these samples and looked at how they responded when cooled from different temperatures.

Key Findings:

Force and Displacement: We observed that the way these glasses react to pressure changes depending on how they were cooled. At first, when the indentation depth was small (less than 3 nm), there wasn’t much difference. But as the depth increased, differences emerged. Stronger glasses resisted more force compared to the fragile ones, which means they were tougher but also more brittle.

Figure 1: Force displacement curves of loading for the fragile glasses and the strong glass at T= 300 K. The dashed line reflects an indentation with a contact modulus of E* = 57.9 GPa. Deviations of the data from linearity at very small d are due to an effectively finite-range repulsion between indenter and samples.

How Shear Bands Form: When we pressed into the glass, we noticed that the stronger glasses formed very distinct radial bands of strain, about 8 nm wide. These bands are where the material starts to deform and eventually fails. In contrast, the fragile glasses had more spread-out and symmetrical bands, showing that they deformed in a ductile manner.

*Figure 2: Local shear strain maps for the strong and fragile glasses loaded to an indentation depth of (a) d = 10 nm, (b) d = 12 nm, and (c) d = 15 nm.*

Using AI to Understand Patterns: To get a better look at these shear bands, we used artificial intelligence models to analyze them. This helped us see that the transition from fragile to strong states made a big difference in how these bands formed.

*Figure 3: Cosine similarity matrix using ViTg14 for the snapshots at different depths. The first number in each row and column gives Tq in Kelvin, while the second gives d in nm.*

Not as Random as It Seems: One of the surprising findings was that the way these shear bands formed wasn’t as random as we thought. Despite their chaotic appearance, they followed a pattern, which means the way the glass was before it cooled had a big impact on where and how these bands showed up.

Figure 4: Von Mises strain of (a) the default setup for Tq = 812 K and a depth of d = 15 nm and (b) and (c) for a slightly larger sample at a depth of d = 15.2 nm. The starting configurations in (b) and (c) differed, i.e., the original glass sample in (c) was rotated by 90° before duplication so that the original yz surface rather than the xy surface was indented.

Finally, the question of why BMGs appear to have extremely small internal friction but unfavorable tribological properties like large wear and large sliding friction was also addressed. The small internal friction makes small BMG spheres bounce off surfaces like rubber balls. This performance is commonly attributed to the lack of dislocations, whose impact-induced motion dissipates much energy. But then, why is their sliding friction high? To address this question, we scratched a BMG surface with an indenter consisting of a perfectly smooth, repulsive indenter, also known as a mathematical wall. Using such a smooth indenter suppresses all explicit (static) friction at the surface since the indenter is translationally invariant. The friction-simulation setup is shown in Fig. 5(a). It is equivalent to a pin-on-disk tribometer due to the use of periodic boundary conditions. Fig. 5(a) shows that our BMG ‘runs in’ as regular metal surfaces do; that is, with each pass of the surface, the friction coefficient m decreases with the first few strides but later levels off, in our case near m = 0.1. While m = 0.1 is a relatively small friction coefficient, it must be remembered that m it would instantly double if the counter face were also composed of a BMG and supposedly more than triple if the scratching tip were atomically rather than perfectly smooth. Increasing the load from 300 nN to 600 nN makes the friction more than triple, i.e., m increases by 65%. The results on BMG friction can be rationalized significantly by the results on plasticity presented earlier. First, BMG work softens. This is already problematic before effects like unstable friction, increased material transfer, or wear set in: Massive, energy-consuming plasticity occurs each time, even when the tip scratches over an existing ‘wear track’, in contrast to what would happen on a work-hardening surface. Second, due to plasticity happening in localized zones, the scale dependence of plasticity in BMGs is relatively minor.

Figure 5: (a) Geometry of the set-up of a rigid tip (purple) scratching over a surface. Bright colors in the BMG substrate indicate a large plastic shear strain. (b) Side-view of the sliding process during the second pass at a normal load of L = 300 nN. (c) Friction coefficient as a function of sliding distance at a normal load of 300 nN (three traces) and 600 nN (one trace).

Realization of the Project

We used a LAMMPS code to run all our simulations on the JUWELS supercomputer. It took about a year’s worth of computing time to complete everything we needed, especially the nanoindentation simulations. Along the way, we had to adjust our methods, like increasing the system size, to make sure we were accurately capturing the formation of these shear bands.

Publications and Presentations

Achraf Atila, Sergey V. Sukhomlinov, Marc J. Honecker, Martin H. Müser, “Brittleness of metallic glasses dictated by their state at the fragile-to-strong transition temperature” (under review and arxiv.org/abs/2408.00536)
Achraf Atila, USTV-DGG joint meeting, Orleans (France) 2023
Martin H. Müser, APS March Meeting, Las Vegas (USA) 2023.
Martin H. Müser, 25th anniversary of NIC, Köln (Germany) 2023
Martin H. Müser, Tribochemistry, Beppu (Japan) 2023

Conclusions

Ultimately, we learned how bulk metallic glasses behave under nanoindentation has much to do with their state at the fragile to strong transition temperature. The stronger the liquids are, the more brittle they become. The more fragile the liquids are, the more ductile they are, but this comes at the cost of being weaker. Using large-scale atomistic simulations, we could see how these changes happen on an atomic level, and that’s interesting to material scientists and engineers as well as physicists working on glasses.They also revealed that bulk metallic glasses are ideal candidates for low-friction applications if coated with nano-crystalline metals.

MATERIALS SCIENCE AND CHEMISTRY