Efficient perturbations for basin hopping in amorphous glasses

Matter that lacks a rigid, repeating pattern is everywhere, from the glass in a window to the volcanic rock beneath our feet. Unlike crystals, where atoms line up in perfect, predictable rows, these amorphous materials have atoms arranged in a chaotic jumble. This disorder makes them difficult to study because scientists cannot simply look at a pattern and deduce the structure; the information is hidden in the way atoms scatter light or electrons. To understand these materials, researchers use computer simulations to build models that match experimental data. However, finding the right arrangement of atoms is like trying to find the lowest point in a vast, foggy mountain range. The computer starts at a high spot and tries to move the atoms to lower the energy, but it often gets stuck in small dips, or local valleys, that look like the bottom but are not. The challenge is to find a way to jump over the ridges that separate these valleys to discover the true, most stable arrangement of the material.

A team of researchers at Cuthbertson High School and the University of North Carolina at Chapel Hill tackled this problem by testing different ways to shake up the atomic structure of amorphous aluminum oxide, a material used in everything from ceramics to electronics. Instead of making tiny, cautious steps that often leave the computer stuck in the same small valley, they tried larger, more disruptive moves. They designed four specific types of changes to the atomic network. One method involved swapping the positions of atoms entirely. Another tweaked the angles between atoms slightly. A third tried to change how many atoms shared a connection with their neighbors. The fourth, and most successful, involved moving a single oxygen atom from one aluminum atom to another, effectively changing the number of connections each aluminum atom had. After making each of these bold changes, the computer was allowed to relax the structure, letting the atoms settle into a new, lower-energy position naturally.

The researchers found that the size and type of the move mattered immensely. The method that moved an oxygen atom to change the coordination of two aluminum atoms proved to be the most efficient way to drop the energy of the system quickly. This single move caused a large shift in the network's shape, allowing the simulation to leap over high energy barriers that would have trapped a more cautious approach. In their tests, this method reduced the energy of the structure significantly in just a few steps, whereas traditional, small steps would have required millions of attempts to achieve the same result. However, the researchers also discovered that being too aggressive has a limit. While the large jumps got the system out of bad spots quickly, they sometimes ran out of new places to go after only a handful of moves.

In contrast, a method that made very small adjustments to the angles between atoms was slower to start but proved more thorough in the long run. Because these moves were gentle, the computer could accept many more of them, allowing the simulation to explore the landscape more deeply. This approach eventually found the lowest energy state observed in the entire study, reaching a value of -768.11 eV, but it took many more steps to get there. The study suggests that the best strategy is not to rely on just one type of move, but to combine them. A large, non-local jump could be used first to escape a deep trap and reach a lower region of the landscape, followed by smaller, local adjustments to fine-tune the structure and find the absolute lowest point.

The work also revealed a clear rule about how these materials stabilize. The simulations showed that the structure became more stable when the aluminum and oxygen atoms shared more connections with one another, creating a more tightly knit network. Moves that increased this sharing lowered the energy, while moves that broke these connections did little to improve the structure. This insight helps explain why certain atomic arrangements are preferred in nature. By understanding that specific types of atomic reorganization lead to greater stability, scientists can improve the computer programs used to model amorphous materials. This could lead to better designs for new glasses and ceramics, ensuring that the digital models used to predict their properties are as accurate as possible. The study confirms that the way we choose to move atoms in a computer simulation is just as important as the rules we use to judge the results, offering a new path to solving the puzzle of disordered matter.