100-Billion-Atom Molecular Dynamics Simulation of Acoustic Cavitation in a Simple Liquid

TL;DR

Ultrasonic cavitation involves complex, multi-bubble dynamics that are difficult to capture experimentally or with conventional simulations. This study leverages a 100-billion-atom MD simulation on the Fugaku supercomputer to directly observe nucleation and growth of bubble clouds localized near the ultrasonic horn under periodic driving. Bubbles form near the horn into a localized cluster that undergoes periodic fragmentation and recombination synchronized with the horn, with pressure and temperature spikes accompanying fragmentation and subharmonic dynamics of amplitudes; nonetheless, bulk acoustic properties show little change. The results provide molecular-scale insight into cavitation mechanisms and subharmonic generation relevant to sonochemistry and offer guidance for designing efficient ultrasonic reactors and biomedical applications.

Abstract

A large-scale molecular dynamics (MD) simulation of acoustic cavitation in a simple liquid was performed using the supercomputer Fugaku. The system, consisting of approximately 100 billion atoms, was subjected to ultrasonic irradiation. Direct observation of multi-bubble dynamics has been challenging in both experimental measurements and conventional numerical fluid mechanics simulations. Moreover, previous MD simulations involving only hundreds of millions of atoms were unable to generate multiple bubbles within a system. Our results reveal that cavitation bubbles nucleate and grow near the ultrasonic horn, forming a large bubble cluster that periodically splits into multiple small clusters and subsequently merges again. This cycle is synchronized with the oscillation period of the horn. Pressure and temperature inside the bubbles exhibit sharp increases during cluster fragmentation, and their oscillation amplitudes vary on a timescale longer than the driving period of the horn, indicating the presence of subharmonic behavior consistent with experimental observations. Despite bubble formation, the effect on the acoustic properties of the sound wave was almost negligible, indicating that cavitation near the horn surface has limited influence on bulk acoustic properties. These findings provide new insights into the molecular-scale mechanisms of cavitation and offer guidance for optimizing ultrasonic systems in chemical and biomedical applications. Future work will focus on quantifying long-period oscillations, analyzing attenuation effects, and extending simulations to complex fluids.

Figures (9)

• Figure 1: Snapshots of the density field at (a) time $t=4t_{\rm p}$, (b) $t=6t_{\rm p}$, (c) $t=8t_{\rm p}$, and (d) $t=10t_{\rm p}$. Time is normalized by $t_{\rm p}$, the driving period of the ultrasonic horn. Only low-density regions are shown to highlight bubble nucleation and growth near the ultrasonic horn. The insets of each panel depict the enlarged view near the horn.
• Figure 2: Void-fraction distributions at (a) $t=4t_{\rm p}$, (b) $t=6t_{\rm p}$, (c) $t=8t_{\rm p}$, and (d) $t=10t_{\rm p}$. Time is normalized by $t_{\rm p}$. The purple lines denote the average void fraction and the standard deviation of subcells at each $x$, while the green lines denote the maximum void fraction. Because the initial state was set at the gas–liquid coexistence point, a uniform void field ($\sim0.003$) exists at the initial time.
• Figure 3: Cross-sectional snapshots of the density field on the $yz$-plane at $x=15$ for (a) $t=4t_{\rm p}$, (b) $t=6t_{\rm p}$, (c) $t=8t_{\rm p}$, and (d) $t=10t_{\rm p}$. Time is normalized by $t_{\rm p}$. Numerous small bubbles appear and expand over time.
• Figure 4: Gas clusters classified by ID at (a) $t=4t_{\rm p}$, (b) $t=6t_{\rm p}$, (c) $t=8t_{\rm p}$, and (d) $t=10t_{\rm p}$. Time is normalized by $t_{\rm p}$ Clusters are labeled in descending order of size, and the largest cluster is shown in green. IDs of $10$ or higher are shown in red, and single-cell clusters are shown in yellow.
• Figure 5: Time evolution of the total gas-phase volume and the volume of the largest gas cluster. Time is normalized by $t_{\rm p}$. Subcells with $\rho <0.32$ are identified as gas phase.