Modelling turbulent flow of superfluid $^4$He past a rough solid wall in the $T = 0$ limit
Authors:
Matthew J Doyle,
Andrei I Golov,
Paul M Walmsley,
Andrew W Baggaley
Abstract:
We present a numerical study, using the vortex filament model, of vortex tangles in a flow of pure superfluid $^4$He in the $T = 0$ limit through a channel of width $D = 1$ mm for various applied velocities $V$. The flat channel walls are assumed to be microscopically rough such that vortices terminating at the walls are permanently pinned; vortices are liberated from their pinned ends exclusively…
▽ More
We present a numerical study, using the vortex filament model, of vortex tangles in a flow of pure superfluid $^4$He in the $T = 0$ limit through a channel of width $D = 1$ mm for various applied velocities $V$. The flat channel walls are assumed to be microscopically rough such that vortices terminating at the walls are permanently pinned; vortices are liberated from their pinned ends exclusively through self-reconnection with their images. Sustained tangles were observed, for a period of 80 s, above the critical velocity $V_c \sim 0.20$ cm s$^{-1} = 20 κ/D$. The coarse-grained velocity profile was akin to a classical parabolic profile of the laminar Poiseuille flow, albeit with a non-zero slip velocity $\sim$ 0.20 cm s$^{-1}$ at the walls. The friction force was found to be proportional to the applied velocity. The effective kinematic viscosity was $\sim 0.1κ$, and effective Reynolds numbers within $\mathrm{Re'} < 15$. The fraction of the polarized vortex length varied between zero in the middle of the channel and $\sim$ 60% within the shear flow regions $\sim D/4$ from the walls. Therefore, we studied a state of polarized ultraquantum (Vinen) turbulence fuelled at short lengthscales by vortex reconnections, including those with vortex images due to the relative motion between the vortex tangle and the pinning rough surface.
△ Less
Submitted 14 February, 2024;
originally announced February 2024.
Interactive Volume Visualization via Multi-Resolution Hash Encoding based Neural Representation
Authors:
Qi Wu,
David Bauer,
Michael J. Doyle,
Kwan-Liu Ma
Abstract:
Neural networks have shown great potential in compressing volume data for visualization. However, due to the high cost of training and inference, such volumetric neural representations have thus far only been applied to offline data processing and non-interactive rendering. In this paper, we demonstrate that by simultaneously leveraging modern GPU tensor cores, a native CUDA neural network framewo…
▽ More
Neural networks have shown great potential in compressing volume data for visualization. However, due to the high cost of training and inference, such volumetric neural representations have thus far only been applied to offline data processing and non-interactive rendering. In this paper, we demonstrate that by simultaneously leveraging modern GPU tensor cores, a native CUDA neural network framework, and a well-designed rendering algorithm with macro-cell acceleration, we can interactively ray trace volumetric neural representations (10-60fps). Our neural representations are also high-fidelity (PSNR > 30dB) and compact (10-1000x smaller). Additionally, we show that it is possible to fit the entire training step inside a rendering loop and skip the pre-training process completely. To support extreme-scale volume data, we also develop an efficient out-of-core training strategy, which allows our volumetric neural representation training to potentially scale up to terascale using only an NVIDIA RTX 3090 workstation.
△ Less
Submitted 29 June, 2023; v1 submitted 23 July, 2022;
originally announced July 2022.
