Search

Fluid dynamics simulations with the lattice Boltzmann method (LBM) are very memory intensive. Alongside reduction in memory footprint, significant performance benefits can be achieved by using FP32 (single) precision compared to FP64 (double) precision, especially on GPUs. Here we evaluate the possibility to use even FP16 and posit16 (half) precision for storing fluid populations, while still carrying arithmetic operations in FP32. For this, we first show that the commonly occurring number range in the LBM is a lot smaller than the FP16 number range. Based on this observation, we develop customized 16-bit formats—based on a modified IEEE-754 and on a modified posit standard—that are specifically tailored to the needs of the LBM. We then carry out an in-depth characterization of LBM accuracy for six different test systems with increasing complexity: Poiseuille flow, Taylor-Green vortices, Karman vortex streets, lid-driven cavity, a microcapsule in shear flow (utilizing the immersed-boundary method), and, finally, the impact of a raindrop (based on a volume-of-fluid approach). We find that the difference in accuracy between FP64 and FP32 is negligible in almost all cases, and that for a large number of cases even 16-bit is sufficient. Finally, we provide a detailed performance analysis of all precision levels on a large number of hardware microarchitectures and show that significant speedup is achieved with mixed FP32/16-bit.

We study the scalar one-component two-dimensional (2D) ϕ 4 model by computer simulations, with local Metropolis moves. The equilibrium exponents of this model are well established, e.g., for the 2D ϕ 4 model γ = 1.75 and ν = 1 . The model has also been conjectured to belong to the Ising universality class. However, the value of the critical dynamical exponent z c is not settled. In this paper, we obtain z c for the 2D ϕ 4 model using two independent methods: (a) by calculating the relative terminal exponential decay time τ for the correlation function ⟨ Φ ( t ) Φ ( 0 ) ⟩ , and thereafter fitting the data as τ ∼ L z c , where L is the system size, and (b) by measuring the anomalous diffusion exponent for the order parameter, viz., the mean-square displacement ⟨ Δ Φ 2 ( t ) ⟩ ∼ t c as c = γ / ( ν z c ) , and from the numerically obtained value c ≈ 0.80 , we calculate z c . For different values of the coupling constant λ , we report that z c = 2.17 ± 0.03 and z c = 2.19 ± 0.03 for the two methods, respectively. Our results indicate that z c is independent of λ , and is likely identical to that for the 2D Ising model. Additionally, we demonstrate that the generalized Langevin equation formulation with a memory kernel, identical to those applicable for the Ising model and polymeric systems, consistently captures the observed anomalous diffusion behavior.

We study the scalar one-component two-dimensional (2D) $\phi^4$ model by computer simulations, with local Metropolis moves. The equilibrium exponents of this model are well-established, e.g. for the 2D $\phi^4$ model $\gamma= 1.75$ and $\nu= 1$. The model has also been conjectured to belong to the Ising universality class. However, the value of the critical dynamical exponent $z_c$ is not settled. In this paper, we obtain $z_c$ for the 2D $\phi^4$ model using two independent methods: (a) by calculating the relative terminal exponential decay time $\tau$ for the correlation function $\langle \phi(t)\phi(0)\rangle$, and thereafter fitting the data as $\tau \sim L^{z_c}$, where $L$ is the system size, and (b) by measuring the anomalous diffusion exponent for the order parameter, viz., the mean-square displacement (MSD) $\langle \Delta \phi^2(t)\rangle\sim t^c$ as $c=\gamma/(\nu z_c)$, and from the numerically obtained value $c\approx 0.80$, we calculate $z_c$. For different values of the coupling constant $\lambda$, we report that $z_c=2.17\pm0.03$ and $z_c=2.19\pm0.03$ for the two methods respectively. Our results indicate that $z_c$ is independent of $\lambda$, and is likely identical to that for the 2D Ising model. Additionally, we demonstrate that the Generalised Langevin Equation (GLE) formulation with a memory kernel, identical to those applicable for the Ising model and polymeric systems, consistently capture the observed anomalous diffusion behavior., Comment: 14 pages, 4 figures, 6 figure files, to appear in Phys. Rev. E

For Sn$_2$P$_2$S$_6$ family of uniaxial ferroelectrics, the anharmonic quantum oscillators model, based on local three-well potential for spontaneous polarization fluctuations, has been used for a description of anharmonic properties, namely thermal expansion. The calculated pseudospin fluctuations spectra in ferroelectric phase demonstrate negative Gr\"{u}neisen parameters for excitations that satisfy Curie-Weiss-like temperature and pressure dependencies in the vicinity of second order phase transition. Thermal expansion coefficient is calculated by evaluation of pseudospins entropy baric dependence. Negative thermal expansion in the ferroelectric phase of Sn$_2$P$_2$S$_6$ crystal is obtained which is in agreement with observed experimental data., Comment: 6 pages, 5 figures

For Sn$_2$P$_2$S$_6$ ferroelectrics, the appearance of spontaneous polarization is related to stereoactivity of tin cations and valence fluctuations of phosphorous cations. Here, the continuous phase transition and its behavior under pressure is determined by thee-well local potential and can be described in an anharmonic quantum oscillator model. For such a model, the spectrum of pseudospin fluctuations at different temperatures and pressures has been calculated and compared with the data of Raman spectroscopy. It was revealed that the ferroelectric lattice instability is related to several low energy optic modes., Comment: 8 pages, 8 figures