Filename
stringlengths 22
64
| Paragraph
stringlengths 8
5.57k
|
|---|---|
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Figure 14: Comparison of phase φ((cid:126)x) and magnitude ρ((cid:126)x) clusters at T = 0.89(1) TC after four sweeps of stout-link smearing, using the isovolume renderer, cut much closer into the peaks than Fig. 13. Here, the rendering thresholds for φ((cid:126)x) and ρ((cid:126)x) are 0.9 and 0.5 respectively. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Figure 15: Scatter plot of center phase proximity and modulus, showing a clear correlation between proximity to a center phase and maxima of the modulus. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Figure 16: Two of the three percolating clusters in a representative 243 figuration at T = 0.89(1) TC when using a cut parameter of zero. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | to the core of the cluster. Above the critical temperature, the peaks become much broader so the gradient is significantly smaller and the restoring force is greatly reduced. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | If we include every site in a cluster (i.e. we set the cut parameter δ to zero), then below the critical temperature each sector is equally occupied so our clustering is comparable to random site percolation theory with an occupation 0.3333. This is above the critical percolation probability of probability of p pc = 0.3116 [29]. Thus, with a cut parameter of zero, we expect to see at least one percolating cluster (i.e. a cluster that has at least one site in each of the 3Ns spatial planes) for each phase. Thus, if we define the clusters in this way, due to the nature of three dimensional space they are not localized, and instead extend over the entire lattice, as seen in Fig. 16. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Figure 17: One of the non-percolating clusters broken off from each of the clusters in Fig. 16 when the cut parameter is increased to 0.3. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | clusters scattered across the lattice. Fig. 17 illustrates results for δ = 0.3. We can see that these clusters have a one dimensional, finger-like quality to their structure and this supports the determination by Endrodi, Gattringer and 1.7. Schadler [13] that the center clusters have a fractal dimension of D As illustrated in Fig.15, the selection of a cut in φ((cid:126)x) identifies a subset of points where ρ((cid:126)x) is small and attributes the points to a domain wall of finite thickness. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Figure 18: Comparison of real component of correlation function for points in the same and in different clusters and averaged across all points. Generated from 243 8 lattice at T = 0.89(1) TC with a cut parameter of 0.0. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | That is, the phases at (cid:126)x and (cid:126)y cancel out, and the correlation function is non-zero. Thus, Fq ¯q((cid:126)x (cid:126)y) is finite and quark-antiquark pairs can be − separated within a cluster. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Figure 19: Comparison of real component of correlation function for points in the same and in different clusters and averaged across all points. Generated from 243 8 lattice at T = 0.89(1) TC with a cut parameter of 0.3. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Figure 20: Comparison of real component of correlation function for points in the same and in different clusters and averaged across all points. Generated from 243 8 lattice at T = 0.89(1) TC with a cut parameter of 0.5. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Figure 21: Volume and cut parameter dependence of the plateau in the correlation function for different clusters at T = 0.89(1) TC . Cut parameters are indicated in the legend. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | a finite size and a volume significantly larger than that size. This allows us to access an even distribution between the three phases so that it can average to zero. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Thus, the cluster size has a physical significance, governing the confining scale of the theory and the size of mesons. Below the critical temperature, when (given a sufficiently large cut parameter), the clusters have a limited finite size, this results in confinement. From a model perspective, this scale governs the size of the quark core of hadrons which is dressed by the mesonic cloud. Above the critical temperature, as a single cluster grows to encompass the entire space (regardless of the cut used), the quarks become deconfined [9]. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Polyakov loop, and clearly displays their phase (and/or absolute value). This allows us to observe the evolution of center clusters with HMC simulation time. For the first time, we were able to reveal the evolution of center clusters as they transitioned from the confined to the deconfined phase. The cluster behavior is consistent with that of spin aligned domains in the three dimensional 3-state Potts model. This supports the idea that the phase transition in QCD is comparable to that of a three state spin system. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | We also observe an approximate one to one correspondence between peaks in the magnitude of local Polyakov loops and the locations of center clusters defined by the proximity of the phase of the Polyakov loops to a center phase. Gaps between the clusters illustrate the magnitude of the Polyakov loop is suppressed within the domain walls. This observation supports the underlying assumption of Ref. [5], linking the center domain walls to phenomena observed at RHIC [28] and the LHC [6–8]. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | The creation of a domain wall of finite thickness between clusters through the cut parameter δ [9] produces clusters in which the quark-antiquark correlation function (cid:10)L((cid:126)x)L†((cid:126)y)(cid:11) vanishes when both quarks do not reside in the same cluster, setting a scale for confinement. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | These peaks in the magnitude are sharp below the critical temperature. Through considerations of the free energy of multi-quark systems, this results in a minimum of the quark-antiquark potential in the core of each center cluster with a steep gradient resulting in a strong restoring force confining the quarks within the cluster. Above the critical temperature the peak structure becomes smooth and the region covered by the domain becomes large. When averaged over an ensemble, any remaining fluctuations in the Polyakov loops are smoothed In this way, quarks become out and the restoring force becomes negligible. deconfined. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | In order to perform the visualisation, we use a ray-traced volume renderer. For each pixel in the final image, a single ray is traced out directly away from the viewer through the volume to be rendered, accruing color and opacity based on the volumetric data. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | In our particular case, we adapt a technique by Kruger and Westermann [32] which involves repeatedly rendering a single cube with a sequence of different shader pairs. The vertex shader is the same every time and performs a simple transformation on the cube and calculates the mapping between points on the surface of the cube and points in the volume data that is being rendered. We then have three different fragment shaders that are run in sequence to produce the desired output. A feature that we make extensive use of in order to store interim data is rendering to a framebuffer, an image in memory that serves as a virtual screen, allowing us to store the result of one fragment shader and then use it in a later shading run. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | The first fragment shader is run with only the outside faces of the cube visible, so the vertex shader gives the coordinates of the point a ray cast through the current pixel would enter the volume. We store these directly in the red, green, and blue channels of a framebuffer. We then run the second shader with only the inside faces of the cube visible, so the vertex shader gives the coordinates of the point the ray would leave the volume. We then access the entry coordinates from the framebuffer and calculate the direction and length of the ray inside the volume and store them in the red, green, blue, and alpha channels of a new framebuffer. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | We can then use a more complicated fragment shader to preform the inte- gration, indexing into a 3D texture containing the volume data and calculating C vol((cid:126)x) and αvol((cid:126)x) at each step. By setting the interpolation mode on the texture, we can tell OpenGL to automatically and efficiently perform trilinear interpolation on the data. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | By transforming the cube, we can transform the volume being rendered. Thus we perform standard OpenGL model/view and projection transformations on the cube to place the volume in the center of the screen with perspective and continuously rotate it so that it is possible to see all sides of the volume and observe the 3D structure it contains. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | memory and convert it into a range of formats for later use. In particular, we use DevIL[33] to convert a single frame into a static image, or we use FFmpeg[34] to combine a sequence of frames into a video. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | Implementing this algorithm naively is rather inefficient for all but the most sparse data sets, as much of the volume is obscured by opaque or nearly opaque regions and has little to no effect on the final image. In order to eliminate this inefficiency, we introduce early ray termination, that is we stop integrating rays once they reach a certain opacity threshold. |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | We do this by introducing another fragment shader that writes to the depth buffer without changing the color or opacity. The depth buffer is a special texture used by OpenGL to determine what geometry should be obscured by other geometry. If the opacity has reached a certain threshold, our shader writes the minimum possible value to the depth buffer, effectively terminating the ray. This shader is then interleaved with the integrating shader, which writes to a framebuffer to store its interim result. This method is also used to stop integrating rays that have left the volume, simply by setting their opacity to 1 if they are outside the volume (equivalent to hitting a solid black backdrop). |
Processed_Visualisations_of_coherent_centre_domains_in_local.txt | This maps φ = 0 (h = 0) to red, φ = 2π to blue. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | where perturbation theory holds, we derive the one-loop self-energy correction to the single-gluon spectrum, for arbitrary N and m. We discuss the transition from small to large volumes that has been investigated by means of Monte-Carlo simulations. We argue that the energy of electric flux ~e, for the lowest gluon momentum, depends solely on ~e/N and the dimensionless variable x = l Nl, with l = g2N the ’t Hooft coupling. The variable x can be interpreted as the dimension- less ’t Hooft coupling for an effective box size given by Nl. This implies a version of reduction that allows to trade l by N without modifying the electric-flux energy. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | c(cid:13) Copyright owned by the author(s) under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike Licence. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | In this paper we will explore the electric-flux spectrum of a 2+1-dimensional SU(N) Yang- Mills theory defined on a 2-dimensional spatial torus endowed with a chromo-magnetic flux. The aim is to disentangle the dependence on the number of colours N and the size of the torus l, fol- lowing the idea of reduction introduced by Eguchi and Kawai [1, 2]. The size of the torus will act as a control parameter that permits to explore the onset of non-perturbative dynamics. This is so, because for small l the effective coupling constant becomes small and perturbation theory holds. As l grows, the finite size effects, including the magnetic flux, should become irrelevant. The way in which this takes place, and in particular the interplay with the large N limit, might shed some light into the processes relevant for non-perturbative dynamics. We will focus here for simplicity on the 2+1 dimensional case that shares many of the non-perturbative properties of the 4-dimensional theory. There is an extensive literature on the subject of Yang-Mills 3-d fields and large N dynamics, for recent lattice reviews we refer the reader to [3]. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | Figure 1: We show, for electric flux ~e = (Nq , 0) (left), and ~e = (Nq , Nq ) (right), the function G(q ) that gives the one-loop correction to the energy of a one-gluon state, through Eq. (2.5). The blue line corresponds to Eq. (2.6), while the red points are derived using a lattice regularization in the calculation of the self-energy. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | formalism looks hence as if the theory had no colour, but was defined on a bigger spatial box with effective size Nl. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | We will now analyze in perturbation theory the energy spectrum of one-gluon states. The first remark is that, since zero momentum is excluded, the theory has a mass-gap even for small Nl. At tree level in perturbation theory, the mass of a state with colour momentum ~p c is E0 = = 2p /(Nl). Such states can be characterized in a gauge-invariant way in terms of the electric flux | ~e, which is defined modulo N and related to ~n through: ei = e jin j ¯k. The gauge invariant operator projecting onto a state of electric flux ~e is a Polyakov loop winding ei times around each of the cycles of the 2-torus. The non-perturbative spectrum of such states can hence be obtained from Polyakov loop correlators. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | The dependence of G on ~e/N is exhibited in Fig. 1, for electric fluxes of the form ~e = (Nq , 0) and ~e = (Nq , Nq ). |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | One important remark at this stage is that, once ~n and ~e/N have been fixed, the energy de- pends exclusively on the variable x = l Nl, which can be interpreted as the dimensionless ’t Hooft coupling corresponding to an effective box of size Nl. Based on our numerical results, we will argue that this is the case even in the non-perturbative regime. This will imply thus a kind of re- duction in which N and l become interchangeable. A second remark in order is the non-analyticity of G(q ) at q = 0 (mod 1). This could give rise to a tachyonic instability in the energy of electric flux for N , which has been previously discussed in the context of non-commutative geometry in Ref. [9]. There it was presented as an evidence of spontaneous ZN symmetry breaking. We will come back to this issue when discussing the numerical results in section 4. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | Let us finally mention that in order to arrive at the one-loop perturbative expression we have used three different approaches: to compute the Euclidean gluon self-energy in the continuum and on the lattice, and to use the Hamiltonian formulation. The lattice calculation has made use of the results in Refs. [7],[8]. In Fig. 1, the lattice results for G(q ), extrapolated to the continuum limit, are compared with the formula given in Eq. (2.6). There is perfect agreement between the two. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | i n with m the magnetic flux. We have performed Monte Carlo simulations with this action employing a heat-bath algorithm based on a proposal by Fabricius and Haan [10]. Four sets of (N, L, L0) values have been generated: (5,14,48), (5,22,72), (7,10,32), and (17,4,32), with values of the magnetic flux: m = 1, 2, for SU(5), m = 1, 2, 3, for SU(7), and m = 1, 3, 5, 8, for SU(17). |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | Figure 2: x = l Nl, where E is the energy of electrix flux e with momentum is the lattice tree-level perturbative result. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | with e = e 21 n ¯k. An analogous expression holds for C2(t, e, n). The energy is extracted from the exponential decay at large t of the correlator. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | A detailed account of the results will be presented in [6]. Here we will focus on a few examples that illustrate the scaling of the electric-flux energies with x = l Nl. We will focus on the minimal = ¯k. According to our perturbative momentum states with calculation, for given e/N we expect energies to depend solely on x. Given that N is prime, identical ratios of e/N are not possible, nevertheless our data falls in three sets with similar values for the ratio: (N, e) = . The dependence of the energy (5, 2), (7, 3) ; (7, 2), (17, 5) } { } { squared on x is presented in Fig. 2. The scaling holds in all the range of x analyzed which goes far beyond the perturbative regime. A discussion of the results will be presented below. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | Figure 3: the same as in Fig. 2. Different lines correspond to: (combined) a = G(e/N), b and g a = G(e/N), b = g = 0; (string) a = 0, b and g free. The lower panels display d E 2(e)x2/l 2. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | Fitting this expression to our data gives the results displayed in Fig. 3. The fits are restricted to the large x region with x > 20 (x > 30, for N = 17, e = 3). Two different fits are presented: (a) fixing a = G(e/N), denoted by ‘combined’, and (b) fixing a = 0, denoted by ‘string’. In all cases, the ‘combined’ fit has a slightly better c 2 per degree of freedom than the ‘string’ fit. In addition, we display the 1-loop expression. Although the combined formula fails to reproduce the results in the intermediate x region it describes qualitatively quite well the data. One remarkable observation is that for small x it improves the one-loop result in the correct direction. This is clearly observed in the low panels of Fig. 3 where d E 2x2/l 2 is plotted to enhance this effect. We are at present exploring alternative parametrizations to improve the fit in the intermediate x regime. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | One final remark refers to the possible existence of tachyonic instabilities. One could use Eq. (4.4) to see in which instances the energy remains non-tachyonic for all values of x. A more thorough analysis will be presented in [6]. Let us just mention here that our results indicate that it suffices to keep both m and ¯k of order N to avoid the tachyonic behaviour. This coincides with the criteria introduced in [2] for twisted Eguchi-Kawai reduction to hold at large N. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | M.G.P. and A.G-A acknowledge financial support from the MCINN grants FPA2009-08785 and FPA2009-09017, the Comunidad Autónoma de Madrid under the program HEPHACOS S2009/ESP- 1473, and the European Union under Grant Agreement number PITN-GA-2009-238353 (ITN STRONGnet). They participate in the Consolider-Ingenio 2010 CPAN (CSD2007-00042). M. O. is supported by the Japanese MEXT grant No 23540310. We acknowledge the use of the IFT clusters for our numerical results. |
Processed_Volume_dependence_in_2+1_Yang-Mills_theory.txt | D. Daniel, A. González-Arroyo and C. P. Korthals Altes, Phys. Lett. B 251, 559 (1990). |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | ABSTRACT Turbulence in the interstellar medium (ISM) is crucial in the process of star formation. Shocks produced by supernova explosions, jets, radiation from massive stars, or galactic spiral-arm dynamics are amongst the most common drivers of turbulence in the ISM. However, it is not fully understood how shocks drive turbulence, in particular whether shock driving is a more solenoidal (rotational, divergence-free) or a more compressive (potential, curl-free) mode of driving turbulence. The mode of turbulence driving has profound consequences for star formation, with compressive driving producing three times larger density dispersion, and an order of magnitude higher star formation rate than solenoidal driving. Here, we use hydrodynamical simulations of a shock inducing turbulent motions in a structured, multi-phase medium. This is done in the context of a laser-induced shock, propagating into a foam material, in preparation for an experiment to be performed at the National Ignition Facility (NIF). Specifically, we analyse the density and velocity distributions in the shocked turbulent medium, and measure the turbulence driving parameter 𝑏 = (𝜎2Γ 𝜌/(cid:104)𝜌(cid:105))−1/2M−1Γ−1/2, with the density dispersion 𝜎𝜌/ (cid:104)𝜌(cid:105), the turbulent Mach number M, and the polytropic exponent Γ. Purely solenoidal and purely compressive driving correspond to 𝑏 ∼ 1/3 and 𝑏 ∼ 1, respectively. Using simulations in which a shock is driven into a multi-phase medium with structures of different sizes and Γ < 1, we find 𝑏 ∼ 1 for all cases, showing that shock-driven turbulence is consistent with strongly compressive driving. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Turbulence in the interstellar medium plays a key role in the process of star formation (Mac Low & Klessen 2004; Elmegreen & Scalo 2004; McKee & Ostriker 2007; Hennebelle & Falgarone 2012; Padoan et al. 2014). The statistics of turbulence can be used to determine the star formation rate (SRF) (Krumholz & McKee 2005; Padoan & Nordlund 2011; Hennebelle & Chabrier 2011; Federrath & Klessen 2012), as well as the initial mass function (IMF) of stars (Padoan & Nordlund 2002; Hennebelle & Chabrier 2008; Hopkins 2012). However, the driving mechanisms of this turbulence are not well understood and remain an active area of research. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | and typically found values 𝑏 (cid:38) 0.5 in the molecular clouds IC5146, Taurus, and GRSMC43.30-0.33, located in the spiral arms of the Milky Way, indicating primarily compressive driving of turbulence in these clouds. Such driving may be the result of supernova explo- sions, shock compression in the spiral arms, gravitational collapse, or radiation feedback from massive stars. The latter was confirmed by Menon et al. (2021) in several gas/dust pillars in the Carina Nebula, where they find 𝑏 ∼ 0.7–1.0, i.e, the radiation from nearby massive stars that sculpted these pillars also drives compressive turbulence inside them. In other regions of the galaxy, the turbulence driving is more solenoidal, such as in the Central Molecular Zone (CMZ) cloud ‘Brick’ (G0.253+0.016), where Federrath et al. (2016) found 𝑏 ∼ 0.2–0.3. This predominantly solenoidal driving mode of turbu- lence is due to the strong shearing motions in the CMZ, driving turbulence in the clouds near the Galactic Centre. Thus, we expect different physical drivers of turbulence in various galactic environ- ments to produce different values of 𝑏 (Federrath et al. 2017). |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | The turbulence driving parameter 𝑏 plays a key role for star for- mation. Compressive driving (𝑏 ∼ 1) can produce star formation rates more than an order magnitude higher than solenoidal driving (𝑏 ∼ 0.3) (Federrath & Klessen 2012; Federrath 2018). Thus, it is crucial to determine the driving mode 𝑏 of different physical drivers of turbulence, in order to understand and predict the star formation activity in different environments. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Here we aim to determine the driving mode of turbulence in- duced by hydrodynamical shocks. Such shock-driven turbulence is very common in the interstellar medium (ISM). For example, super- nova (SN) explosions or radiation from massive stars drive shocks in the ISM, which drives turbulence (Mac Low & Klessen 2004). Simulations of SN-driven turbulence find that the turbulence that the SN explosions produce is relatively solenoidal in the velocity field (Pan et al. 2016). However, even purely compressive driving typically produces a ∼ 50% fraction of solenoidal modes in the velocity field (Federrath et al. 2010), by non-linear shock interac- tions and baroclinic vorticity generation in a multi-phase (multi- temperature) medium (Del Sordo & Brandenburg 2011). The pres- ence of solenoidal modes in the velocity field, however, does not imply a dominance of solenoidal driving of the turbulence, and in fact, the density fluctuations may still follow Eq. (1) with 𝑏 ∼ 1. This is particularly true in a medium of nearly constant temperature, where the baroclinic term is ineffective, such as in dense molecular clouds, where cooling is very efficient and the gas remains roughly isothermal (e.g., the pillars in the Carina nebula, studied by Menon et al. 2021). Further examples of flow driven turbulence include Gi- acalone & Jokipii (2007), Guo et al. (2012), and Drury & Downes (2012), in which turbulence is generated through the differential ac- celeration of the upstream ISM, which occurs as a result of density inhomogeneities. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | In order to determine the driving parameter 𝑏 of shock-driven tur- bulence, we run a set of numerical simulations, where we investigate the properties of the turbulence driven by a radiation-induced hydro- dynamical shock impacting a pre-structured, multi-phase medium (modelled as a porous medium, i.e., foam). Our simulations are designed to mimic laboratory experiments of laser-induced, shock- driven turbulence, to be conducted at the National Ignition Facility (NIF), Lawrence Livermore National Laboratory. While these sim- ulations provide predictions and calibrations for the laboratory ex- periments, they also provide the opportunity to study the statistics of shock-driven turbulence in relatively simple, controlled numerical experiments. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | drodynamical equations solved, and the numerical methods used here. In Section 3 we discuss the density and velocity PDFs of the turbulence. We also discuss the thermodynamic properties of the multi-phase medium. We then study the dependence of the turbu- lence properties on the size of the structures in the foam (Section 4), by considering three simulations with different foam void diameters. Using these results, we calculate the turbulence driving parameter of the shock-driven turbulence in Section 5. We also check if the driving parameter depends on the choice of analysis volume, time, or numer- ical resolution (Appendices). We compare our results with those of astrophysical observations and other simulations in Section 6. Sec- tion 7 summarises the results and conclusions of this study. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | We consider a cylindrical tube with an outer diameter of 2.0 mm and a total length of 3.8 mm as shown in Fig. 1. The walls of the tube with a thickness of 0.1 mm are made of beryllium, which has a density of 1.85 g/cm3. The tube is filled with foam whose void diameter is 50 𝜇m; however, we are also considering cases with smaller and bigger void sizes below. The foam has an adiabatic index 𝛾 = 5/3. At the bottom of the tube is a 0.3 mm thick plastic (CH-based) ablator, on which the laser beam is incident. A thin aluminium plate is placed on top of the ablator. The density of the ablator is 1.4 g/cm3 and that of the aluminium plate is 2.7 g/cm3. The details of this setup, including the specific materials used, are only important for the laboratory experiment, however, these specifics will likely change during the current design phase of the experiment, most notably the details of the shock tube (Nagel et al. 2017) and foam material (Hamilton et al. 2016). However, none of these specifics and details are important for the numerical simulations studied here, as we will only extract information in the foam material sufficiently far away from the boundaries, such that we can study the statistics of the turbulence in the foam, without any influence from the cylinder walls. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | All materials are in pressure equilibrium at the start of the sim- ulation. The laser is directed at the tube from the bottom (inci- dent from the negative 𝑦 direction) and induces a hydrodynamic shock wave, which is initialised in the ablator material. As an ini- tial condition for this hydrodynamical shock wave, we use a detailed radiation-hydrodynamical simulation of a 1D shock (see details in Appendix A). We then interpolate this 1D shock into the 3D domain, spanning the whole width (in 𝑥 and 𝑧 of the ablator). This interpo- lation serves to initialise a plane-parallel shock travelling in the 𝑦 direction, starting in the ablator at 𝑦 = 0.3 mm. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | The assumed foam material (excluding voids) in these simulations has a density of 0.1 g/cm3. The mean density of the foam (including the voids) is set to 0.05 g/cm3. The voids are modelled as spheres with a diameter of 50 𝜇m for the standard simulation shown in Fig. 1, and have the density of air (1.225×10−3 g/cm3). An increasing num- ber of voids is placed randomly (sampled from a uniform distribution of positions) in the foam, until the total mean density of the foam has reached the target mean density of 0.05 g/cm3. The voids are allowed to overlap. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 1. Volume density slices at time 𝑡 = 0 (initial conditions), showing the geometry of the basic setup. The top panel shows the 𝑥𝑧-slice of the material (volume) density taken at 𝑦 = 2 mm, and the bottom panel shows a respective 𝑥 𝑦-slice taken at 𝑧 = 0 mm. The hydrodynamical shock wave is initialised in the plastic ablator material (𝜌0 = 1.4 g/cm3), seen in the bottom panel, and then travels into the foam (here shown with void diameter of 50 𝜇m). The circular white structures represent voids, meant to roughly correspond to the cell structure of an actual foam. The non-void material (red) is a CH-based polymer with a density of 𝜌0 = 0.1 g/cm3. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | domain. The same applies in all spatial directions, i.e., ultimately, the wall materials will also leave the computational domain. This means that there is no influence of the numerical boundary on the results, because we are only extracting information from within a small portion of the domain, where the medium can be considered turbulent, and no information is allowed to propagate into the domain from the boundaries. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | We use a modified version of the FLASH code (Fryxell et al. 2000; Dubey et al. 2008) (v4), to solve the three-dimensional compressible hydrodynamical equations, with the setup and initial conditions spec- ified in Sec. 2.1 and 2.2. The system of hydrodynamical equations is solved with the robust HLL5R approximate Riemann scheme (Waa- gan et al. 2011). We also compared the present simulations using the HLL5R solver with the PPM solver (Colella & Woodward 1984), and find very good agreement. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | The computational grid has 768 × 1024 × 768 compute cells, cap- turing the large-scale turbulent dynamics and sufficient to converge on the statistical properties of the turbulent gas (Kitsionas et al. 2009; Price & Federrath 2010; Kritsuk et al. 2011b). This is also confirmed by a resolution study presented in Appendix G. The number of cells in the 𝑦-direction is larger than in the 𝑥 and 𝑧 directions, which ex- actly compensates for the different computational domain length in 𝑦 as opposed to 𝑥 and 𝑧. Therefore, the physical cell width is the same in all three directions. We are not using any explicit viscosity in these simulations, i.e., we are conducting Implicit Large Eddy Simulations (ILES). Thus, the Reynolds number is not directly con- trolled, but increases with the numerical grid resolution. Considering a resolution of 𝑁 cells in linear direction across the turbulent region, grid codes in ILES mode usually result in Reynolds numbers of the order of Re ∼ 𝑁4/3 (Benzi et al. 2008; Federrath et al. 2011; McKee et al. 2020), which means that we expect to develop turbulent flow in the simulations, i.e., typically Re > 103 (Frisch 1995). In the real experiments, we also expect turbulent flow (Robey et al. 2003). |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | For the thermodynamics, we use an ideal equation of state (EOS), where all materials have an adiabatic index 𝛾 = 5/3. The foam is modelled as ionised CH-based polymer, which has a mean molecular weight of 6.5, and the foam voids are assumed to be filled with air at a mean molecular weight of 29. This is a simplified approach, where phase transitions, chemistry, radiation and cooling of the ma- terials are neglected. While these processes would introduce some quantitative changes, e.g., in the details of the shock Mach num- ber, temperature of the foam, and mean molecular weight during the shock passage, the qualitative evolution remains the same. In particular, for the purposes of understanding the turbulence driving mode of this system, the details of the thermodynamical modelling do not affect our basic results and conclusions. This is because the driving parameter is ultimately a property of the shock driving it- self, and the response of the system being shocked is such that we are only interested in the ratio of relative density dispersion 𝜎𝜌/(cid:104)𝜌(cid:105) and turbulent Mach number; see Eq. (1). Thus, while the details of the thermodynamics, chemistry, radiation and cooling will alter the density dispersion and Mach number in their absolute values (for in- stance, strong cooling would allow for higher M), the ratio of these dimensionless numbers, i.e., the driving parameter 𝑏, is expected to remain largely invariant. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | eter only depends on dimensionless numbers that characterise the turbulent flow, which can be compared for systems of vastly different absolute scale, because turbulence is a ubiquitous, scale-free process. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Our goal is to study the properties of shock-driven turbulence in a pre-structured medium. Here we measure the density and velocity distributions in the turbulent shocked foam, and finally obtain the turbulence driving parameter 𝑏. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | We first consider a foam with void size of 50 𝜇m. In order to study the properties of the turbulence, we select a time after the initiation of the shock, at which there is a large enough region of well-developed turbulence. A cuboidal region of sufficiently developed turbulence is selected at time 𝑡 = 75 ns, which is represented by the white rectangle in Fig. 2 at 𝑡 = 𝑡3. The analysis region has dimensions of 1.2 mm × 0.34 mm × 1.2 mm. The density, sound speed and Mach number PDFs are obtained and analysed for this region, and the turbulence driving parameter is calculated. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | We also investigate the dependence of the turbulence properties on the foam void size by considering two more simulations with a smaller (12.5 𝜇m) and a larger (100 𝜇m) foam void size along with the standard 50 𝜇m foam void case. The analysis procedure is the same as that for the 50 𝜇m void case. The time evolution and volume dependence of the turbulence is also studied by considering different time instances and different analysis volumes for the 50 𝜇m foam void size case (Appendices E and F). |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Fig. 2 depicts the time evolution of the shock-driven turbulence in the foam with void size of 50 𝜇m. As the shock passes through the foam from the bottom, it drives turbulent density and velocity fluctuations (see top and bottom panels of Fig. 2, respectively). As the shock wave propagates into the foam tube, the walls melt, and are pushed outwards. The volume of the turbulent region increases as the shock progresses further. Finally, the shocked foam bursts out of the tube at the top (𝑦 = 3.8 mm). The whole process happens within ∼ 100ns. The right-hand panels at 𝑡 = 𝑡3 of Fig. 2 show the time at which we carry out our analysis. The cuboidal region selected for the turbulence analysis is shown as a white rectangle. This region was selected to minimise the impact of systematic biases from the boundaries or the shock passing through the foam. In Appendix E and F, we show that our main results do not depend on the specific choice of the analysis region or the time of the analysis. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | velocity along the direction of the propagation of the shock, i.e., the shock induces a systematic bulk motion along the 𝑦-direction. The average velocities in the 𝑥 and 𝑧 directions are very close to zero. The average physical properties of the analysis region are listed in Table 1. The small bulk velocities in the 𝑥 and 𝑧 directions are because of density inhomogeneities and because the analysis region does not cover the entire simulation volume1. However, these bulk velocities do not affect the results of this study, as we subtract them in the definition of the local Mach number, which is the quantity of interest. Despite the non-zero bulk flow along the 𝑦 direction, the system is largely isotropic. The velocity fluctuations along the shock direction are somewhat larger than perpendicular to it, which we will quantify in detail below. The density fluctuations are relatively independent of the direction. In addition to these bulk flows, we also find that there are weak density and velocity gradients. In Appendix D we investigate the influence of these gradients, and find that they are small compared to the main fluctuations in the velocity and density. Thus, our following analyses are practically independent of the presence of these systematic gradients. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Here we quantify the density fluctuations and thermodynamic prop- erties of the turbulence in the analysis region. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | with the mean density (cid:104)𝜌(cid:105) in the analysis region. The quantity 𝑠 has long been used in the literature to study the density PDF, because in this transformation to the natural logarithm of 𝜌/(cid:104)𝜌(cid:105), the PDF of 𝜌 is often found to be nearly log-normal for isothermal turbulence, that is, the PDF of 𝑠 is Gaussian (Vázquez-Semadeni 1994; Passot & Vázquez-Semadeni 1998; Kritsuk et al. 2007; Federrath et al. 2008, 2010). |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 2. Time evolution of the simulation with 50 𝜇m foam voids, showing 𝑥 𝑦-slices of the material density (top panels) and the 𝑧-component of the velocity (bottom panels) at 𝑧 = 0. The snapshots are taken at times 𝑡1 = 25 ns, 𝑡2 = 50 ns and 𝑡3 = 75 ns (from left to right). The white rectangles in the right-hand panels show the region in which we carry out the subsequent turbulence analysis. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Because of these constraints, only 3 out of the 6 parameters in Eq. (3) are independent. We choose the slope of the PL (𝛼), the width of the LN part (𝜎𝑠), and the transition point (𝑠𝑡 ) as the free parameters. The remaining 3 parameters, 𝑁, 𝑝0, and 𝑠0, are determined from the 3 free parameters and the imposed 3 constraints. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | We fit Eq. (3) to the PDF data. As seen from Fig. 4, the LN+PL PDF provides a very good fit. The parameters of the LN+PL fit are provided in Table C1. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | lainen & Federrath 2017; Alves et al. 2017). The most common ori- gin of these high-density PL tails is gravitational collapse (Klessen 2000; Kritsuk et al. 2011a; Federrath & Klessen 2013; Girichidis et al. 2014; Burkhart & Mocz 2019; Körtgen et al. 2019; Jaupart & Chabrier 2020; Khullar et al. 2021). A gravitational origin of the PL tail seen in our simulation is of course excluded. However, two other physical processes can also produce high-density PL tails: 1) strong shock compression (Tremblin et al. 2014), and 2) an effective polytropic index Γ < 1, in the equation of state for the thermal gas pressure, 𝑝th ∝ 𝜌Γ (Passot & Vázquez-Semadeni 1998; Federrath & Banerjee 2015). |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 3. Maps showing the turbulence analysis region in an 𝑥𝑧 slice of the density (top left panel), an 𝑥𝑧 slice of the 𝑦-velocity (top right panel), an 𝑥 𝑦 slice of the density (bottom left panel), and an 𝑥 𝑦 slice of the 𝑧-velocity (bottom right panel). All the slices pass approximately through the centre of the analysis region, at ( 𝑥, 𝑦, 𝑧) = (0.0, 3.5, 0.0) mm. As seen from the velocity slices, there is a net velocity in the 𝑦-direction, along which the shock propagates. Despite this non-zero bulk speed, the density and velocity structures are largely isotropic, and exhibit a morphology typical of fully-developed turbulence. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 4. Probability density function (PDF) of the logarithmic density con- trast 𝑠 = ln(𝜌/(cid:104)𝜌(cid:105)) in the analysis region. The PDF follows a Gaussian dis- tribution at low 𝑠 up to around the peak, followed by a power-law tail towards high densities. The dashed lines represent the log-normal plus power-law (LN+PL) fit via Eq. (3), with fit parameters reported in Tab. C1. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 5. Sound speed PDF of the gas in the analysis region, where sound speed is defined as 𝑐𝑠 = √︁𝛾 𝑝th/𝜌 with the mean sound speed (cid:104)𝑐𝑠 (cid:105). The dashed lines represent an Exponential + Gaussian fit (see Eq. C1), the details of which are provided in Appendix C. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 6. Maps showing 𝑥𝑧-slices of the density contrast parameter, 𝑠 (left panel), and sound speed (right panel) taken at 𝑦 = 3.5 mm. Both quantities are scaled with respect to their respective mean values in the analysis region, and the natural logarithm is taken for better visualisation and comparison. We see that regions of higher density correspond to lower sound speed and vice versa. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | to colder, denser gas, as opposed to the Gaussian component seen at higher 𝑐𝑠 (warmer, lower-density gas). The sound speed PDF therefore mirrors the shape of the density PDF, suggesting a link between the two. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | To visualise this link, we show 𝑥𝑧-slices of the density contrast parameter 𝑠 and the sound speed in Fig. 6. It is clearly seen that over- dense regions have lower sound speed and vice versa. This is because initially, owing to pressure equilibrium, the foam (which has a higher density) is cooler as compared to the air in the foam voids (which has a lower density). These two phases of the medium, i.e., the foam cell wall material and the air in the voids, mix as the shock passes and the medium becomes turbulent. Since the adiabatic 𝛾 = 5/3, the compressed regions will heat up. However, the previously over-dense foam cell wall material still stays cooler compared to the under-dense voids. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | In order to quantify the link between density and sound speed, we show the density – sound speed correlation PDF of the analysis region in Fig. 7. The PDF indicates a negative correlation between the two quantities. This is in agreement with what is observed from the slice plots. Most of the points have relatively lower density and higher sound speed. However, there is a large tail corresponding to cooler, over-dense gas. For a polytropic gas, 𝑐𝑠 ∝ ( 𝑝th/𝜌)1/2 ∝ (𝜌Γ/𝜌)1/2 ∝ 𝜌 (Γ−1)/2 (Federrath & Banerjee 2015). In the correlation PDF, we plot lines of constant polytropic Γ to find the effective Γ of the system. From Fig. 7, we see that the system has an effectively soft equation of state with Γ ∼ 0.0 to 0.2. As demonstrated and quantified in Passot & Vázquez-Semadeni (1998) and Federrath & Banerjee (2015), a polytropic exponent of Γ < 1 gives rise to a high-density power-law tail in the 𝑠-PDF and a low-𝑐𝑠 power-law region in the sound speed PDF, which is what we observed in Figs. 4 and 5, respectively. Thus, we conclude that the high-density tail in the density PDF and the low-𝑐𝑠 tail in the sound speed PDF are the result of the multi-phase nature of the medium, with an effective polytropic Γ < 1. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 7. Density – sound speed correlation PDF for the gas in the analysis region. We see a clear negative correlation between density and sound speed. For comparison and to guide the eye, three lines of constant polytropic Γ are plotted, one each at the upper and the lower edge of the PDF, and one passing approximately through its centre. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Now that we have gained a basic understanding of the features (LN+PL) in the density PDF, we can focus our attention on the density fluctuations, 𝜎𝜌/ (cid:104)𝜌(cid:105) in the analysis region. The overall tur- bulent density fluctuations are required in order to determine the turbulent driving parameter 𝑏 via Eq. (1). |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | where 𝑛 is the total number of cells in the analysis volume, and 𝜌𝑖 is the density in cell 𝑖. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | The value of 𝜎𝜌/(cid:104)𝜌(cid:105) calculated directly from the data is 0.270, whereas the value obtained by integrating over the fitted density PDF via Eq. (5) is 0.269. Thus, the values of the total density fluctuations obtained using the two methods agree very well, with a deviation of only ∼ 0.4%. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | where 𝑐𝑠 and 𝑣 𝑥 are the sound speed and the 𝑥-component of the velocity at any given point, and (cid:104)𝑣 𝑥(cid:105) is the mean 𝑥-component of the velocity in the analysis volume. M𝑦 and M𝑧 are defined in analogy to M𝑥. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | is due to intermittency and/or the direct influence of the shock. The parameters of the Gaussian fits are listed in Table C1. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Overall, we find that the turbulence behind the shock is subsonic. From the data, we find M = 0.43, while the Gaussian fits above give M = 0.39 (see Tab. C1), i.e., they are nearly identical, with the fitted value being 10% smaller due to the fact that the fit underestimates the velocity fluctuations in the far wings of the distribution. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | In order to study how the turbulence properties change with different foam void size, we now compare the previous 50 𝜇m void size case with simulations using smaller (12.5 𝜇m) and larger (100 𝜇m) voids. Except for the void size, all other simulation parameters are kept unchanged. As in the case of 50 𝜇m foam voids, we use the snapshot at 𝑡 = 75 ns for the turbulence analysis. Appropriate positions for the analysis region are selected for each case, such that there is a sufficiently large volume of well-developed turbulence. The dimen- sions of the selected analysis region are the same for all three cases. The time evolution of all three simulations and the selected analysis regions are shown in Fig. B1. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Fig. 9 shows the density PDFs (top panels) and Mach number PDFs (middle and bottom panels) for the three cases with different foam void sizes. Since the M𝑥 and M𝑧 PDFs are almost identical, we omit the M𝑧 PDF in the figure and only show the M𝑥 and M𝑦 PDFs. We apply the LN+PL fit (Eq. 3) to the density distributions, and the Gaussian fit (Eq. 7) to the Mach number PDFs. We find that both the density and the Mach number PDFs become broader with increasing void size. This is because for larger void sizes, there is more space for the density and velocity fluctuations to develop when the shock passes through it. Relations between the post-shock properties and void size have been derived in a companion paper (Davidovits 2022). |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | The density PDFs in all the three cases are consistent with log- normal distributions at lower densities, with a high-density power- law tail that becomes flatter with increasing void size. When the size of the voids is small, the mixing of the two phases (foam and void material) is more efficient. Because of this, the density PDF for the smallest void size case is closest to a LN distribution, with a very steep power-law tail. With increase in the foam void size, the mixing becomes less and less efficient, there are more anisotropies, and the power-law tail flattens. The Mach number PDFs in all the three cases are close to Gaussian distributions, with non-Gaussian features in the form of wings far away from the mean, as discussed in Sec. 3.3. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | The parameters for the LN+PL fit of the density PDFs and for the Gaussian fits of the Mach number PDFs are summarised for all the three void sizes in Table C1. It is observed that the slope of the high-density power-law tail decreases with increasing foam void size, leading to a more prominent PL contribution. The standard deviation of the LN part (𝜎𝑠) increases with increasing void size, giving a more spread out PDF. The 3D Mach number also increases with increasing void size, but remains subsonic in all three cases. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 8. PDFs of the local Mach number in 𝑥, 𝑦, 𝑧 (from left to right) for shock-driven turbulence with 50 𝜇m foam voids. The Mach number PDFs are consistent with Gaussian fits via Eq. (7). The far wings of the PDFs have exponential tails, typical of intermittent fluctuations in the turbulence. However, these tails have no significant influence on the overall standard deviation of the distributions. We also see that the M𝑦 PDF is slightly wider (by ∼ 30%) than the M𝑥 and M𝑧 PDFs, due to 𝑦 being the shock propagation direction. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 9. Comparison between the PDFs obtained for the three different foam void sizes: 12.5 𝜇m (leftmost column), 50 𝜇m (middle column) and 100 𝜇m (rightmost column). The top panels show the PDFs of logarithmic density, 𝑠. The middle and bottom panels show the PDFs of the local Mach numbers along the 𝑥 and 𝑦 directions, respectively. The dashed curves represent LN+PL fits for the density PDFs and Gaussian fits for the Mach number PDFs. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 10. Density–sound speed correlation PDFs for the three different foam voids cases: 12.5 𝜇m, 50 𝜇m, and 100𝜇m (from left to right). Lines of constant polytropic Γ are plotted to guide the eye. Three lines are drawn on each plot, one each at the top and the bottom edge of the PDF, and one passing approximately through the centre. We find that Γ does not change significantly with different foam void size. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Table 2. Comparison of density dispersion, Mach number, polytropic Γ, and driving parameter, in the three different foam void size cases. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Notes. Columns: (1) simulation name, (2) diameter of foam voids, (3) standard deviation of density fluctuations, (4)–(6) standard deviation of the turbulent Mach number in the 𝑥, 𝑦, 𝑧 direction, (7) 3D turbulent Mach number computed via Eq. (8), (8) polytropic index Γ, (9) turbulence driving parameter (Eq. 10). The errors in 𝜎𝜌/(cid:104)𝜌(cid:105), 𝜎M𝑥,𝑦,𝑧 , and M are calculated from the relative errors obtained by changing the analysis volume (see Tab. E1). These errors are propagated analytically to obtain the uncertainties in 𝑏. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | three cases. As expected from the density PDFs, the spread of the correlation PDFs also increases with increasing foam void size. All the three PDFs exhibit a large tail corresponding to regions with high densities and low sound speeds (hence low temperatures). There is clearly a negative correlation between the density and the sound speed for all the three cases. To determine the approximate polytropic Γ of the systems, we plot lines of constant Γ similar to the ones plotted before for the 50 𝜇m void size case. We find that Γ does not vary much with foam void size, with Γ = 0.2 ± 0.1 for the 12.5 𝜇m void size case, and Γ = 0.1±0.1 for the 50 𝜇m and 100 𝜇m void size cases, as discussed before in Sec. 3.2. Thus, the system has an effectively soft equation of state with Γ < 1, irrespective of the foam void size. We calculate the density variance and 3D Mach number directly from the data for each void-size case. The values are listed in Table 2. We can also compare the values obtained directly from the data, with those obtained from the fits (see Tab. C1), and find that both 𝜎𝜌/ (cid:104)𝜌(cid:105) and M agree between data and fits to within 10% for all three void size cases. Using these measurements, we can now move on to the main objective of this study, which is to calculate the driving parameter of shock-driven turbulence. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | This is identical to Eq. (1), but generalised to magnetised turbulence with plasma 𝛽 = 𝑝th/𝑝mag, i.e., the ratio of the thermal pressure to the magnetic pressure (Molina et al. 2012). In the absence of magnetic fields, 𝑝mag → 0 =⇒ 𝛽 → ∞, simplifying Eq. (9) to 𝜎𝜌/ (cid:104)𝜌(cid:105) = 𝑏M, which is exactly Eq. (1). In the following, we will ignore magnetic fields. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | In Eqs. (1) and (9), the parameter 𝑏 is known as the turbulence driv- ing parameter, which depends on the modes induced by the turbulent driving mechanism. As discussed in the introduction, some physical drivers of turbulence excite more solenoidal modes, other drivers excite more compressive modes (Federrath 2018). The driving pa- rameter covers the range 𝑏 ∼ 1/3–1, with 𝑏 ∼ 1/3 corresponding to a purely solenoidal (divergence-free) driving, and 𝑏 ∼ 1 corresponding to a purely compressive (curl-free) driving. If the turbulence driv- ing modes are randomly distributed in all the three directions, then 𝑏 ∼ 0.4, which is known as the ‘natural mixture’ (Federrath et al. 2008, 2010). Thus, 𝑏 (cid:46) 0.4 means that there are more solenoidal modes as compared to compressive modes, whereas 𝑏 (cid:38) 0.4 cor- responds to more compressive modes as compared to solenoidal modes. |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | Figure 11. Relation between turbulent density fluctuations (𝜎𝜌/(cid:104)𝜌(cid:105)) and sonic Mach number (M). For data points with magnetic field information (here only Taurus, G0.253+0.016, and the blue squares (Molina et al. 2012)), the sonic Mach number is modified by the plasma 𝛽 term from Eq. (9), which simplifies to unity in the limit 𝛽 → ∞. The curves represent purely compressive driving (𝑏 = 1, magenta), naturally mixed driving (𝑏 ∼ 0.4, blue) and purely solenoidal driving (𝑏 = 1/3, gold) for isothermal (Γ = 1, dotted) and polytropic (Γ = 0.1, dashed) cases, respectively, following Eq. (10). The magenta shaded area shows the region where 𝑏Γ = 1 for 0 ≤ Γ ≤ 0.3 (this is the range of Γs in this study), again computed based on Eq. (10). The markers with black error-bars represent observational data points. These include three clouds in the spiral arms of the Milky Way: Taurus (Brunt 2010; Kainulainen & Tan 2013; Federrath et al. 2016), GRSMC43.30-0.33 (Ginsburg et al. 2013) and IC5146 (Padoan et al. 1997b), and the ‘Brick’ cloud (G0.253+0.016) in the Central Molecular Zone (Federrath et al. 2016). The cream coloured circles show the six pillars analysed by Menon et al. (2021) in the Carina Nebula. The black cross labelled as ‘WNM’ represents measurements in the Warm Neutral Medium (WNM) by Marchal & Miville-Deschênes (2021). The cyan square represents a measurement for the star-forming region N159E in the Large Magellanic Cloud (Sharda et al. 2021). The gold, blue and magenta symbols are results of various numerical simulations, with the colour corresponding to the driving parameter of the turbulence used in the given simulation. These include results from studies by Federrath & Banerjee (2015) (inverted triangle), Nolan et al. (2015) (triangles), Konstandin et al. (2012b) (stars), Molina et al. (2012) (squares), Price et al. (2011); Price & Federrath (2010) (pentagon) and Federrath et al. (2008, 2010) (diamonds). The red circle shows the main result of our calculation for laser-induced, shock-driven turbulence for the 50 𝜇m void foam. The 12.5 𝜇m void size case is represented by the light green circle, whereas the 100 𝜇m void size case is shown as the light blue circle. Red crosses correspond to the 50 𝜇m void size case, but for different analysis volumes (see Appendix E). The right-facing triangles show simulations obtained for each void size (using the corresponding colour), but for different analysis times (see Appendix F). Irrespective of the void size, analysis volume or time, we find that shock-driven turbulence is close to purely compressive driving (𝑏 ∼ 1). |
Processed_The_driving_mode_of_shock-driven_turbulence.txt | We can immediately verify that for isothermal turbulence (Γ = 1), Eq. (10) reduces to Eq. (1). Thus, Eq. (10) is a generalised form for 𝑏 in gases with polytropic index Γ. |
Subsets and Splits