<?xml-model href='http://www.tei-c.org/release/xml/tei/custom/schema/relaxng/tei_all.rng' schematypens='http://relaxng.org/ns/structure/1.0'?><TEI xmlns="http://www.tei-c.org/ns/1.0">
	<teiHeader>
		<fileDesc>
			<titleStmt><title level='a'>Length and Velocity Scales in Protoplanetary Disk Turbulence</title></titleStmt>
			<publicationStmt>
				<publisher>The Astrophysical Journal</publisher>
				<date>04/26/2024</date>
			</publicationStmt>
			<sourceDesc>
				<bibl> 
					<idno type="par_id">10554145</idno>
					<idno type="doi">10.3847/1538-4357/ad2c89</idno>
					<title level='j'>The Astrophysical Journal</title>
<idno>0004-637X</idno>
<biblScope unit="volume">966</biblScope>
<biblScope unit="issue">1</biblScope>					

					<author>Debanjan Sengupta</author><author>Jeffrey N Cuzzi</author><author>Orkan M Umurhan</author><author>Wladimir Lyra</author>
				</bibl>
			</sourceDesc>
		</fileDesc>
		<profileDesc>
			<abstract><ab><![CDATA[<title>Abstract</title> <p>In the theory of protoplanetary disk turbulence, a widely adopted ansatz, or assumption, is that the turnover frequency of the largest turbulent eddy, Ω<sub><italic>L</italic></sub>, is the local Keplerian frequency Ω<sub><italic>K</italic></sub>. In terms of the standard dimensionless Shakura–Sunyaev<italic>α</italic>parameter that quantifies turbulent viscosity or diffusivity, this assumption leads to characteristic length and velocity scales given respectively by<inline-formula><tex-math><CDATA/></tex-math><math overflow='scroll'><msqrt><mrow><mi>α</mi></mrow></msqrt><mi>H</mi></math><inline-graphic href='apjad2c89ieqn1.gif' type='simple'/></inline-formula>and<inline-formula><tex-math><CDATA/></tex-math><math overflow='scroll'><msqrt><mrow><mi>α</mi></mrow></msqrt><mi>c</mi></math><inline-graphic href='apjad2c89ieqn2.gif' type='simple'/></inline-formula>, in which<italic>H</italic>and<italic>c</italic>are the local gas scale height and sound speed. However, this assumption is not applicable in cases when turbulence is forced numerically or driven by some natural processes such as vertical shear instability. Here, we explore the more general case where Ω<sub><italic>L</italic></sub>≥ Ω<sub><italic>K</italic></sub>and show that, under these conditions, the characteristic length and velocity scales are respectively<inline-formula><tex-math><CDATA/></tex-math><math overflow='scroll'><msqrt><mrow><mi>α</mi><mrow><mo stretchy='true'>/</mo></mrow><mi>R</mi><mo accent='false'>′</mo></mrow></msqrt><mi>H</mi></math><inline-graphic href='apjad2c89ieqn3.gif' type='simple'/></inline-formula>and<inline-formula><tex-math><CDATA/></tex-math><math overflow='scroll'><msqrt><mrow><mi>α</mi><mi>R</mi><mo accent='false'>′</mo></mrow></msqrt><mi>c</mi></math><inline-graphic href='apjad2c89ieqn4.gif' type='simple'/></inline-formula>, where<inline-formula><tex-math><CDATA/></tex-math><math overflow='scroll'><mi>R</mi><mo accent='false'>′</mo><mo>≡</mo><msub><mrow><mi mathvariant='normal'>Ω</mi></mrow><mrow><mi>L</mi></mrow></msub><mrow><mo stretchy='true'>/</mo></mrow><msub><mrow><mi mathvariant='normal'>Ω</mi></mrow><mrow><mi>K</mi></mrow></msub></math><inline-graphic href='apjad2c89ieqn5.gif' type='simple'/></inline-formula>is twice the Rossby number. It follows that<inline-formula><tex-math><CDATA/></tex-math><math overflow='scroll'><mi>α</mi><mo>=</mo><mover accent='true'><mrow><mi>α</mi></mrow><mrow><mo>˜</mo></mrow></mover><mrow><mo stretchy='true'>/</mo></mrow><mi>R</mi><mo accent='false'>′</mo></math><inline-graphic href='apjad2c89ieqn6.gif' type='simple'/></inline-formula>, where<inline-formula><tex-math><CDATA/></tex-math><math overflow='scroll'><msqrt><mrow><mover accent='true'><mrow><mi>α</mi></mrow><mrow><mo>˜</mo></mrow></mover></mrow></msqrt><mi>c</mi></math><inline-graphic href='apjad2c89ieqn7.gif' type='simple'/></inline-formula>is the root-mean-square average of the turbulent velocities. Properly allowing for this effect naturally explains the reduced particle scale heights produced in shearing box simulations of particles in forced turbulence, and it may help with interpreting recent edge-on disk observations; more general implications for observations are also presented. For<inline-formula><tex-math><CDATA/></tex-math><math overflow='scroll'><mi>R</mi><mo accent='false'>′</mo><mo>></mo><mn>1</mn></math><inline-graphic href='apjad2c89ieqn8.gif' type='simple'/></inline-formula>, the effective particle Stokes numbers are increased, which has implications for particle collision dynamics and growth, as well as for planetesimal formation.</p>]]></ab></abstract>
		</profileDesc>
	</teiHeader>
	<text><body xmlns="http://www.tei-c.org/ns/1.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink">
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="1.">Introduction</head><p>Protoplanetary disks are increasingly regarded as being moderately turbulent, under a handful of magnetohydrodynamic <ref type="bibr">(Balbus &amp; Hawley 1991;</ref><ref type="bibr">Turner et al. 2014;</ref><ref type="bibr">Lesur et al. 2023)</ref> or purely hydrodynamic instabilities <ref type="bibr">(Marcus et al. 2013;</ref><ref type="bibr">Nelson et al. 2013;</ref><ref type="bibr">Lyra 2014)</ref>. In many applications, such as global models of disk evolution under turbulent viscosity <ref type="bibr">(Estrada et al. 2016;</ref><ref type="bibr">Sengupta et al. 2022)</ref>, it is impractical to use more than a simple parameterization of the viscous and diffusive effects of turbulence, and by far the most popular one is the socalled "&#945;-model" <ref type="bibr">(Shakura &amp; Sunyaev 1973;</ref><ref type="bibr">Shakura et al. 1978)</ref>. In this closure model, the spatial and temporal evolution of turbulent protoplanetary gas and particle disks is determined by the large-scale effective viscosity and diffusivity of the gas, as well as the particles in it. For many purposes, it is adequate (and indeed necessary, given our limited understanding) to simplify the complex effects of real turbulence into a scalar viscosity &#957; and diffusivity D. The basic scaling or mixing length approximation for turbulent diffusivity and viscosity is (</p><p>, where L and V L are respectively the characteristic length and velocity scales of the turbulence. In this case, L can be identified with the energy injection scale or the energy-containing large spatial scale, and V L is the velocity at that scale. <ref type="bibr">Shakura &amp; Sunyaev (1973)</ref> introduced the widely used mixing-length-like closure model &#957; &#8801; &#945;cH to incorporate our ignorance about the turbulence into the single parameter &#945;, where c is the gas sound speed and H is the gas vertical scale height. In that classic paper,<ref type="foot">foot_0</ref> it was suggested briefly that L &#8764; H and thus V L &#8764; &#945;c, indeed with &#945; often taken to be of order unity <ref type="bibr">(Cameron 1978;</ref><ref type="bibr">Lin &amp; Bodenheimer 1982;</ref><ref type="bibr">Weidenschilling 1984;</ref><ref type="bibr">Morfill 1985;</ref><ref type="bibr">Weidenschilling &amp; Cuzzi 1993)</ref>, given that early works were more focused on accretion disks around compact objects where trans-sonic turbulence used to be widely believed to exist. Moreover, to study the properties of solid particles (hereafter simply called "particles") in the disk as they grow by collisional sticking and evolve radially by drift and diffusion (e.g., <ref type="bibr">Estrada et al. 2016</ref>, P. R. Estrada &amp; J. N. Cuzzi 2022, in preparation, P. R. <ref type="bibr">Estrada et al. 2022, in preparation)</ref>, it has become necessary to separate the velocity and lengthscale properties of the turbulence that influence the particles <ref type="bibr">(Cuzzi et al. 2001)</ref>.</p><p>A common assumption in such studies is that the turnover time of the largest energy-containing eddy of the nebula turbulence equals the local orbital time <ref type="bibr">(Shakura &amp; Sunyaev 1973)</ref>. <ref type="bibr">Shakura et al. (1978, Appendix I and page 184)</ref> worked through an energy-dissipation argument that implied &#937; L = &#937; K , and a logical chain resulted that was similar to the one given in Section 2.1 below (see Appendix A for more details). The &#937; L = &#937; K assumption had been made earlier by <ref type="bibr">Safronov (1972)</ref>, and there has been some numerical support for this assumption under certain conditions <ref type="bibr">(Coleman et al. 1992</ref>).</p><p>As simulations of particle collisions and growth in turbulence have become more detailed, it has also become important to reasonably estimate particle velocities given V L <ref type="bibr">(V&#246;lk et al. 1980)</ref>. <ref type="bibr">Weidenschilling (1997)</ref> adopted the scaling of <ref type="bibr">Shakura et al. (1978)</ref>, and <ref type="bibr">Cuzzi et al. (2001)</ref> independently reproduced the scaling of <ref type="bibr">Shakura et al. (1978)</ref> in the form most often used today, as described below in Section 2.1. <ref type="bibr">Dubrulle (1992)</ref> generalized the discussion to eddy frequencies &#937; L &gt; &#937; K and &#937; L &lt; &#937; K , but concluded that &#937; L &#8764; &#937; K for selfgenerated nebula turbulence. <ref type="bibr">Dubrulle (1992)</ref> even proposed a scaling of L relevant to these conditions, but without connecting it to &#945; in any clear way. The idea that large-eddy frequencies in turbulence could significantly exceed the orbit frequency has also been discussed in the context of midplane turbulence generated by a settled particle layer <ref type="bibr">(Cuzzi et al. 1993;</ref><ref type="bibr">Sekiya 1998</ref>), but has never been explored using direct numerical simulations.</p><p>&#937; L = &#937; K seems to be a sensible assumption, given that the disk is rotating, and thus any radial or azimuthal velocity deviation will be subsequently dominated by the Coriolis force, leading to a Keplerian turnover time (e.g., <ref type="bibr">Cuzzi et al. 2001)</ref>. While reasonable for the magnetorotational instability (MRI; <ref type="bibr">Balbus &amp; Hawley 1991)</ref>, this result does not appear to be generally valid for turbulence generated by, e.g., vertical shear instability, where <ref type="bibr">Stoll &amp; Kley (2016)</ref> measure eddy lifetimes much shorter than Keplerian, of the order of 0.1 K 1 . Sengupta &amp; Umurhan (2023) find similarly short eddy lifetimes for the turbulence generated by the settled particle layer at the disk midplane. Incidentally, even for the MRI, <ref type="bibr">Johansen et al. (2006)</ref> find eddy lifetimes to be shorter than Keplerian in the case of forcing by a sufficiently strong externally imposed vertical magnetic field.</p><p>Taken all in all, these results suggest that the tendency of the Coriolis force to impose a Keplerian turnover time will be in competition with other external forcing agents acting on the system. If these other forcing agents exert greater strength or amplitude, then the Coriolis force becomes less influential, potentially resulting in the turnover time of the largest eddies deviating from Keplerian. It is this case we investigate in this paper.</p><p>In this contribution, we numerically assess the properties of externally forced turbulence under conditions of a rotating, sheared nebula flow, under the general condition &#937; L &#937; K . The utility of the work is to allow for initial conditions in forced simulations of turbulence with particles that satisfy simultaneously the critical nebula parameters &#945; and &#937; L . Without actually knowing what the appropriate turbulent frequency &#937; L is, the relevant particle Stokes number St &#8801; t s &#937; may be incorrectly defined-here, t s is the aerodynamic stopping time of the particles. This appears to be the case in many prior simulations of particle behavior in nebula turbulence for which the particle properties are defined by an &#937; K -based Stokes number <ref type="bibr">(Birnstiel et al. 2010;</ref><ref type="bibr">Estrada et al. 2016;</ref><ref type="bibr">Sengupta et al. 2019;</ref><ref type="bibr">Gole et al. 2020</ref>), which we define as St K = t s &#937; K , and in which the possibility of &#937; L &gt; &#937; K was not considered. In this paper, we will introduce and utilize an injection-scale eddy Stokes number defined hereafter by St L = t s &#937; L .</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.">Generalizing the Components of &#945;</head></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.1.">The Current Ansatz</head><p>We first revisit the original ansatz, which is generally assumed to be true in the literature in cases where the turbulence parameter is calculated from numerical simulations. We shall call this &#732; , the turbulence intensity when &#937; L = &#937; K . <ref type="bibr">Cuzzi et al. (2001)</ref> restated the separation of the L and V L components of &#957; based on the ansatz that the largest-scale eddies in the nebula would have eddy frequency &#937; L = &#937; K . Situations where &#937; L &lt; &#937; K generally correspond to slow currents rather than fluctuating eddies, such as gently varying zonal flows.</p><p>These structures do not survive longer periods, due to the Coriolis effect introduced by rotation <ref type="bibr">(Zahn 1989;</ref><ref type="bibr">Cuzzi et al. 2001)</ref>.</p><p>When the ansatz &#937; L = &#937; K is assumed true, &#732; relates to the length and velocity scales in a specific way by starting with the usual formulation,</p><p>where c is the local sound speed and we have used H &#8801; c/&#937; K , where H is the local gas scale height. In this case, c and H are considered as the characteristic velocity and length scales of the system. Now, if the ansatz &#937; L = &#937; K is valid, then Equation (1) can be rewritten as</p><p>Combining Equations (1) and</p><p>, this result is widely used to calculate particle relative velocities in growth models (e.g., <ref type="bibr">Ormel &amp; Cuzzi 2007;</ref><ref type="bibr">Birnstiel et al. 2010</ref><ref type="bibr">Birnstiel et al. , 2016;;</ref><ref type="bibr">Estrada et al. 2016;</ref><ref type="bibr">Sengupta et al. 2019)</ref>. The estimate &#732; is also used as a measure of turbulent intensity in forced turbulence models, i.e., based on root-mean-square (rms) velocities <ref type="bibr">(Gole et al. 2020</ref>; see Section 3.2 for more discussion).</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.2.">The General Case</head><p>The above usual ansatz does not address cases where &#937; L &gt; &#937; K , which may be the case in realistic turbulence. In order to address cases where the traditional assumption &#937; L = &#937; K is invalid (as occurs in at least the highly relevant VSI regime; <ref type="bibr">Nelson et al. 2013;</ref><ref type="bibr">Stoll &amp; Kley 2016)</ref>, the analysis of Section 2.1 must be generalized. For the general situation, we denote the turbulence (diffusion) parameter as &#945; (rather than &#732; ). Starting once again with Equation (1), we write the frequency of the largest eddy as &#937; L &#8764; V L /L, where V L is the velocity of the largest (energy-containing) eddy and L is the corresponding length scale. We define</p><p>, where R is twice the traditional Rossby number Ro (Cushman-Roisin 2011).</p><p>With this, Equation (2) now reads as</p><p>As before, setting this result equal to Equation (1) in terms of &#945;, we get</p><p>Using Equation (5), the velocity V L for the largest eddy can be written as</p><p>which is equivalent to the expression Equation (3). Thus, while L and V L change form, their product does not, and we recover</p><p>L From Equation (6), &#945; can be related to &#732; as</p><p>In the limit of &#937; L = &#937; K , R 1 , and the general case reduces to &#732; . Also, substituting the above expression into Equation (6) demonstrates its equivalence to Equation (3).</p><p>It is noteworthy that a similarity can be drawn between the expression for the diffusive &#945; found in Equation (8) and the one given by <ref type="bibr">Youdin &amp; Lithwick (2007;</ref><ref type="bibr"/> see their Section 3), i.e., V g 2 eddy , where V g is the rms of the gas fluctuation velocities and &#964; eddy is the eddy turnover time. Given that V g 2 is dominated by the largest energy-containing scale, R 1 eddy if &#964; K &#8764; 1/&#937; K &#8764; 1 is assumed.<ref type="foot">foot_2</ref> </p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.3.">Why Is &#937; L So Pivotal in Nebular Turbulence?</head><p>In the protoplanetary disk, the gas scale height H = c/&#937; K is a natural length scale arising from hydrostatic balance. In a Keplerian disk, it is also the largest scale on which organized motions like large vortices remain subsonic-so if the turbulence reaches the scales of H (either through that scale being unstable or through inverse cascade), then H will shape the turbulence.</p><p>The most significant length scale for turbulence is the scale associated with the largest eddy, the correlations of which govern major aspects of nebula physics from transport properties to particle dynamics, and it has been found that it can be orders of magnitude smaller than H for many plausible turbulence mechanisms studied to date, even when R 1 (Figure <ref type="figure">1</ref>, left).</p><p>When the more general case is considered (&#937; L &#937; K ), the largest turbulent eddy length scale shifts toward smaller scales (Figure <ref type="figure">1</ref>, right), and under such conditions, it is this scale that dominates the particle transport and diffusion properties in the nebula. Scales longer than this may be important for zonal flows or coherent structures, such as long lived vortices, but they are not relevant to fluctuating turbulent gas dynamics.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.">Simulations with Forced Turbulence</head><p>For the purpose of demonstrating the theoretical concept constructed in Section 2.2, we executed a series of simulations where turbulence is forced externally in a shearing box with domain size 0.2H in the radial and azimuthal directions and 0.4H in the vertical. All the simulations presented in this section are gas-only (Lagrangian particles are introduced in Section 4), and we solve for the equations for mass and momentum conservation under isothermal condition with vertical gas stratification in a shearing box setup:</p><p>where u g is the gas velocity, P is the gas pressure, &#961; g is the gas densityand &#937; K is the local Keplerian frequency. The term &#937; 2 z in Equation (10) represents the vertical gravity responsible for the gas stratification. The third term on the left-hand side of Equation (10) represents the Coriolis force arising from the rotation, and the last term of the right-hand side of the same equation is the shear. The resolution used in all the simulations in this section is (N x &#215; N y &#215; N z ) &#8801; 256 &#215; 256 &#215; 512. We have shear, rotation, and vertical stratification included in the simulations, and we use the PENCIL CODE<ref type="foot">foot_3</ref> (Pencil Code Collaboration et al. 2021) to solve the continuity and momentum equations. The forcing scheme is the same one used in <ref type="bibr">Sengupta &amp; Umurhan (2023)</ref>. The external forcing is controlled by a forcing parameter f 0 that sets the amount of energy injected in the system at a previously chosen wavenumber. For the gasonly simulations presented in this section, we have chosen two forcing wavenumbers k f = 3 and 6, and used f 0 = [0.003, 0.005, 0.01, 0.03, 0.05, 0.1, 0.3] for each k f , totaling 14 simulations. The details of the forcing scheme are presented in Appendix F. We have also used sixth-order hyperdiffusion and hyperviscosity in the simulations, allowing the fields to dissipate their energy near the smallest scales while preserving the power spectra at the larger scales (for details see Sengupta &amp; Umurhan 2023). We calculate &#945; and &#732; directly from our code results</p><p>where v r and v are the local velocity fluctuations relative to the mean flow, and V rms is a spatial and temporal average of the fluctuating velocity amplitudes. It is generally the case that V L &#8776; V rms (see Appendix B for details).</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.1.">The Importance of Turbulent Kinetic Energy Spectra</head><p>Here, we illustrate how relevant turbulence parameters are determined directly from the turbulent kinetic energy spectrum E(k) as a function of spatial wavenumber k (see Figure <ref type="figure">1</ref>). The first step to test the general case through our numerical simulations is to estimate &#937; L . The velocity v k of an arbitrary eddy with wavenumber k can be written as</p><p>k where E(k)dk is the energy per unit mass contained in eddies with wavenumber ranging between k and k + dk, where E(k) has units of H K 3 2 . Then, the turnover time for eddies with wavenumber k is</p><p>where &#8467; &#8764; 1/k is the eddy length scale <ref type="bibr">(Tennekes &amp; Lumley 1972;</ref><ref type="bibr">Davidson 2004</ref>) and the corresponding eddy frequency is simply &#937; e &#8801; 1/&#964;. We note that &#964; here is the same as the nonlinear timescale or the timescale associated with the scale-to-scale energy transfer. Using Equations (13-14), the eddy turnover frequency can be expressed as</p><p>For the energetically dominant eddy with length scale L, Equation (15) reads as</p><p>where k f is the energy injection scale or the forcing scale in our simulations. Correspondingly, <ref type="figure">2</ref>(a) illustrates these relationships for two energy spectra with f 0 = 0.05 with forcing at k f = 3 and 6. One should take note of the positions of the largest energycontaining scales, which determine &#937; L for the two cases.</p><p>Appendix C shows how these spectral expressions also lead to &#732;k H R f (Equation ( <ref type="formula">8</ref>)).</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.2.">Connection between &#945; and Tangential Reynolds Stresses</head><p>Following <ref type="bibr">Shakura &amp; Sunyaev (1973)</ref>, the coefficient of turbulent dynamical viscosity &#956; is defined by the mixing length form ( ) LV , 18</p><p>L and using the correspondence between Reynolds stress W rf and viscosity,</p><p>W in which r is the cylindrical radius coordinate and where Keplerian flow is assumed for the azimuthal velocity v f , i.e., v f /r = &#937; K &#8764; r -3/2 9 Re-expressing the stress-viscosity relationship Equation (19) in terms of variables we use here for the box-including the use of Equation ( <ref type="formula">7</ref>) and the assumption that the density &#961; is constant on the scales of interest-we find</p><p>where the above has been expressed in terms of the negative of the specific tangential stress r r W . Similarly to V rms , r is calculated from the spatiotemporal average of the correlations between the radial and azimuthal velocity fluctuations (see Appendix B). Equation (20) simply states that turbulent kinematic viscosity can be estimated from the specific with the tangential Reynolds stress r 2 3</p><p>and R , calculated based on the KE spectrum and Equation ( <ref type="formula">16</ref>). The relationship shows a power law with exponent 0.91 &#177; 0.05 for k f = 6, close to our expectation of unity, and 0.71 &#177; 0.03 for k f = 3 (see Section 3.2 for details). 9 In isotropic turbulence, the ensemble-averaged correlations of off-diagonal fluctuating velocity components are equal to zero <ref type="bibr">(Davidson 2004</ref>). However, in anisotropic systems like those involving Keplerian shear, off-diagonal velocity correlations like W rf are nonzero in general, and can lead to transport. This explains the original motivation behind the mixing length model Equation (19).</p><p>tangential stresses via the simple form c 2 3 r 2</p><p>(Equation ( <ref type="formula">11</ref>)). However, the quantity r (also commonly written as v v r</p><p>) is sensitive to the correlations of the velocity fluctuations in addition to their magnitudes, while V rms is not, corroborating the analysis of <ref type="bibr">Balbus &amp; Papaloizou (1999)</ref>. Indeed, in several cases, turbulent Reynolds stresses have been numerically calculated from actual velocity correlations rather than only their magnitudes (e.g., as done for the VSI turbulence, <ref type="bibr">Nelson et al. 2013;</ref><ref type="bibr">Stoll &amp; Kley 2014)</ref>. Such stresses can be directly associated with &#945; (also see <ref type="bibr">Youdin &amp; Lithwick 2007, Appendix B)</ref>.</p><p>This important distinction is accounted for by our R scaling term as shown in Figure <ref type="figure">2</ref>. Using the k f = 6 simulation results in Figure <ref type="figure">2</ref> . For k f = 3, however, we get b = 0.71 &#177; 0.03, which is about 30% less than the expected value.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.3.">&#937; L in Forced Turbulence as a Function of f 0</head><p>In Figure <ref type="figure">3</ref>(a), we show the variation of R L K as a function of f 0 for two different injection wavenumbers k f = 3 (blue star) and 6 (orange diamond). For a particular forcing wavenumber k f , &#937; L -and hence R -increases with the forcing amplitude. Increasing the injection energy at the same wavenumber and associated length scale results in a faster rotation for the largest (energy-containing) eddy and a corresponding increase in the turnover frequency &#937; L . It is also evident from Figure <ref type="figure">3</ref>(a) that, for a fixed value of f 0 , R increases when energy is injected at a higher wavenumber and smaller scale, and as for fixed energy injection, smaller eddies have a smaller turnover time.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.4.">&#945; in Forced Turbulence as a Function of f 0</head><p>In Figure <ref type="figure">3(b)</ref>, a calibration of &#732;R as a function of f 0 is shown for forcing at two different wavenumbers: k f = 3 (blue star) and 6 (orange diamond). It is evident from the figure that, for a fixed k f , &#945; increases with the increasing energy injection. Diffusion and viscosity are primarily governed by the largest energy-containing scales, so increasing energy at those scales will result in an increased diffusion and hence a higher &#945;. However, for a fixed f 0 , &#945; decreases when energy is injected at a higher wavenumber (see k f = 3 versus k f = 6 plots in Figure <ref type="figure">3</ref>(b), as the length scale for the largest energycontaining eddy decreases when a higher wavenumber is chosen.</p><p>Interestingly, as &#732; is not a function of R and it depends on the rms of the fluctuating velocities throughout the inertial range, which measures the total energy injected into the system, variation in &#732; for a fixed f 0 is minimal with the injection wavenumber k f as shown in Figure <ref type="figure">3(c</ref>). This result is also Calibration for R as a function of forcing-energy injection, given in terms of the forcing parameter f 0 for two different injection wavenumbers: k f = 3 (blue star) and k f = 6 (orange diamond). The straight lines are the power-law fits for each data set. For fixed energy injection, R increases with k f . (b): Calibration for &#945; (corrected by R ) vs. f 0 for k f = 3 (blue star) and k f = 6 (orange diamond). (c): Variation of &#732; with f 0 for k f = 3 and 6. It should be noted that the values of &#732; are almost the same for two different injection scales. With &#732;V c rms 2 2 over the full inertial range, &#732; is expected to reflect the total energy injected into the system, irrespective of the injection scale. (d): Variation of &#945; with R for k f = 3 and 6. A power-law fit for y = ax b gives b = 0.98 &#177; 0.07 for k f = 3 and 1.02 &#177; 0.02 for k f = 6 (see Section 3.5).</p><p>consistent with our findings in Equations ( <ref type="formula">6</ref>) and (8) that V L 2 remains reasonably constant with the forcing wavenumber and R implicitly conveys the length-scale information in the realization of &#945;.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.5.">Scaling between &#945; and R</head><p>The present model, developed on the basis of the general case &#937; L &gt; &#937; K , can be tested by analyzing the scaling between &#945; and R , using the parameter f o . Consider that the total energy injected varies with f 0 with some power-law index p as ( )</p><p>and V L &#8733; E 1/2 , as in Equation (13); thus, we can write E f p 1 2 0</p><p>2 . So, we expect that, for a particular value of k f , R is constant and the slope of the plot for log versus R log will be unity. In Figure <ref type="figure">3(d)</ref>, we show the variation of log versus R log for k f = 3 (blue star) and 6 (orange diamond), with the calibrations shown by dotted lines. When a power-law fitting in the form y = ax b is done for the respective data sets, we get b = 0.98 &#177; 0.07 for k f = 3 and b = 1.02 &#177; 0.02 for k f = 6,</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.6.">Physical Processes Leading to Different R</head><p>The model developed in this article is relevant to several kinds of nebula turbulence, such as that driven by vertical shear instability (VSI; <ref type="bibr">Nelson et al. 2013;</ref><ref type="bibr">Flock et al. 2020)</ref> or convective overstability (COV; Lyra 2014). For example, <ref type="bibr">Stoll &amp; Kley (2014)</ref> found that their VSI simulations had &#937; L significantly higher than &#937; K (eddy turnover time &#964; eddy &lt; 1). More recently, the energy spectra presented for global VSI simulations in <ref type="bibr">Flock et al. (2020)</ref> show that energy injection happens around k &#8764; 100 and the stratification-induced inverse cascade sets the largest energy-containing scale around k &#8764; 20, suggesting that the value of R may significantly exceed unity. More recently, Sengupta &amp; Umurhan (2023) found Ro &gt; 1 in particle-laden disk midplane when turbulence is driven by Kelvin-Helmholtz and symmetric instability <ref type="bibr">(Stamper &amp; Taylor 2017)</ref>. If the disk turbulence is primarily triggered and sustained by such hydrodynamic processes, the method of approximating &#945; as ( &#732;) v c 2 2 will be an overestimation, whereas values for &#945; from the analyses of edge-on disks will probably be more consistent with the actual level of disk turbulence (see Section 5.1).</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="4.">Numerical Validation</head><p>There are several situations in which the value of R becomes relevant-particle scale height being one of them. In order to investigate the validity of &#937; L &gt; &#937; K and the effect of R , we have performed numerical tests where inertial (St &#8800;0) particles have been introduced in the simulation box with rotation and vertical stratification, with the turbulence being generated and maintained by external forcing. The particles are then allowed to settle under the influence of vertical gravity until a balance between the settling and turbulent diffusion is reached. We then investigated the scale heights obtained in the simulations, the main objective being to test whether inertial particles react to the nebula turbulence following (&#945;, St L ) or ( &#732;) , St K .</p><p>In a protoplanetary disk, a steady state between particle settling toward the disk midplane and turbulent diffusion in the vertical direction sets the particle scale height <ref type="bibr">(Cuzzi et al. 1993;</ref><ref type="bibr">Dubrulle et al. 1995;</ref><ref type="bibr">Youdin &amp; Lithwick 2007</ref>; also see Appendix E):</p><p>In order to test the validity of the theory and insights developed in Sections 2 and 3, we now introduce Lagrangian particles in our simulations and compare their scale heights h with Equation (21). We use a monodisperse particle distribution with St K = 0.06 and an initial vertically averaged particleto-gas mass ratio Z = 0.1%. Thus, while the particle backreaction is formally in play, the particle local density never gets large enough for it to be significant. The particles are coupled to the gas through aerodynamic drag by their stopping time t s (Section 1)-the time required for a particle to lose all its momentum to the differentially moving gas. Here, the Lagrangian equations of motion for a particle's vector position (x p ) and velocity (u p ) can be written as</p><p>The second term on the right-hand side in Equation ( <ref type="formula">23</ref>) represents momentum exchange via aerodynamic drag that depends upon the gas velocity evaluated at the particle's position. In the gas momentum equation Equation (10), we must introduce a term representing the equal and opposite momentum transfer. This modifies the Equation (10) to</p><p>where &#961; p is the local volume mass density of the particle phase.</p><p>In Figure <ref type="figure">4</ref>, we present the results of our numerical experiment where turbulence is forced with f 0 = 0.05 at k f = 3 (blue dashed curve) and 8 (orange solid curve) with k &#8764; 45 and 120, respectively, when normalized by domain size, producing &#732;8 10 3 in a simulation box, including vertical gravity, rotation, and density stratification. The vertically integrated solid-to-gas mass ratio in the simulation is kept low (Z = 0.1%) in order to minimize the effects of particle mass loading. The scale height for the particles when calculated using &#732; and St K is h = 0.34H, significantly larger than what the simulations show. In Figure <ref type="figure">4</ref>(a), the obtained z 2 is shown to be different for the two runs, even though all the parameters are the same and the runs differ only by the forcing wavenumber, clearly indicating that R plays an important role in setting &#945; and hence the vertical distributions of the particles. The values of R for k f = 3 and 8 are approximately 5 and 12, producing St L &#8764; 0.3 and 0.7. respectively. When h is calculated using Equation ( <ref type="formula">21</ref>) with these values, we get h &#8764; 0.14 for k f = 3 and &#8764;0.08 for k f = 6. In Figure <ref type="figure">4</ref>(b), the particle density for the two simulations is shown with a Gaussian fit in the form y ae z 2 2 2 on the top. The fitted values for |&#963;| (the particle scale height of the distribution) for the two wavenumbers come out to be 0.13 and 0.08, closely matching the calculated scale heights from Equation (21). So, it is evident from our numerical experiment that the vertical diffusion of particles under nebula turbulence should be effectively parameterized by &#945;, not &#732; , and without the scaling by R as in Equations ( <ref type="formula">8</ref>) and (B2), the reported &#732; will be an overestimation of the true diffusivity of the particles in the domain.</p><p>We calculate R using equation Equation ( <ref type="formula">16</ref>) directly from the KE spectra produced from the simulation data. In Figure <ref type="figure">4</ref>(c), the kinetic energy spectra for the particle-gas simulation with k f = 8 is shown. The associated eddy turnover times &#937; e for all wavenumbers are shown in Figure <ref type="figure">4(d)</ref>. The vertical and horizontal dashed black lines in the figure indicate the forcing wavenumber (length scale L) and the associated R L K . We note that the estimated particle scale heights h using &#945; and St L match nicely with the values for |&#963;| obtained from the Gaussian vertical particle distribution. However, when compared with z 2 , the estimate agrees well with k f = 8 but differs by &#8764;35% for k f = 3. This difference is due to the effect of a relatively small vertical domain used in the simulation. h gets smaller either when the level of turbulence is low or when the energy is injected at a higher wavenumber (smaller scale) such that the velocity correlations at the largest energycontaining scales are small. This is the reason why z 2 agrees well with the actual scale height for k f = 8. We also expect that z 2 and |&#963;| will be close for k f = 3, when using a smaller value for f 0 (lower level of turbulence) where particlelayer thickness is small, or if a larger vertical domain is used. So, we can infer that z 2 is a good representation of particle scale height only when the particles are settled sufficiently away from the vertically periodic boundary of the simulation domain.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="5.">Discussion</head></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="5.1.">Determination of &#945; from Observations</head><p>There are several situations in which the value of R is relevant-estimating the level of nebula turbulence from disk observations is one of them. Observations of protoplanetary disks reveal that the systems are turbulent to varying degrees, and the observed levels differ by the mode of observation as well. 2 on the top. The fitting gives the values of the parameter |&#963;| = 0.13 &#177; 0.0007 for k f = 3 and 0.08 &#177; 0.0001 for k f = 8. The particle distribution is Gaussian with the midplane particle-to-gas mass ratio reaching &#8764;0.009 for k f = 3 and &#8764;0.012 for k f = 8, implying that the effect of particle mass loading is negligible. (c): Kinetic energy spectra for the particle-gas simulation forced at k f = 8 (k &#8764; 120 when normalized by domain size). A curve fit for the slope of the inertial range gives -1.36 &#177; 0.2, shallower than -5/3 even in the presence of inertial particles. (d): A plot for eddy turnover time &#937; e /&#937; K for all the wavenumbers. The location for the largest energy-containing wavenumber k f and the corresponding value of R are noted by the dashed straight lines.</p><p>One way of assessing &#945; for disks through observations is by analyzing the turbulent line broadening in order to obtain &#948;v/c and estimating &#945; as &#8764;&#948;v 2 /c 2 using velocity information only <ref type="bibr">(Flaherty et al. 2017</ref><ref type="bibr">(Flaherty et al. , 2018;;</ref><ref type="bibr">Teague et al. 2018;</ref><ref type="bibr">Flaherty et al. 2020)</ref>. Generally, this method gives an upper bound of &#945; -for example, <ref type="bibr">Flaherty et al. (2017)</ref> found &#948;v/c &#8764; 0.01 for HD 163296 and &#8764;0.3 at &#8764;1-2 scale heights above the midplane for DM Tau (also see <ref type="bibr">Lesur et al. 2023</ref> for a discussion). Similar analysis by <ref type="bibr">Teague et al. (2018)</ref> on TW-Hya using the J = 7 -6, 5 -4, and 3-2 transitions of CS imaged at &#8764;0 5 spatial resolution estimated &#948;v/c &#61576; 0.1. As these observations improve, it will be critical to remember that they are fundamentally V rms 2 -based and thus measure &#732; , not &#945;. On the other hand, the observed vertical thicknesses of edgeon disks with ALMA and the VLA have been used to infer values of &#945;/St using Equation (21). Then, values of &#945; are inferred using assumptions of St based on these observations and nebula-growth models <ref type="bibr">(Doi &amp; Kataoka 2021;</ref><ref type="bibr">Villenave et al. 2022;</ref><ref type="bibr">Doi &amp; Kataoka 2023;</ref><ref type="bibr">Villenave et al. 2023)</ref>. In this way, <ref type="bibr">Villenave et al. (2022)</ref> estimated &#945; &#8764; 10 -5 for Oph 163131. For HD 163296, which is an older Class II disk, estimates for &#945; vary based on which ring system is being analyzed: for the inner, more vertically extended ring, the turbulence is vigorous, with &#945; = 6 &#215; 10 -2 (Doi &amp; Kataoka 2023), while the corresponding values for the outer flatter ring are systematically lower but with little agreement between published values with &#945; 6 &#215; 10 -3 (Doi &amp; Kataoka 2023) or &#945; &#8776; 4.5 &#215; 10 -3 <ref type="bibr">(Pizzati et al. 2023)</ref>.</p><p>We underscore that particle-layer thicknesses are sensitive to &#945;, not &#732; (Section 4 and Figure <ref type="figure">4</ref>). It is interesting to note that, while uncertainties remain significant (especially in St), the values of &#945; estimated from the analyses of particle scale heights seem to be consistently smaller compared to the ones obtained from the turbulent line broadening, and this is just what would be expected from the difference between values of particle height derived from &#732; and &#945; if R 1 . It is possible to combine the independent and distinct observational constraints discussed above (on h/H from particle-layer thicknesses and V rms from spectral line observations) with other observed constraints on particle size and gas disk surface mass density (e.g., <ref type="bibr">Carrasco-Gonz&#225;lez et al. 2019;</ref><ref type="bibr">Mac&#237;as et al. 2021;</ref><ref type="bibr">Doi &amp; Kataoka 2023)</ref> to actually constrain the critical value of R and thus the nature of the underlying processes causing disk turbulence in the first place. We sketch the logic by which this may be done in Appendix H.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="5.2.">R and Sc: Effects on Global Particle Redistribution</head><p>Here, we comment on how R may affect particle radial distributions. As noted in Section 1, the ratio &#957;/D &#8801; Sc (the Schmidt number) is usually taken as unity; indeed, there has been some confusion in the literature as to the definition of Sc itself (see <ref type="bibr">Jacquet et al. 2012 and</ref><ref type="bibr">Estrada &amp;</ref><ref type="bibr">Cuzzi 2016, their Appendix B)</ref>. As shown in Section 3.2, this quantity is the same as the ratio v v V R 1 r rms 2 . To our knowledge, <ref type="bibr">Hughes &amp; Armitage (2010</ref><ref type="bibr">, 2012)</ref> represent the most complete (or only) study of the role of variable Sc in global redistribution of particles undergoing growth, drift, and diffusion. They only varied Sc from 0.5 to 2.0, and over this range found the effects minor. Unsurprisingly, (what we find to be) the more realistic case, Sc&lt;1, favors diffusion over drift or advection, leading to a slightly greater retention of small particles in outer disks after longer times. To the degree that future studies of different kinds of nebula turbulence might be found to favor larger values of R , the implications for particle radial transport might need to be reassessed. This would include studies where the radial structure of a particle band is used to infer the properties of the local particle size and/or &#957; <ref type="bibr">(Dullemond et al. 2018)</ref>.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="5.3.">Effects on Collision Velocities and Particle Growth</head><p>Following <ref type="bibr">V&#246;lk et al. (1980)</ref>, <ref type="bibr">Ormel &amp; Cuzzi (2007)</ref> derived closed-form solutions for the collision velocities for particles in the turbulent nebula for three different regimes: (1) the tightly coupled particles when t s &lt; t &#951; , where t &#951; is the stopping time at the Kolmogorov scale (their Equation (27); (2) the inertial range regime where t &#951; &lt; t s &lt; t L (their Equation (28)); and (3) the heavy particle limit when t s &gt; t L (their Equation (29)). The response of a particle to turbulence is determined by St L , which is a factor of R larger than the St K assumed by previous studies along these lines (e.g., <ref type="bibr">Birnstiel et al. 2010;</ref><ref type="bibr">Zsom et al. 2010;</ref><ref type="bibr">Estrada et al. 2016;</ref><ref type="bibr">Sengupta et al. 2019</ref><ref type="bibr">Sengupta et al. , 2022))</ref>. This has different implications in the three different particle size ranges, as noted below. Regarding the equations below, recall that <ref type="bibr">Ormel &amp; Cuzzi (2007)</ref> defined St with the meaning of St L , while at the time implicitly associating it with St K as has been done by all dust growth models to date.</p><p>The velocities for the tightly coupled small particles in the sub-Kolmogorov regime are mostly dominated by Brownian motion <ref type="bibr">(Sengupta et al. 2019</ref>, see their Figure <ref type="figure">2</ref>) and are unlikely to change significantly. However, with the R effect included, the collision velocities &#916;V 12 between particles with large-eddy Stokes numbers St 1L and St 2L (Equation ( <ref type="formula">27</ref>) Ormel &amp; Cuzzi 2007) for the tight coupling limit become</p><p>Similarly, in the most important range (2) above, Equation (28) of <ref type="bibr">Ormel &amp; Cuzzi (2007)</ref> shows the collision velocity, which can be written with the new scaling included as</p><p>where the term &#923; is only weakly dependent on St L for St L = 1 <ref type="bibr">(Ormel &amp; Cuzzi 2007)</ref>. Thus, Equation (26), along with Equation (6), shows that the collision velocity is increased by a factor of R 1 2 for a given &#945;, and thus particles reach the bouncing or fragmentation barriers at physical sizes that are smaller by a related factor.</p><p>In the heavy particle limit, a higher St L with R 1 reduces collision velocities, allowing for larger particles. The modified expression from <ref type="bibr">Ormel &amp; Cuzzi (2007, Equation (29)</ref>) would read as</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="5.4.">Planetesimal Formation</head><p>Our results have implications for the process of planetesimal formation, an outstanding problem in planet formation theory.</p><p>For example, we have seen that turbulence with R 1 can significantly reduce the particle scale height (Figure <ref type="figure">4</ref>) and increase &#242; = &#961; p /&#961; g at the midplane, for particles of some given St K or nominal physical size, relative to estimates based on the rms velocity-based turbulence parameter &#732; . This is primarily because the (more appropriate) value of &#945; will decrease, and because St L for the particles will increase <ref type="bibr">(however,</ref><ref type="bibr">Equation (21)</ref> shows the St L effect to be relatively minor when St L = 1). Planetesimal formation by either streaming instability (SI; <ref type="bibr">Youdin &amp; Goodman 2005;</ref><ref type="bibr">Johansen et al. 2007)</ref> or turbulent concentration (TC; <ref type="bibr">Cuzzi et al. 2010;</ref><ref type="bibr">Hartlep &amp; Cuzzi 2020</ref>) is aided when the near-midplane &#242; is larger; even though this effect may only increase &#242; by a factor of order unity, this can be important.</p><p>TC has a more subtle dependence on St L than SI, but comparison of planetesimal formation models with observations of, e.g., observed chondrule or aggregate sizes will be affected. <ref type="bibr">Hartlep &amp; Cuzzi (2020)</ref> find reasonable sizes and abundances for their resulting planetesimals if the initial pebbles are centimeter-and decimeter-sized, implicitly assuming R</p><p>1 . However, if R 10 , for example, the same outcome would be achieved with mm-cm sized particles.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="6.">Conclusions</head><p>The main objective of the work presented here is to investigate how the particle behavior in a turbulent fluid under nebular conditions is altered with varying large-eddy frequency &#937; L &gt; &#937; K . Below, we present a summary of our findings: , up to an undetermined correction factor, supposedly of order unity. Here, we quantify this correction factor to be R , which can be significantly larger than unity. In this context, it is important to note that the calculation of &#945; from the correlations of velocity fluctuations always incorporates the R dependence and is itself a proper measure of diffusivity or viscosity. 3. The effective particle Stokes number is best defined by the large-eddy frequency &#937; L rather than the orbital frequency</p><p>St L is the parameter that governs the particle transport properties and should be used for turbulent concentration simulations when comparing with the cascade model predictions from, e.g., <ref type="bibr">Hartlep &amp; Cuzzi (2020)</ref>. 4. When R 1 (&#937; L &gt; &#937; K ), the particle-layer scale heights h in our 3D simulations show excellent matching with the predictions obtained from (&#945;, St L ) instead of from ( &#732;) , St K . 5. The self-consistent generation of turbulence by hydrodynamic processes, such as vertical shear instability <ref type="bibr">(Nelson et al. 2013</ref>) consistently shows R 1 <ref type="bibr">(Stoll &amp; Kley 2016;</ref><ref type="bibr">Flock et al. 2020)</ref>. Our findings imply that particle scale heights inferred from turbulent broadening ( &#732;) v c 2 2 will be overestimates, while those obtained from direct observations of particle scale heights in edgeon disks or actual calculated Reynolds stresses are diffusion-based and probably more appropriate. 6. The collision velocity of particles, under the condition &#937; L &gt; &#937; K , is scaled by different positive powers of R , suggesting that, for R 1 turbulence, the maximum particle sizes should be smaller compared to R 1 turbulence, all other things being equal. 7. The reduction of particle-layer scale height and the enhancement of midplane &#242; = &#961; p /&#961; g for high R turbulence in the nebula should favor planetesimal formation by both streaming instability <ref type="bibr">(Youdin &amp; Goodman 2005;</ref><ref type="bibr">Johansen et al. 2007</ref>) and turbulent Concentration <ref type="bibr">(Hartlep et al. 2017;</ref><ref type="bibr">Hartlep &amp; Cuzzi 2020</ref>). 8. Calibration of a numerical code for the purpose of particle-gas simulations under external forcing should be done separately for different particle sizes, as the mass loading changes the values for &#945; and R for same injection energy. Furthermore, for external forcing in the presence of vertical stratification, choosing a forcing scale away from the domain size is a safer choice and is advisable (preferably k f = 4 or higher).</p><p>In conclusion, we would like to emphasize the importance of turbulent kinetic energy spectra in the context of numerical studies of &#945;, and in particular, directly determining the value of &#937; L based on these spectra. separate L and V L . The ansatz is thus underlain by the physically plausible assumption that the turbulent kinetic energy budget of the large eddies (Equation (A1)) is supplied by the local release of gravitational energy by inwardly drifting mass, which is in turn caused by viscous outward transport of angular momentum (Lynden-Bell &amp; Pringle 1974). The gravitational energy released is locally transformed to mechanical turbulent motions by viscosity acting across the local Keplerian gradient (Equation <ref type="formula">A2</ref>). In all this, of course, "viscosity" is generalized to a turbulent viscosity. This turbulent kinetic energy budget is transformed to heat on the smallest scales of turbulence and radiated away locally. On this basis, <ref type="bibr">Shakura et al. (1978)</ref> then used a logic parallel to that of our Section 2.1, which they expressed as</p><p>, where is a Mach number. In their Section 3, they had already found that 2 2 , leading directly, as in our Section 2.1, to V c L and L H . Extending this "energetics" logic to physical conditions where other nebula structures are driven by non-Keplerian shear and have different local energy-dissipation rates, it is not hard to see how the large, energy-containing eddies can have &#937; L &#8800; &#937; K . For example, <ref type="bibr">Cuzzi et al. (1993)</ref> studied midplane turbulence driven by a settled particle layer, which imposes a vertical velocity shear considerably larger than the radial Keplerian shear, recently demonstrated explicitly in <ref type="bibr">Sengupta &amp; Umurhan (2023</ref><ref type="bibr">). et al. (1993)</ref> argued that the largest eddies would have large R and adjusted their treatment of St accordingly. Given the forcing implicit in the shearing vertical sheets of VSI <ref type="bibr">(Nelson et al. 2013;</ref><ref type="bibr">Stoll &amp; Kley 2014)</ref>, a similar situation there seems inevitable. Other turbulent instabilities should be considered from this standpoint, as well.</p><p>where for practical intents and purposes &#961; is assumed to be constant. Expressing W rf in terms of a negative quantity ensures that gas accretion in the traditional sense of the effect occurs toward the star.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head>Appendix C Mixing Length Phenomenology</head><p>In light of our introductory remarks in Section 2.1, we seek to relate turbulent viscosity D and &#945; to the KE spectrum of the gas. Diffusion in a turbulent medium can be phenomenologically assessed by the dominant length (L) and velocity scales (V L ) associated with the turbulent dynamics, which usually occurs at the start of the inertial range in 3D isotropic turbulence. According to standard mixing length theory (e.g., <ref type="bibr">Knobloch 1978)</ref>, when the turbulence is isotropic, the diffusion becomes a scalar and can be written as</p><p>Considering that a particular length scale &#8467; may reasonably be taken to contain two counter-rotating eddies of size &#8467;/2, Equation (C1) reads</p><p>Using Equations (6) and (16), Equation (C2) above can be further manipulated into</p><p>The approach leading to Equation (C3) captures both the form of the original mixing length ansatz and the Reynolds stress approach (Section 2.1) as well as the spectral approach to identifying V L , for which we generally find that indeed &#7804; V c L 2 rms 2 2 . Using the expression for &#937; L found in Equation ( <ref type="formula">16</ref>), and &#732;( )</p><p>which in turn leads to to an equivalent restatement of Equation (C3) as</p><p>which demonstrates the equivalence introduced in Section 3.1 between &#945; and &#732;R Hk f . Thus, in practice, one can estimate &#945; using either of these preceding formulae by identifying the wavenumber k f corresponding to the peak value of ( ) ( )</p><p>is the vertical component of solar gravity at height z. The particle diffusion coefficient <ref type="bibr">Youdin &amp; Lithwick 2007)</ref>. We note that the settling velocity depends on St K and the diffusion coefficient on St L . One approximates &#961; p &#8764; &#961; po and (&#8706;&#961; p /&#8706;z) &#8764; &#961; po /h, so</p><p>The St L 2 term from the diffusion coefficient plays relatively little role in the process, given that St L = 1 for most realistic cases <ref type="bibr">(Carballido et al. 2011;</ref><ref type="bibr">Estrada et al. 2016)</ref>. This is also true in cases where there is a balance between radial diffusion and drift <ref type="bibr">(Dullemond et al. 2018;</ref><ref type="bibr">Umurhan et al. 2020</ref>) because radial drift, like vertical drift, is determined by St K .</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head>Appendix F The Forcing Scheme</head><p>In order to drive turbulence in the simulation box by external forcing, we shall use the same prescription as <ref type="bibr">Brandenburg (2001)</ref>, <ref type="bibr">Haugen &amp; Brandenburg (2004)</ref>, and Sengupta &amp; Umurhan (2023) and drive turbulence with a random, nonhelical, and delta-correlated forcing function f(x, t), written as</p><p>Here, k(t) = (k x , k y , k z ) is a time-dependent wavevector, and f (t) with | | is the random phase. On a purely dimensional argument, the normalization factor can be written as ( ) f c kc t 0 1 2 , where c is the sound speed set as 1 in the code unit, &#948;t is the time step, k = |k| is the wavenumber, and f 0 is a dimensionless factor we call the forcing amplitude. In the code, f 0 is the knob we adjust in order to control the forcing and achieve a desired value of &#945; (also see below). We choose to force the system at some k = k f , in which case, at each step, a randomly chosen possible wavenumber within a narrow band of k f -0.5 &lt; |k| &lt; k f + 0.5 is forced. The forcing is executed with the eigenfunctions of the curl operator</p><p>e 1 1 . F2 k 2 2 2 2 where we have used the relationships &#732;R found in Equation (8) and the definition R St St L K . Equation (H1) can be viewed as a cubic equation for St L , in which the turbulent Stokes number is a function of the observables h/H and &#732;V c rms 2 2 (see Appendix H for details): ( ) St St , H2 L L 3 where &#732;&#732;&#732;( ) H h H V h c V h . H3 K 2 2 2 rms 2 2 2 rms 2 2 2 Cubic equations are awkward to solve, but one can obtain simple expressions in two limiting regimes to illustrate the use of the technique. It is important to note that St L itself may be determined directly from j, which is composed entirely of the observable quantities, h, V rms , and &#937; K . First, consider the highly likely regime j = 1 or R St St 1 L K . In this regime, St St L L 3 , so from Equation (H2), &#732;( ) R H h St St K L</p><p>2 . We can now solve directly for R in terms of observables, making use of the standard approximation for the small-particle (Epstein) regime that St K = t s &#937; K = 2r&#961; p /&#931; g :</p><p>that is, R is constrained by a combination of potentially observable parameters including &#732; , and once R is determined, &#945; itself can be immediately calculated.</p><p>The alternate limiting regime is j ?</p><p>1 or R St St 1 L K . Here, St St L L 3 , so from Equation (H2), R St St K L 1 3 and &#9115; &#9117; &#9118; &#9120; &#9115; &#9117; &#9118; &#9120; &#732;&#732;( ) R H h r H h St 2 . H5 K g p 1 3 2 3 1 3 2 3</p><p>It will always be evident from the observed values of h/H (ALMA or VLA) and &#732; (spectral line broadening) which regime is of interest. Again, once R is determined, &#945; itself can be immediately calculated. The usually cited combined parameter loosely referred to as &#945;/St is thus seen to be merely an intermediate step in a complete determination of the properties of the system.</p></div><note xmlns="http://www.tei-c.org/ns/1.0" place="foot" n="6" xml:id="foot_0"><p>We note that the first description and application of mixing length theory to disks may be found in<ref type="bibr">Prendergast &amp; Burbidge (1968)</ref>, but an explicit mathematical prescription was not given.</p></note>
			<note xmlns="http://www.tei-c.org/ns/1.0" place="foot" xml:id="foot_1"><p>The Astrophysical Journal, 966:90 (13pp), 2024 May 1 Sengupta et al.</p></note>
			<note xmlns="http://www.tei-c.org/ns/1.0" place="foot" n="7" xml:id="foot_2"><p>The interested reader is referred to<ref type="bibr">Carballido et al. (2011)</ref> for a validation and discussion of the<ref type="bibr">Youdin &amp; Lithwick (2007)</ref> expression for particle diffusion.</p></note>
			<note xmlns="http://www.tei-c.org/ns/1.0" place="foot" n="8" xml:id="foot_3"><p>http://pencil-code.nordita.org/</p></note>
			<note xmlns="http://www.tei-c.org/ns/1.0" place="foot" n="10" xml:id="foot_4"><p>We note that Gole et al. (2020) write &#732; as &#945; drive .</p></note>
		</body>
		</text>
</TEI>
