Beyond the α \alpha model: scaling the wind-driven accretion rate in protoplanetary disks using systematic non-ideal magnetohydrodynamical simulations Haruhi Enomoto, \altemailmark \orcid 0000-0001-6390-8700 Shoji Mori, \orcid 0000-0002-7002-939X and Satoshi Okuzumi \altemailmark \orcid 0000-0002-1886-0880 Abstract Magnetically driven mass accretion in protoplanetary disks plays a crucial role in understanding disk evolution and planet formation. However, the α \alpha prescription lacks a direct connection to the underlying physical processes, and no systematic scaling law yet exists that describes the accretion rate as a function of disk quantities. While local shearing-box simulations offer a powerful approach to analyzing the accretion structure at low computational cost, they suffer from a fundamental problem: the toroidal magnetic field generated by Keplerian shear accumulates within the computational domain, disrupting a field-line geometry consistent with global wind-driven accretion. In this study, we introduce the super-box-scale diffusion (SBD) scheme into non-ideal MHD shearing-box simulations. The SBD scheme continuously damps the horizontally averaged horizontal magnetic field components, thereby mitigating this problem and maintaining the field-line symmetry required for global wind-driven accretion for more than 500 orbital periods. Comparison with self-similar solutions supports the SBD method’s quantitative usefulness, with both the vertical structure and the plasma-beta dependence of the accretion rate agreeing to within 23–28%. We then conduct a parameter survey of 46 cases using a magnetic diffusivity table constructed from ionization equilibrium calculations, covering a wide range of disk radius, surface density, magnetic field strength, and dust-to-gas ratio. We find that the surface field-line pitch and the mass accretion rate each follow a power-law scaling with the midplane plasma beta, an effective ambipolar Elsasser number that characterizes the ionization structure of the magnetically active layer, and the normalized thickness of that layer. The scaling relations reproduce the numerical results to within a factor of 2–3 across the explored parameter space and, in most cases, to within a factor of 2. The present scaling relations provide a framework for predicting the mass accretion rate from local disk physical quantities without invoking an α \alpha parameter. 2026 \Received \Accepted \KeyWords accretion,accretion disks —magnetohydrodynamics(MHD) — magnetic fields — methods:numerical —protoplanetary disks — planets and satellites: formation 1 Introduction Accretion in protoplanetary disks plays a crucial role in determining the final masses and compositions of planets. Systematic surveys of nearby star-forming regions have been characterizing disk accretion rates across a wide range of stellar masses and ages ( 34 ) . However, significant uncertainties remain in current theoretical models of accretion disks. The classical viscous disk model ( 33 ) has been widely used as a phenomenological prescription of disk accretion. However, this model parameterizes the mass transport efficiency through a single dimensionless viscosity parameter \alpha ( 46 ) whose connection to the underlying physical processes is ambiguous. In particular, it is widely assumed that \alpha is spatially and temporally constant, but there is no guarantee that this assumption holds true ( 16 , e.g.,) . Identifying the physical mechanism of mass transport is crucial for a better understanding of disk evolution. The leading candidate for the mechanism driving mass transport in protoplanetary disks is angular momentum transport by magnetic fields (e.g., 12 ; 27 ; 9 ; 24 ). Three-dimensional magnetohydrodynamics (MHD) simulations (e.g., 3 ; 30 ; 21 ; 31 ; 26 ) that incorporate non-ideal MHD effects (e.g., 18 ; 44 ; 65 ; 60 ; 29 ) have shown that magnetic winds serve as the dominant angular momentum transport mechanism in weakly ionized regions where magnetic turbulence is suppressed (for a review, see, e.g., 29 ). In this picture, angular momentum is removed vertically by magnetically driven outflows launched from the disk surface, thereby allowing the disk gas to accrete. The magnetically driven mass accretion rate in such regions depends strongly on both the ionization fraction and the net vertical magnetic field strength (e.g., 47 ; 11 ; 8 ; 20 ; 31 ; 15 ). However, no empirical model yet exists that describes how the accretion rate depends on all these disk parameters. Developing such a model requires systematic MHD simulations of disk accretion under realistic ionization fraction distributions. Previous MHD simulations of protoplanetary disk accretion can be classified into radially global and local approaches. The radially global approach ( 54 ; 11 ; 51 ; 52 ; 55 ; 20 ; 56 , e.g.,) has the advantage of naturally treating global magnetic field-line geometry and radial magnetic flux transport. However, this approach generally requires high computational costs and is therefore not well-suited for extensive parameter surveys. Recently, 31 proposed a self-similar approach that reduces the global problem to a one-dimensional one by assuming radial self-similarity. This method can capture global accretion properties at low computational cost; however, it is not applicable to realistic ionization fraction distributions, which are not necessarily self-similar in radius. The radially local approach ( 24 ; 50 ; 53 ; 40 ; 3 ; 48 ; 43 , e.g.) employs a shearing box, in which the dynamical equations in a local Cartesian domain centered at a given orbital radius are solved in the presence of Keplerian shear. Local simulations are a powerful tool for computing the vertical structure at any given radius with high resolution and low computational cost. However, because the shearing box assumes uniform shear and periodic boundary conditions, it does not fully capture the radial gradient of the toroidal field produced by Keplerian rotation, nor the associated radial diffusion. As a result, the toroidal magnetic flux generated by shear tends to accumulate within the local domain (e.g., 61 ). This excessive accumulation of toroidal magnetic flux makes it difficult to maintain a field-line configuration consistent with the system’s global geometry, resulting in an unphysical accretion structure that is perfectly anti-symmetric with respect to the disk midplane, thereby yielding zero net accretion ( 3 ; 30 , e.g.,) . In this study, we present a systematic study of magnetically driven accretion in weakly ionized protoplanetary disks by using a refined local shearing-box approach that mitigates the accumulation of horizontal magnetic fields. Our method employs the super-box-scale diffusion (SBD) scheme ( 19 ) , which regularly removes the horizontally averaged component of the horizontal magnetic fields from the local box. We demonstrate that this SBD scheme enables us to compute vertical profiles of gas accretion that are consistent with global self-similar solutions ( 31 ) . Using this approach, we derive an empirical law for the magnetically driven accretion rate, including the effects of Ohmic and ambipolar diffusion, as a function of the disk’s vertical ionization structure and the net vertical magnetic flux. This paper is organized as follows. In section 2 , we describe the numerical methods and the formulation of the SBD scheme. In section 3 , we examine the numerical stability of the SBD scheme and then validate the scheme against self-similar solutions. In section 4 , we present the results of an extensive parameter survey using a magnetic diffusivity table based on ionization equilibrium calculations, systematically investigate the effects of disk radius, surface density, and dust-to-gas ratio on the accretion rate, and propose the resulting scaling laws. Finally, in section 5 , we discuss implications of our findings and future directions. 2 Method In this study, we numerically solve the conservative non-ideal MHD equations in the local shearing-box approximation (subsection 2.1 ). To mimic the global diffusive relaxation of the mean magnetic field, we add an SBD term to the induction equation (subsection 2.2 ). We adopt two prescriptions for the magnetic diffusivities: an analytic parametric model and a tabulated model based on ionization equilibrium calculations (subsection 2.3 ). Subsection 2.4 summarizes the numerical scheme and run parameters, and subsection 2.5 defines the mass accretion rate and explains our treatment of the field-line geometry. 2.1 Governing equations We introduce a local Cartesian coordinate system (x,y,z)=(r-r_{0},\,r_{0}(\phi-\Omega_{\rm K}t),\,z) centered at an orbital radius r_{0} and corotating with the local Keplerian angular velocity \Omega_{\rm K} . Here, r is the cylindrical radius, \phi is the azimuthal angle, z is the vertical distance from the disk midplane, and t is time. Throughout this paper, we identify the x - and y - directions with the r - and \phi - directions, respectively, unless otherwise noted. For example, the radial component of the magnetic field is denoted by either B_{x} or B_{r} , depending on the context. For clarity, we hereafter omit the subscript 0 from r_{0} . The continuity equation is \frac{\partial\rho}{\partial t}+\nabla\cdot(\rho\bm{v})=0, (1) where \rho is the gas density. The momentum equation is \displaystyle\frac{\partial(\rho\bm{v})}{\partial t}+\nabla\cdot\left[\rho\bm{v}\bm{v}-\frac{\bm{B}\bm{B}}{4\pi}+\left(P+\frac{B^{2}}{8\pi}\right)\bm{I}\right]= (2) \displaystyle-2\rho{\Omega}_{\rm K}\hat{\bm{z}}\times\bm{v}+3\rho\Omega_{\rm K}^{2}x\,\hat{\bm{x}}-\rho\,\Omega_{\rm K}^{2}z\,\hat{\bm{z}}, where \bm{v} is the gas velocity vector, \bm{B} is the magnetic field vector, P is the gas pressure, \bm{I} is the unit tensor, and \hat{\bm{x}} and \hat{\bm{z}} are the unit vectors in the x - and z -directions, respectively. We adopt the local isothermal approximation and relate P to \rho as P=c_{\rm s}^{2}\rho, (3) where c_{\rm s} is the isothermal sound speed. The isothermal scale height is given by H\equiv c_{\rm s}/\Omega_{\rm K} . We include Ohmic and ambipolar diffusion, while neglecting the Hall effect. The magnetic field evolves according to the induction equation, \frac{\partial\bm{B}}{\partial t}=\nabla\times\left(\bm{v}\times\bm{B}-\eta_{\rm O}\bm{J}-\eta_{\rm A}\bm{J}_{\perp}\right)+\left.\frac{\partial\bm{B}}{\partial t}\right|_{\rm SBD}, (4) where the current density \bm{J} is defined by Ampère’s law, \bm{J}=\frac{c}{4\pi}\nabla\times\bm{B}. (5) The last term, \left.\partial\bm{B}/\partial t\right|_{\rm SBD} , represents the SBD contribution, which damps the horizontally averaged magnetic field. Its explicit form is given in subsection 2.2 . The initial density profile is given by the Gaussian profile of an isothermal hydrostatic atmosphere, \rho(z)=\rho_{0}\exp\!\left(-\frac{z^{2}}{2H^{2}}\right), (6) where \rho_{0} is the initial density at the midplane ( z=0 ). We initialize the magnetic field with a spatially uniform vertical component B_{z0} only, with no radial or azimuthal component. Defining the initial plasma beta as \beta_{0} , the isothermal equation of state (equation ( 3 )) gives \beta_{0}\equiv\frac{8\pi\rho_{0}c_{s}^{2}}{B_{z0}^{2}}, (7) and the plasma beta \beta is defined as \beta\equiv\frac{8\pi P}{B^{2}}. (8) 2.2 SBD prescription The local shearing-box approximation cannot treat global radial diffusion of magnetic flux self-consistently. As a result, the toroidal magnetic field persistently accumulates within the shearing box and can eventually lead to unphysical field configurations ( 3 ; 30 , see, e.g.,) . In reality, any magnetic field that accumulates locally can diffuse away over a global diffusion timescale. The SBD prescription employed by 19 enables us to mimic this large-scale relaxation in local simulations. The SBD prescription decays the horizontally averaged horizontal magnetic fields, \langle B_{i}\rangle_{xy}(z,t) ( i=x,y ), at a constant rate. The decay term is given by \left.\frac{\partial\langle B_{i}\rangle_{xy}}{\partial t}\right|_{\rm SBD}=-\,\frac{\langle B_{i}\rangle_{xy}}{\tau_{\rm diff}(z)}\qquad(i=x,y), (9) where \tau_{\rm diff}(z) is the global magnetic diffusion timescale. Approximating the global variation scale of the magnetic field as \sim r and defining the effective magnetic diffusivity as \bar{\eta}\equiv\eta_{\rm A}+\eta_{\rm O} , the global diffusion timescale can be estimated as \tau_{\rm eff}\sim r^{2}/\bar{\eta} . We therefore take \tau_{\rm diff} to be \tau_{\rm diff}(z)=\frac{r^{2}}{C\bar{\eta}(z)}, (10) where C is a dimensionless coefficient. The corresponding SBD term can then be rewritten as \left.\frac{\partial\langle B_{i}\rangle_{xy}}{\partial t}\right|_{\rm SBD}=-\frac{C\bar{\eta}(z)}{r^{2}}\langle B_{i}\rangle_{xy}\qquad(i=x,y). (11) In the numerical implementation, C serves as a parameter that controls the efficiency of the global field relaxation. Unless otherwise stated, we adopt C=\pi^{2} throughout this paper. This choice corresponds to approximating the large-scale spatial gradient of the magnetic field by the wavelength of the longest Fourier mode (equation (B3) of 19 ). The SBD prescription does not violate the divergence-free condition \nabla\!\cdot\!\bm{B}=0 as it only subtracts the uniform components of the horizontal fields. 2.3 Magnetic diffusivities Ohmic diffusion and ambipolar diffusion are characterized by the diffusivities \eta_{\rm O} and \eta_{\rm A} , respectively. The relative importance of magnetic diffusion compared to magnetic induction can be quantified by the Ohmic and ambipolar Elsasser numbers, \Lambda_{\rm O}\equiv\frac{v_{\rm A}^{2}}{\eta_{\rm O}\Omega_{\rm K}}, (12) {\rm Am}\equiv\frac{v_{\rm A}^{2}}{\eta_{\rm A}\Omega_{\rm K}}, (13) where v_{\rm A}=|\bm{B}|/\sqrt{4\pi\rho} is the Alfvén speed. Ohmic and ambipolar diffusion become important when \Lambda_{\rm O}\lesssim 1 and {\rm Am}\lesssim 1 , respectively (e.g., 44 ; 2 ). The magnetic Reynolds number, {\rm Rm}\equiv\frac{c_{\rm s}^{2}}{\eta_{\rm O}\Omega_{\rm K}}=\left(\frac{c_{\rm s}}{v_{\rm A}}\right)^{2}\Lambda_{\rm O}, (14) also serves as a useful diagnostic for Ohmic diffusion (e.g., 17 ). In this study, we employ two prescriptions for the magnetic diffusivities: an analytic parametric model used by 31 and a tabulated model based on ionization equilibrium calculations. These prescriptions are described below. 2.3.1 Parametric diffusivity model In the parametric model, we prescribe {\rm Rm} and {\rm Am} as functions of z , following 31 : \displaystyle{\rm Rm}(z) \displaystyle={\rm Rm}_{0}\exp\!\left[\left(\frac{z}{3H}\right)^{4}\right]\left(\frac{\rho(z)}{\rho(0)}\right)^{-1}, (15) \displaystyle{\rm Am}(z) \displaystyle={\rm Am}_{0}\exp\!\left[\left(\frac{z}{3H}\right)^{4}\right] where {\rm Rm}_{0} and {\rm Am}_{0} are the midplane values. The corresponding diffusivities \eta_{\rm O}={v_{\rm A}^{2}}/({\Lambda_{\rm O}\Omega_{\rm K}}) and \eta_{\rm A}={v_{\rm A}^{2}}/({{\rm Am}\Omega_{\rm K}}) increase gradually from the midplane toward the disk surface, mimicking the resistivity structure in a protoplanetary disk with ionized surfaces. The upper panel of figure 1 shows the vertical profiles of \Lambda_{\rm O} and {\rm Am} for (\beta_{0},{\rm Rm}_{0},{\rm Am}_{0})=(10^{5},1,1) , where the density profile \rho(z) follows equation ( 6 ). Figure 1: Vertical profiles of the Elsasser numbers \Lambda_{\rm O} (solid) and {\rm Am} (dashed) as a function of height z/H . The dotted line marks {\rm Rm},{\rm Am}=1 . and the lower panel shows the result for the tabulated diffusivity model (subsubsection 2.3.2 ) with r=1~{\rm au} , \Sigma=10^{3}~{\rm g~cm^{-2}} , \beta_{0}=10^{4} , f_{\rm dg}=10^{-5} , and T=110~{\rm K} . Both panels use the same axis scales. Alt text: Two line graphs arranged vertically, sharing the same logarithmic vertical axis from 10 to the minus 6 to 10 to the 6 and horizontal axis from minus 8 to 8 scale height. In each panel, the solid line traces the Ohmic Elsasser number, and the dashed line traces the ambipolar Elsasser number Am as functions of height. 2.3.2 Tabulated diffusivity model The tabulated diffusivity model uses the magnetic diffusivity table adopted by 35 . This table is constructed based on ionization equilibrium calculations (see subsection 2.3 of 37 ) that account for ionization by cosmic rays, stellar X-rays, and radioactive nuclide decay. Assuming local ionization equilibrium, we compute the number densities of charged particles, including dust-grain charge states, and then evaluate \eta_{\rm O} and \eta_{\rm A} . The table is constructed for a fixed dust grain size of 0.1\,\mu{\rm m} . At fixed grain size, varying the dust-to-gas mass ratio f_{\rm dg} approximately changes the total dust surface area, which primarily controls the recombination rate in ionization equilibrium. For each simulation run, we specify the dust-to-gas mass ratio f_{\rm dg} and the gas temperature T , and use the corresponding table to evaluate the diffusivity coefficients. We also include the ionization of C and S by FUV radiation in addition to the tabulated magnetic diffusivities ( 42 ; see 3 for details). To avoid excessively small timesteps due to strong magnetic diffusion, we impose an upper limit of 100c_{\rm s}H on both \eta_{\rm O} and \eta_{\rm A} . This diffusivity cap is not applied to the SBD term. We confirmed that even increasing the cap value by a factor of 10 does not significantly change the quasi-steady structure or the mass accretion rate. Using the tabulated diffusivity model, we run a total of 46 simulations covering r=\{1,3,10,30\}\,{\rm au} , \Sigma=\{10^{2},10^{3},10^{4}\}\,{\rm g~cm^{-2}} , \beta_{0}=\{3.2\times 10^{3},10^{4},3.2\times 10^{4},10^{5},3.2\times 10^{5},10^{6}\} , and f_{\rm dg}=\{10^{-5},10^{-4},10^{-3}\} . The input parameters for these runs are listed in table 1 (appendix A ). The lower panel of figure 1 shows the vertical profiles of \Lambda_{\rm O} and {\rm Am} from the tabulated diffusivity model with r=1~{\rm au} , \Sigma=10^{3}~{\rm g~cm^{-2}} , \beta_{0}=10^{4} , f_{\rm dg}=10^{-5} , and T=110\,{\rm K} . 2.4 Numerical scheme and run parameters We solve the non-ideal MHD equations in the local shearing-box approximation using Athena ( 49 ) . We compute MHD fluxes with the Harten–Lax–van Leer–discontinuities Riemann solver and use a second-order Godunov scheme. The constrained transport method maintains the divergence-free constraint \nabla\!\cdot\!\bm{B}=0 . We use a second-order Runge–Kutta scheme for time integration, with the Courant–Friedrichs–Lewy number set to 0.3 . The diffusion terms are updated using super-time-stepping for numerical efficiency and stability in regions with high diffusivities. The computational domain covers x\in[-0.5,\,0.5]H , y\in[-1,\,1]H , and z\in[-8,\,8]H , with a uniform grid of (N_{x},\,N_{y},\,N_{z})=(16,\,16,\,512) cells. We apply shearing-periodic boundary conditions in the x -direction, periodic boundary conditions in the y -direction, and outflow boundary conditions in the z -direction, where inflow is prohibited. We add small random perturbations to the initial density and velocity fields to seed instabilities. To avoid numerical instabilities caused by excessive magnetization in low-density regions, we impose a density floor determined by \beta_{0} . This floor serves to stabilize the calculation during the initial transient phase, and we confirmed that it is not reached after each run has entered a quasi-steady state. Each simulation is run for 500 orbital periods to confirm that the disk’s vertical structure reaches a quasi-steady state. We adopt c_{s} , \Omega_{\rm K}^{-1} , and \rho_{0} as the units of velocity, time, and density, respectively. The unit of length is the scale height H , and the unit of magnetic field is B_{\rm u}\equiv\sqrt{4\pi\rho_{0}}\,c_{s} . Hereafter, we explicitly indicate the units for clarity. 2.5 Accretion rate and field-line geometry The goal of this study is to model the mass accretion rate. Following equation (15) of 65 , we define the mass accretion rate at cylindrical radius r as \dot{M}\equiv-2\pi r\int_{-z_{\rm surf}}^{+z_{\rm surf}}\rho v_{r}\,dz, (16) where z_{\rm surf} denotes the height of the disk surface, defined by the condition that the magnetic pressure equals the gas pressure, i.e., \beta(z_{\rm surf})=1. (17) Under the assumptions of axisymmetry and steady state, vertical integration of the angular momentum equation gives \dot{M} in terms of the magnetic field (equation (17) of 65 ): \dot{M}\approx\frac{1}{r\Omega_{K}}\frac{\partial}{\partial r}\left[r^{2}\int_{-z_{\rm surf}}^{+z_{\rm surf}}\left(-B_{r}B_{\phi}\right)\,dz\right]-\frac{r}{\Omega_{K}}\left[B_{z}B_{\phi}\right]_{-z_{\rm surf}}^{+z_{\rm surf}}. (18) On the right-hand side of the above equation, the first and second terms represent the contributions from radial angular momentum transport by the magnetic stress within the disk and from vertical angular momentum transport by the Maxwell stress \pm(B_{z}B_{\phi}/4\pi)_{\pm z_{\rm surf}} acting on the upper and lower surfaces, respectively. In wind-driven accretion, magnetic stresses acting on the upper and lower disk surfaces extract angular momentum from the disk. Because the outward normal vectors on the two surfaces point in opposite directions, the surface stress \pm(B_{z}B_{\phi}/4\pi)_{\pm z_{\rm surf}} contributes to angular momentum extraction in the same sense on both sides. The second term on the right-hand side of equation ( 18 ) therefore represents the net vertical extraction of angular momentum through the disk surface and drives accretion. This field-line configuration has been confirmed in global simulations and self-similar solutions (e.g., 12 ; 31 ). We refer to this field-line configuration as the physical field-line geometry ( 3 , figure. 9) . In contrast, when B_{\phi} has an even-function symmetry about the midplane, the contributions from the upper and lower surfaces cancel each other, and no net vertical angular momentum transport arises. We refer to this configuration as the unphysical field-line geometry . By omitting the direction toward the central star and neglecting curvature terms, the standard shearing-box approximation loses the global geometric constraint that field lines should incline in the same radial direction on both sides of the disk. As a result, both the physical and unphysical field-line geometries can arise as numerical solutions. Indeed, 3 reported that both configurations can appear in local simulations, and that the physical field-line geometry is difficult to maintain for long periods in the shearing box, so that the unphysical field-line geometry tends to be selected. In the next section, we show that the SBD scheme stabilizes the physical field-line geometry. 3 Validation of the SBD prescription In this section, we validate the SBD scheme using the parametric diffusivity model (section 2.3.1 ), which is the same prescription adopted by 31 and thus enables a direct comparison with their self-similar solutions. In subsection 3.1 , we demonstrate that SBD maintains the physical field-line geometry over long periods. In subsection 3.2 , we compare the resulting vertical structure and mass accretion rate with the self-similar solutions. 3.1 Time evolution with and without SBD Figure 2: Space–time diagram of the toroidal field B_{\phi} for the parametric diffusivity model with (\beta_{0},{\rm Rm}_{0},{\rm Am}_{0})=(10^{5},1,1) . The upper and lower panels show the results with and without SBD, respectively. The vertical white lines mark the time at which the field-line geometry transitions from the physical to the unphysical configuration. Alt text: Two two-dimensional color maps arranged vertically. The horizontal axis shows time from 0 to 500 orbits and the vertical axis shows height from minus 8 to 8 scale heights. The color indicates the toroidal field ranging from minus 0.05 to 0.05 in units of Bu. Figure 3: Magnetic field structure at 75 and 200 orbits for the parametric diffusivity model with (\beta_{0},{\rm Rm}_{0},{\rm Am}_{0})=(10^{5},1,1) . Streamlines show the poloidal field lines of (B_{x},B_{z}) , and the color indicates the toroidal field B_{\phi} . Although the radial extent of the computational domain is -0.5Hz_{\rm surf} , where \beta<1 , the magnetic field accelerates winds, with the vertical velocity increasing with height. The magnetic field and winds remove angular momentum from the accretion layers, a mechanism called MHD wind-driven accretion. Figure 6 demonstrates that our shearing-box simulation with SBD reproduces the overall wind-driven accretion structure of the self-similar model. Focusing on the disk interior defined by |z|1 , i.e., where Ohmic resistivity is negligible. Hereafter, we refer to the regions with \Lambda_{\rm O}>1 and \Lambda_{\rm O}<1 within the disk interior ( |z|z_{\rm dead} , ambipolar diffusion generally determines the gas–field coupling. In our disk ionization model, {\rm Am} tends to take nearly constant values of \sim 1 in the magnetically active layers, as shown in figure 10 (see also figure 1 of 3 for a similar example). However, this tendency does not always hold in general ionization models, because the ionization fraction can vary by orders of magnitude with height owing to the attenuation of external ionizing sources such as cosmic rays and stellar UV/X-rays. Figure 34 of 26 provides an example in which {\rm Am} exhibits steep variation in the surface region where \Lambda_{\rm O}>1 . With this in mind, we define the effective ambipolar Elsasser number {\rm Am}_{\rm eff} by \log{\rm Am}_{\rm eff}=\langle\log{\rm Am}\rangle_{\rm active} , where \langle\log{\rm Am}\rangle_{\rm active} denotes the average of \log{\rm Am} over the magnetically active layer. The logarithmic average is adopted to prevent {\rm Am}_{\rm eff} from being biased by values of {\rm Am} at specific heights. 4.3 Scaling the field-line pitch and accretion rate with key diagnostics We now derive a scaling law for the field-line pitch \left|B_{\phi}/B_{z}\right|_{\rm surf} at the disk surface in terms of the midplane plasma beta \beta_{0} and key diagnostics of magnetic diffusivities introduced in the previous subsection. To begin with, we plot in figure 11 (a) the relation between \left|B_{\phi}/B_{z}\right|_{\rm surf} and \beta_{0} for all runs using tabulated and parametric diffusivity models. Our results show \left|B_{\phi}/B_{z}\right|_{\rm surf}\sim 1\text{--}20 , consistent with those from previous global simulations (e.g., 13 ; 11 ; 8 ). Note that the scaling \dot{M}\propto\beta_{0}^{-0.78} found by 31 for {\rm Rm}_{0}=1 and {\rm Am}_{0}=1 (see figure 8 ) is equivalent to \left|B_{\phi}/B_{z}\right|_{\rm surf}\propto\beta_{0}^{0.22} . However, it is clear from figure 11 (a) that \beta_{0} alone does not uniquely determine \left|B_{\phi}/B_{z}\right|_{\rm surf} in more general cases. To isolate the \beta_{0} dependence without contamination from Ohmic diffusion, we focus on runs with {\rm Rm}_{0}=\infty and {\rm Am}_{0}=1 , finding \left|B_{\phi}/B_{z}\right|_{\rm surf}\propto\beta_{0}^{0.27} , which we show as the reference line in figure 11 (a). We next plot \left|B_{\phi}/B_{z}\right|_{\rm surf} normalized by \beta_{0}^{0.27} as a function of {\rm Am}_{\rm eff} (figure 11 (b)). To isolate the effect of ambipolar diffusion, we here focus on the runs for the parametric diffusivity model with {\rm Rm}_{0}=\infty (plus symbols in figure 11 ), where Ohmic dissipation is negligible. In these runs, the normalized \left|B_{\phi}/B_{z}\right|_{\rm surf} increases with {\rm Am}_{\rm eff} , suggesting that a larger {\rm Am}_{\rm eff} leads to stronger gas–field coupling in the active layer and hence stronger amplification of B_{\phi} . A power-law fit to these runs yields \left|B_{\phi}/B_{z}\right|_{\rm surf}/\beta_{0}^{0.27}\propto{\rm Am}_{\rm eff}^{0.48} . Finally, we isolate the effect of Ohmic resistivity by normalizing \left|B_{\phi}/B_{z}\right|_{\rm surf} with \beta_{0}^{0.27}{\rm Am}_{\rm eff}^{0.48} . We characterize the effect of Ohmic resistivity using the normalized thickness of the magnetically active layer, \Delta\equiv\frac{|z_{\rm surf}-z_{\rm dead}|}{|z_{\rm surf}|}, (21) where the active layer vanishes when \Delta\to 0 and the Ohmic dead zone vanishes when \Delta\to 1 . Figure 11 (c) plots \left|B_{\phi}/B_{z}\right|_{\rm surf}/(\beta_{0}^{0.27}{\rm Am}_{\rm eff}^{0.48}) versus \Delta for all runs. The figure shows that the normalized \left|B_{\phi}/B_{z}\right|_{\rm surf} is positively correlated with \Delta , suggesting that a thicker active layer leads to stronger amplification of B_{\phi} . A power-law fit yields \left|B_{\phi}/B_{z}\right|_{\rm surf}/(\beta_{0}^{0.27}{\rm Am}_{\rm eff}^{0.48})\propto\Delta^{1.2} . To summarize, we have shown that the field-line pitch \left|B_{\phi}/B_{z}\right|_{\rm surf} can be scaled as \left|\frac{B_{\phi}}{B_{z}}\right|_{\rm surf}\approx k\,\beta_{0}^{0.27}\,{\rm Am}_{\rm eff}^{0.48}\,\Delta^{1.2}, (22) where k is a numerical coefficient with a best-fit value of k=0.84 . Substituting equation ( 22 ) into equation ( 19 ), we also obtain a scaling formula for the mass accretion rate, \dot{M}\approx k\frac{2r}{\Omega_{K}}\,B_{z0}^{2}\beta_{0}^{0.27}\,{\rm Am}_{\rm eff}^{0.48}\,\Delta^{1.2}. (23) Figure 12 compares the predicted and directly measured values of \left|B_{\phi}/B_{z}\right|_{\rm surf} and \dot{M} for all runs. We find that equations ( 22 ) and ( 23 ) reproduce the measured values to within a factor of three across the explored parameter space ( \beta_{0}=3\times 10^{3} – 10^{6} , {\rm Am}_{\rm eff}\sim 0.1 – 10 , \Delta\sim 0.1 – 1 ). With the exception of the least magnetized cases ( \beta_{0}=10^{6} ), our formulae agree with the simulation results to within a factor of 2–3. Figure 11: Decomposition of the parameter dependence of the field-line pitch |B_{\phi}/B_{z}|_{\rm surf} on key parameters. Panel (a) shows the dependence on \beta_{0} , panel (b) shows the dependence on {\rm Am}_{\rm eff} after dividing by \beta_{0}^{0.27} , and panel (c) shows the dependence on \Delta after further dividing by {\rm Am}_{\rm eff}^{0.48} . Filled circles show the results of shearing-box simulations with SBD using the tabulated diffusivity model, with color indicating \beta_{0} . Plus and cross symbols show the results of shearing-box simulations with SBD using the parametric diffusivity model for {\rm Rm}_{0}=\infty and {\rm Rm}_{0}\leq 10^{2} , respectively. The solid lines represent the reference scalings \propto\beta_{0}^{0.27} , \propto{\rm Am}_{\rm eff}^{0.48} , and \propto\Delta^{1.2} , respectively. Alt text: Three scatter plots with logarithmic axes arranged horizontally. In panel (a), the horizontal axis shows beta 0 and the vertical axis shows the field-line pitch. In panel (b), the horizontal axis shows Am eff. In panel (c), the horizontal axis shows Delta. Figure 12: Validation of the scaling laws. The left panel compares the model prediction k\beta_{0}^{0.27}{\rm Am}_{\rm eff}^{0.48}\Delta^{1.2} with the measured |B_{\phi}/B_{z}|_{\rm surf} , and the right panel compares the mass accretion rate derived from equation ( 16 ) with the model prediction based on equation ( 22 ). Each data point represents the result of a shearing-box simulation with SBD. Filled circles show the results from the tabulated diffusivity model, and plus and cross symbols show the results from the parametric diffusivity model. The dashed line indicates where the model prediction and the measured value agree exactly. The dark and light gray bands indicate agreement within a factor of 2 and a factor of 3, respectively. Alt text: Two scatter plots with logarithmic axes arranged horizontally. In the left panel, the horizontal axis shows the model prediction of the field-line pitch, and the vertical axis shows the measured field-line pitch. In the right panel, the horizontal axis shows the model prediction of the mass accretion rate in units of rho, sound speed, r, and H, and the vertical axis shows the measured mass accretion rate in units of rho, sound speed, r, and H. 5 Discussion 5.1 Toward global modeling of wind-driven accretion disks without \alpha parameters The scaling formula (equation ( 23 )) derived in subsection 4.3 allows us to predict the accretion rate \dot{M} from local disk quantities, namely \beta_{0} , {\rm Am}_{\rm eff} , and \Delta . In one-dimensional disk evolution models, the magnetic diffusivity table can be evaluated locally for the evolving disk radius r and surface density \Sigma . Equation ( 23 ) then gives the corresponding accretion rate, allowing the scaling formula to be incorporated into global disk evolution calculations without introducing an additional \alpha parameter. Several one-dimensional disk evolution models incorporating magnetically driven accretion have been proposed in recent years ( 1 ; 54 ; 7 ; 23 ; 14 ; 57 ) . For instance, the model of 57 extends the classical viscous \alpha model to include MHD wind-driven accretion and parametrizes the mass accretion rate as \dot{M}=3\pi\Sigma\frac{c_{s}^{2}}{\Omega_{K}}\alpha_{\rm DW}, (24) where \alpha_{\rm DW} is a dimensionless parameter that corresponds to the \alpha parameter in the classical viscous model. However, since our equation ( 23 ) gives \dot{M} explicitly as a function of disk physical quantities, there is no need to parametrize \dot{M} with \alpha_{\rm DW} as in equation ( 24 ). For given B_{z} , equation ( 23 ) directly determines how the mass distribution in a wind-driven accretion disk evolves. The remaining uncertainty lies in the evolution of B_{z} , which is discussed in subsection 5.3 . Nonetheless, given that the model of 57 is widely used in the literature ( 34 , e.g.,) , it is useful to analyze how the parameter \alpha_{\rm DW} relates to disk properties in light of our scaling formula. Comparing equations ( 19 ) and ( 24 ) and using \Sigma=\sqrt{2\pi}\rho_{0}H (which follows from equation ( 6 ) with \int\rho dz=\Sigma ), we can obtain \alpha_{\text{DW}}=\frac{16}{3\sqrt{2\pi}}\frac{1}{\epsilon\beta_{0}}\left|\frac{B_{\phi}}{B_{z}}\right|_{\text{surf}}, (25) where \epsilon\equiv H/r is the disk aspect ratio. By substituting the empirical range of |B_{\phi}/B_{z}|_{\rm surf} from global simulations into the definition of \alpha_{\rm DW} , 57 estimate (see their equation (69)) \alpha_{\rm DW}\approx(0.2\text{--}4)\times 10^{-2}\left(\frac{\beta_{0}}{10^{4}}\right)^{-1}\left(\frac{\epsilon}{0.1}\right)^{-1}, (26) where the range of the pre-factor reflects the uncertainty in |B_{\phi}/B_{z}|_{\rm surf} ( 13 ; 11 ; 8 ) . Our scaling law for |B_{\phi}/B_{z}|_{\rm surf} enables us to reduce the uncertainty in \alpha_{\rm DW} ; substituting equation ( 22 ) into equation ( 25 ) yields \displaystyle\alpha_{\rm DW} \displaystyle\approx\frac{16k}{3\sqrt{2\pi}}\epsilon^{-1}\beta_{0}^{-0.73}{\rm Am}_{\rm eff}^{0.48}\Delta^{1.2} \displaystyle\approx 2.1\times 10^{-2}\left(\frac{\beta_{0}}{10^{4}}\right)^{-0.73}\left(\frac{\epsilon}{0.1}\right)^{-1}{\rm Am}_{\rm eff}^{0.48}\Delta^{1.2}. (27) This expression is accurate to within a factor of 2–3 (see figure 12 ). The \beta_{0}^{-0.73} dependence of \alpha_{\rm DW} corresponds to the B_{z0}^{2}\beta_{0}^{0.27}\propto\beta_{0}^{-0.73} dependence of \dot{M} (see equation ( 23 )). The factors {\rm Am}_{\rm eff}^{0.48} and \Delta^{1.2} reflect the fact that stronger ambipolar diffusion and Ohmic diffusion each lead to a smaller |B_{\phi}/B_{z}|_{\rm surf} (see section 2.5 ). 5.2 Comparison with previous local simulations without SBD With SBD, local shearing-box simulations can maintain the physical field-line geometry over long periods, allowing direct measurement of \dot{M} . In contrast, previous local studies without SBD indirectly estimated \dot{M} —by measuring the magnetic stress B_{z}B_{\phi} at either the upper or lower disk surface and substituting it into equation ( 18 )—because the physical field-line geometry breaks down in some cases ( 3 ; 6 ) . In this section, we compare our scaling formula for \dot{M} with the previous stress-based estimate by 5 ; 6 . 6 presented a stress-based empirical formula for \dot{M} based on Hall-free shearing box simulations by 5 , which adopts the minimum-mass solar nebula (MMSN) model of 25 . The formula reads (see equation (28) of 6 ) \dot{M}=0.47\times 10^{-8}\,r_{\rm au}^{1.90}\left(\frac{B_{z}}{10\,{\rm mG}}\right)^{1.32}M_{\odot}\,{\rm yr}^{-1}, (28) where r_{\rm au}=r/1\,{\rm au} . To allow direct comparison with equation ( 28 ), we now rewrite our scaling law (equation ( 23 )), using \rho_{0}=\Sigma/(\sqrt{2\pi}H) together with the MMSN surface density, temperature, and rotation profiles of \Sigma_{\rm MMSN}=1700\,r_{\rm au}^{-3/2}\ {\rm g\,cm^{-2}} , T_{\rm MMSN}=280\,r_{\rm au}^{-1/2}\ {\rm K} , and \Omega_{\rm K}=2.0\times 10^{-7}\,r_{\rm au}^{-3/2}\ {\rm s^{-1}} , as \displaystyle\dot{M} \displaystyle=1.2\times 10^{-8}{\rm Am}_{\rm eff}^{0.48}\,\Delta^{1.2}\,r_{\rm au}^{1.62}\left(\frac{B_{z}}{10\,{\rm mG}}\right)^{1.46} \displaystyle\times\left(\frac{\Sigma}{\Sigma_{\rm MMSN}}\right)^{0.27}\left(\frac{T}{T_{\rm MMSN}}\right)^{0.14}\,M_{\odot}\,{\rm yr}^{-1}. (29) Unlike equation ( 23 ), equation ( 29 ) accounts for the dependence on the surface density, temperature, and magnetic diffusivities. In both formulae, the power-law index of B_{z} is smaller than 2. If |B_{\phi}/B_{z}|_{\rm surf} were independent of B_{z} , equation ( 19 ) would give \dot{M}\propto B_{z}^{2} . As discussed in subsection 4.3 , however, |B_{\phi}/B_{z}|_{\rm surf}\propto\beta_{0}^{0.27}\propto B_{z}^{-0.54} , and \Delta also depends on B_{z} , both of which reduce the effective power-law index below 2. It is therefore natural that the explicit B_{z} dependences in the two formulae do not match exactly. For typical values of {\rm Am}_{\rm eff}\sim 1 and \Delta\sim 0.4 – 0.5 , the prefactor 1.2\times 10^{-8}{\rm Am}_{\rm eff}^{0.48}\,\Delta^{1.2} in our formula yields 0.4 – 0.5\times 10^{-8} , consistent with the prefactor 0.47\times 10^{-8} in the formula of 6 . Our formula suggests that \dot{M} depends weakly on \Sigma and T , a dependence that arises from the \beta_{0} dependence of |B_{\phi}/B_{z}|_{\rm surf} . 5.3 Limitations of this work The scaling law derived in this study provides a practical framework for estimating the mass accretion rate \dot{M} from local disk quantities. The largest uncertainty in applying this scaling law to global disk evolution models is the vertical magnetic field strength B_{z} , which is represented by \beta_{0} in our simulations. Direct observational constraints on the large-scale magnetic field threading protoplanetary disks remain limited, and current knowledge relies mainly on indirect evidence from disk observations ( 59 ; 39 ) and from the remanent magnetization of meteorites (e.g., 66 ; 38 ; 45 ). On the theoretical side, no framework has yet been established that uniquely determines the global distribution and time evolution of B_{z} . Global transport models based on magnetic flux advection and diffusion have long been proposed and applied to protoplanetary disks. However, they still do not uniquely determine the distribution and time evolution of B_{z} ( 32 ; 41 ; 58 ; 22 ) . Furthermore, because local shearing-box simulations cannot self-consistently treat global radial magnetic flux transport, \beta_{0} itself must be specified externally. The scaling law derived in this study therefore characterizes the systematic dependence of \dot{M} on disk conditions for a given \beta_{0} , and its application to real disks requires a global model for the distribution and evolution of B_{z} ( 4 ; 20 ; 31 ; 15 ; 26 ) . The model adopted in this study includes several additional simplifications beyond those already noted above. We mention them here as issues to be addressed in future work. First, our simulations include Ohmic and ambipolar diffusion but neglect the Hall effect. The Hall effect is known to modify both the horizontal magnetic field structure and the efficiency of angular momentum transport, depending on the relative orientation between the vertical magnetic field and the disk rotation axis ( 64 ; 10 ; 28 ; 63 , e.g.,) . Local simulations have shown that, when the vertical magnetic field is aligned with the disk rotation axis, the horizontal magnetic field is amplified and wind-driven accretion is enhanced, whereas in the anti-aligned case, angular momentum transport and outflow are weakened ( 30 ; 6 ) . In global simulations, by contrast, although the Hall effect significantly changes the horizontal magnetic field structure and the relative contributions of the radial and wind-driven angular momentum transport terms, the total accretion rate, including the wind-driven component, may remain comparable to that in the Hall-free case ( 8 ) . Therefore, extending the present scaling law to Hall-dominated regimes requires careful treatment of the magnetic-field polarity. Second, as shown in equation ( 3 ), we assume an isothermal equation of state throughout the disk. In reality, however, disk surface layers can be heated by irradiation and magnetic dissipation ( 35 ; 62 ; 36 ) . Such changes in the thermal structure can affect the ionization structure by modifying the vertical density distribution, and may consequently alter z_{\rm dead} and \Delta . Third, the tabulated diffusivity model adopted in this study is based on a specific ionization model (subsubsection 2.3.2 ). The values of {\rm Am}_{\rm eff} and \Delta may therefore change if the chemical reaction network, dust grain size distribution, or treatment of ionization sources is modified. The scaling law derived in this study, however, is expressed not through the ionization model itself but through the effective quantities {\rm Am}_{\rm eff} and \Delta determined from the ionization structure. Therefore, the same functional form may be applicable to other ionization models, provided that {\rm Am}_{\rm eff} and \Delta are re-evaluated from the corresponding ionization structure. 6 Summary and Conclusion In this study, we introduced the super-box-scale diffusion (SBD) scheme into non-ideal MHD shearing-box simulations to quantitatively describe magnetically driven accretion in protoplanetary disks. Local shearing-box simulations are a powerful tool for studying the vertical structure at any orbital radius with high resolution and low computational cost. However, they suffer from a fundamental problem: the toroidal magnetic field tends to accumulate within the computational domain, making it difficult to maintain the physical field-line geometry over long periods. In this study, we addressed this problem with SBD, thereby establishing a local framework for wind-driven accretion. The main results are summarized below. 1. The SBD scheme (subsection 2.2 ) suppresses the excessive accumulation of toroidal magnetic flux in the local shearing box and maintains the physical field-line geometry over long periods. Without SBD, B_{\phi} accumulates in the same direction throughout the domain and the field-line geometry transitions to the unphysical configuration after approximately 110 orbits. In contrast, with SBD, the physical field-line geometry is maintained for at least 500 orbits (figures 2 and 3 ). We further confirmed that, as long as C is chosen large enough to keep \tau_{\rm diff} shorter than the amplification timescale of B_{\phi} by shear (figure 4 ), the quasi-steady field structure is nearly independent of the value of C (figure 5 ). These results demonstrate that SBD enables a stable local framework for describing wind-driven accretion (subsection 3.1 ). 2. Local simulations with SBD quantitatively agree with the self-similar solutions of 31 . Comparisons using the parametric diffusivity model show that the accretion layer position and magnetic field structure agree with the self-similar solutions across a wide parameter range (figures 6 and 7 ). The \beta_{0} dependence of the mass accretion rate also agrees with that of the self-similar solutions to within 23–28% (figure 8 ). This confirms that local shearing-box simulations incorporating SBD provide an effective framework for approximating the global accretion structure at low computational cost (subsection 3.2 ). 3. A parameter survey of 46 runs using the magnetic diffusivity table shows that the mass accretion rate \dot{M} is not uniquely determined by the net vertical magnetic field strength \beta_{0} alone. It also depends systematically on the effective parameters {\rm Am}_{\rm eff} and \Delta (equation ( 21 )), which represent the strength of magnetic coupling and the normalized thickness of the magnetically active layer, respectively (figures 10 and 11 ). We formulate this dependence and derive a scaling law for the field-line pitch \left|B_{\phi}/B_{z}\right|_{\rm surf} (equation ( 22 )) and a predictive formula for the mass accretion rate \dot{M} (equation ( 23 )). These formulas reproduce the simulation results to within a factor of 2–3 across the parameter space explored in this study (figure 12 ). 4. Our scaling law determines the \alpha_{\rm DW} parameter in the widely used parametric wind-driven accretion model of 57 as a function of \beta_{0} , {\rm Am}_{\rm eff} , and \Delta (equation ( 27 )). These results advance the description of wind-driven accretion in protoplanetary disks beyond the conventional phenomenological \alpha prescription by directly linking the accretion rate to local physical quantities. In particular, this study clarifies the physical origin of magnetically driven accretion and provides a framework for directly predicting \dot{M} from \beta_{0} , {\rm Am}_{\rm eff} , and \Delta . This study, however, has several limitations. First, the simulations assume an isothermal equation of state and neglect the Hall effect. In addition, SBD is a prescription for incorporating large-scale magnetic relaxation into local simulations and does not self-consistently solve the radial transport of the net vertical magnetic flux. The scaling law derived in this study therefore provides a model for predicting the local mass accretion rate for a given B_{z} . An important direction for future work is to extend the local scaling law derived here to a global evolutionary model coupled to the radial transport of the net vertical magnetic flux. In practice, equation ( 23 ) can be incorporated relatively easily into one-dimensional disk evolution calculations by updating the magnetic diffusivity table as the disk radius r and surface density \Sigma evolve. The present results therefore provide a foundation for constructing long-term evolutionary models of protoplanetary disks. {ack} We thank Xuening Bai for providing the initial numerical setup for this study, Geoffroy Lesur for making the data of his self-similar solutions publicly available, and Kazunari Iwasaki and Takeru Suzuki for discussions on global MHD accretion. Funding This work was supported by JST SPRING, Japan Grant Number JPMJSP2106, and by JSPS KAKENHI Grant number JP22KJ0155, JP22K14081, and JP23K25923. Appendix A Data Table 1 summarizes the input parameters and resulting accretion quantities for all 46 runs of the tabulated diffusivity model. Table 1: List of simulations using the tabulated diffusivity model. Run r [au] \Sigma [g cm -2 ] \log_{10}\beta_{0} f_{\rm dg} {\rm Am}_{\rm eff} z_{\rm dead}[H] z_{\rm surf}[H] \dot{M} [ \rho_{0}c_{s}rH ] |B_{\phi}/B_{z}|_{\rm surf} R1-B4-S2-D4 1 10^{2} 4 10^{-4} 1.90 1.12 3.19 4.16\times 10^{-2} 8.20 R10-B4-S3-D4 10 10^{3} 4 10^{-4} 0.798 0.102 3.21 3.93\times 10^{-2} 7.83 R1-B4-S3-D3 1 10^{3} 4 10^{-3} 1.47 2.40 3.57 1.96\times 10^{-2} 3.88 R1-B4-S3-D4 1 10^{3} 4 10^{-4} 1.85 2.20 3.56 2.08\times 10^{-2} 4.16 R1-B4-S3-D5 1 10^{3} 4 10^{-5} 1.92 2.14 3.56 2.10\times 10^{-2} 4.16 R30-B4-S3-D4 30 10^{3} 4 10^{-4} 0.924 0.00 2.96 5.57\times 10^{-2} 11.3 R3-B4-S3-D4 3 10^{3} 4 10^{-4} 1.27 1.65 3.45 2.59\times 10^{-2} 5.20 R1-B4-S4-D4 1 10^{4} 4 10^{-4} 1.89 2.92 3.85 9.57\times 10^{-3} 1.85 R1-B5-S2-D4 1 10^{2} 5 10^{-4} 2.17 1.27 3.32 1.02\times 10^{-2} 20.3 R10-B5-S3-D4 10 10^{3} 5 10^{-4} 0.840 0.229 3.31 1.04\times 10^{-2} 20.7 R1-B5-S3-D3 1 10^{3} 5 10^{-3} 2.15 2.52 3.83 3.94\times 10^{-3} 7.85 R1-B5-S3-D4 1 10^{3} 5 10^{-4} 2.38 2.37 3.83 4.05\times 10^{-3} 8.03 R1-B5-S3-D5 1 10^{3} 5 10^{-5} 2.38 2.33 3.84 4.20\times 10^{-3} 8.20 R30-B5-S3-D4 30 10^{3} 5 10^{-4} 0.931 0.00 3.03 1.66\times 10^{-2} 33.1 R3-B5-S3-D4 3 10^{3} 5 10^{-4} 1.51 1.81 3.62 6.00\times 10^{-3} 11.9 R1-B5-S4-D4 1 10^{4} 5 10^{-4} 2.11 3.05 4.17 2.01\times 10^{-3} 3.96 R10-B6-S3-D4 10 10^{3} 6 10^{-4} 6.62 0.841 4.01 1.14\times 10^{-3} 21.3 R1-B6-S3-D3 1 10^{3} 6 10^{-3} 2.69 2.61 3.97 9.55\times 10^{-4} 18.9 R1-B6-S3-D4 1 10^{3} 6 10^{-4} 2.67 2.48 3.97 9.63\times 10^{-4} 19.1 R1-B6-S3-D5 1 10^{3} 6 10^{-5} 2.71 2.45 3.97 9.73\times 10^{-4} 19.3 R30-B6-S3-D4 30 10^{3} 6 10^{-4} 11.8 0.247 4.08 1.25\times 10^{-3} 24.5 R3-B6-S3-D4 3 10^{3} 6 10^{-4} 1.56 1.86 3.65 1.75\times 10^{-3} 34.8 R1-B6-S4-D4 1 10^{4} 6 10^{-4} 2.58 3.18 4.38 4.09\times 10^{-4} 8.11 R1-B3.5-S2-D4 1 10^{2} 3.5 10^{-4} 1.66 0.982 3.05 9.18\times 10^{-2} 5.81 R10-B3.5-S3-D4 10 10^{3} 3.5 10^{-4} 0.766 0.0484 3.12 8.13\times 10^{-2} 5.15 R1-B3.5-S3-D3 1 10^{3} 3.5 10^{-3} 1.35 2.34 3.39 4.21\times 10^{-2} 2.72 R1-B3.5-S3-D4 1 10^{3} 3.5 10^{-4} 1.75 2.12 3.38 4.79\times 10^{-2} 3.07 R1-B3.5-S3-D5 1 10^{3} 3.5 10^{-5} 1.79 2.04 3.38 4.90\times 10^{-2} 3.14 R30-B3.5-S3-D4 30 10^{3} 3.5 10^{-4} 0.906 0.00 2.87 1.18\times 10^{-1} 7.34 R3-B3.5-S3-D4 3 10^{3} 3.5 10^{-4} 1.11 1.55 3.30 5.60\times 10^{-2} 3.65 R1-B4.5-S2-D4 1 10^{2} 4.5 10^{-4} 2.05 1.21 3.28 1.97\times 10^{-2} 12.5 R10-B4.5-S3-D4 10 10^{3} 4.5 10^{-4} 0.824 0.148 3.28 1.94\times 10^{-2} 12.4 R1-B4.5-S3-D3 1 10^{3} 4.5 10^{-3} 1.76 2.45 3.73 8.69\times 10^{-3} 5.49 R1-B4.5-S3-D4 1 10^{3} 4.5 10^{-4} 2.09 2.28 3.73 8.98\times 10^{-3} 5.66 R1-B4.5-S3-D5 1 10^{3} 4.5 10^{-5} 2.12 2.23 3.73 9.02\times 10^{-3} 5.69 R30-B4.5-S3-D4 30 10^{3} 4.5 10^{-4} 0.934 0.00 3.00 2.95\times 10^{-2} 18.9 R3-B4.5-S3-D4 3 10^{3} 4.5 10^{-4} 1.41 1.74 3.55 1.23\times 10^{-2} 7.79 R1-B4.5-S4-D4 1 10^{4} 4.5 10^{-4} 1.97 2.98 4.02 4.50\times 10^{-3} 2.87 R1-B5.5-S2-D4 1 10^{2} 5.5 10^{-4} 2.31 1.37 3.38 5.22\times 10^{-3} 33.1 R10-B5.5-S3-D4 10 10^{3} 5.5 10^{-4} 0.884 0.480 3.35 5.43\times 10^{-3} 34.5 R1-B5.5-S3-D3 1 10^{3} 5.5 10^{-3} 2.34 2.58 3.92 1.87\times 10^{-3} 11.9 R1-B5.5-S3-D4 1 10^{3} 5.5 10^{-4} 2.55 2.44 3.91 1.96\times 10^{-3} 12.4 R1-B5.5-S3-D5 1 10^{3} 5.5 10^{-5} 2.57 2.40 3.91 1.97\times 10^{-3} 12.5 R30-B5.5-S3-D4 30 10^{3} 5.5 10^{-4} 7.17 0.0674 3.76 3.85\times 10^{-3} 22.9 R3-B5.5-S3-D4 3 10^{3} 5.5 10^{-4} 1.59 1.84 3.67 3.08\times 10^{-3} 19.6 R1-B5.5-S4-D4 1 10^{4} 5.5 10^{-4} 2.40 3.11 4.30 8.81\times 10^{-4} 5.59