betaFeedback

WorksPhysics and Astronomy

Turbulent transport of radiation in the solar convective zone

Abstract

A turbulent transport of radiation in the solar convective zone is investigated. The mean-field equation for the irradiation intensity is derived. It is shown that due to the turbulent effects, the effective penetration length of radiation can be increased in several times in comparison with the mean penetration length of radiation (defined as an inverse mean absorption coefficient). Using the model of the solar convective zone based on the mixing length theory, where the mean penetration length of radiation is usually much smaller than the turbulent correlation length, it is demonstrated that the ratio of the effective penetration length to the mean penetration length of radiation increases in 2.5 times in the vicinity of the solar surface. The main reason are the compressibility effects that become important in the vicinity of the solar surface where temperature and density fluctuations increase towards the solar surface, enhancing fluctuations of the radiation absorption coefficient and increasing the effective penetration length of radiation.

Introduction

Turbulent transport of temperature, particles and magnetic fields have been studied analytically, in laboratory and field experiments and in numerical simulations for more than a century (see, e.g., [33, 34, 31, 18, 25, 12, 40]), however, some fundamental questions remain. This is particularly true in applications to astrophysics, where the governing parameter values are too extreme to be modelled either experimentally or numerically.

In astrophysical turbulent flows radiative transport can be affected by turbulence. This effect is different for optically thick and optically thin regimes of the radiative transport. For optically thick regime, the mean free path of photons is much smaller in comparison with the typical scales of the flow. In the opposite case, i.e., for optically thin regime, the mean free path of photons is much larger than the fluid motion scales, so the photons propagate over large distances before they are absorbed and re-emitted again (see, e.g., [11, 32, 2, 30, 20]).

The relaxation time of small temperature perturbations by radiative diffusion has been determined by Spiegel [44], where a time-dependent equation for the temperature field of a medium with deviations from radiative equilibrium has been derived assuming that the medium is graylike and there are no internal motions or compressional effects, and heat is exchanged only radiatively. It has been shown that perturbations of small amplitude imposed on a homogeneous medium decay exponentially in time, and the decay time depends on a characteristic length of the perturbations [44].

Turbulent diffusion can be increased by a radiative diffusion, i.e., by a photon diffusion. For instance, the decay rates of sinusoidal large-scale temperature perturbations in the optically thick and thin regimes have been determined by Brandenburg & Das [4] using radiative hydrodynamic direct numerical simulations of forced turbulence. It has been shown there that the rate of decay increases with the wavenumber. However, this effect is much weaker in comparison with the effect of the standard turbulent diffusion [4].

In the present study, we investigate the turbulent transport of radiation in the solar convective zone. We derive a mean-field equation for the irradiation intensity, and show that the effective penetration length of radiation can be increased by turbulence in several times in comparison with the mean penetration length of radiation which is defined as an inverse mean absorption coefficient. This effect has been tested using the model of the solar convective zone (Spruit 1974) based on the mixing length theory. According to this model, the mean penetration length of radiation is much less than the turbulent correlation length. We show that the ratio of the effective penetration length to the mean penetration length of radiation increases in 2.5 times in the vicinity of the solar surface.

This paper is organized as follows. In § 2, we discuss a general concept of turbulent transport of radiation, and in § 3 we derive the mean-field radiation transport equation. In Section 4, we derive expression for the effective penetration length of radiation in turbulent flows, which depends on the ratio of fluctuations of the radiation absorption coefficient to the mean penetration length of radiation. To calculate the effective penetration length of radiation, in Section 4 we determine fluctuations of the radiation absorption coefficient which are caused by fluctuations of fluid temperature and density, and the temperature–density correlations. In § 5, we apply obtained results to the solar convective zone. Finally, in § 6, we discuss our results and draw conclusions. In Appendix § 7 we derive expressions for the level of temperature and density fluctuations as well as temperature–density correlations which allow us to determine fluctuations of the radiation absorption coefficient.

General concept of turbulent transport of radiation

In solar and stellar convective zones, the convective transport of energy is more effective than the radiative transfer. The Schwarzschild criterion for the onset of convection does not take into account an effect of turbulence on radiative transfer. However, the absorption coefficient of radiation depends on temperature and density, and there are strong fluctuations of the fluid temperature and density in solar and stellar convective zones. These fluctuations affect the absorption coefficient of radiation, and therefore they affect turbulent transfer of radiation.

In solar and stellar convective zones, the characteristic times of turbulent motions are much larger than the radiation time, and the integral turbulent scales are much larger than the mean penetration length of radiation defined as the inverse absorption coefficient of radiation. The latter implies that turbulent eddies are optically thick, and inhomogeneities in the fluid temperature and density can strongly affect the radiation transfer.

To describe the radiation transfer, we use the radiation transport equation. This equation represents a steady state version of the equation for electromagnetic energy transfer, since the time of photon propagation (the radiation time) is very short. This equation is characterized by the absorption coefficient of radiation and the black-body radiation intensity of the gas. The radiative transport equation for the intensity I(r,s^,ω)I(r,\hat{s},\omega) reads (see, e.g., Chandrasekhar 1960; Mihalas & Mihalas 1984; Apresyan & Kravtsov 1996; Liou 2002; Howell et al. 2010):

(s^⋅∇)I(r,s^,ω)=−κ(r,ω)[I−Ib(T,ω)],(1)(\hat{s}\cdot\nabla)I(r,\hat{s},\omega)=-\kappa(r,\omega)\left[I-I_b(T,\omega)\right], \tag*{(1)}

where rr is the position vector, s^=k/k\hat{s}=k/k is the unit vector in the direction of radiation kk, kk is the wave vector, κ(r,ω)=ρκop\kappa(r,\omega)=\rho\kappa_{\rm op} is the absorption coefficient of gas, κop=κ0ρaTb\kappa_{\rm op}=\kappa_0\rho^aT^b is the opacity of the gas, TT and ρ\rho are the gas temperature and density, Ib(T,ω)I_b(T,\omega) is the black-body radiation intensity of the gas, and ω\omega is the radiation frequency. Here we take into account the radiation absorption in gases and neglect the radiation scattering in gases. The function Ib(T,ω)I_b(T,\omega) in a local thermodynamic equilibrium is given by

Ib(T,ω)=ℏω3π2c3[exp⁡(ℏωkBT)−1]−1,(2)I_b(T,\omega)=\frac{\hbar\omega^3}{\pi^2c^3}\left[\exp\left(\frac{\hbar\omega}{k_{\rm B}T}\right)-1\right]^{-1}, \tag*{(2)}

where ℏ\hbar is Planck’s constant, cc is the speed of light and kBk_{\rm B} is the Boltzman constant. The integral ∫Ib(T,ω) dω∝σT4\int I_b(T,\omega)\,d\omega\propto\sigma T^4 yields the Stefan–Boltzmann law.

Our goal is to derive effective radiation transport equation with effective transport coefficients: effective absorption coefficient of radiation and the effective source of the radiation intensity. To take into account the turbulence effects, we apply a mean-field approach and average the radiation transport equation (1) over ensemble of fluctuations. In the framework of the mean-field approach, all quantities are decomposed into the mean and fluctuating parts: I=I‾+I′I=\overline{I}+I', Ib=Ib‾+Ib′I_b=\overline{I_b}+I_b' and κ=κ‾+κ′\kappa=\overline{\kappa}+\kappa'. We adopt the Reynolds averaging, where I‾=⟨I⟩\overline{I}=\langle I\rangle, Ib‾=⟨Ib⟩\overline{I_b}=\langle I_b\rangle, κ‾=⟨κ⟩\overline{\kappa}=\langle\kappa\rangle are the mean fields, and I′I', Ib′I_b', κ′\kappa' are the fluctuating fields with zero mean, and the angular brackets denote ensemble averaging. To derive the mean-field radiation transport equation, we adopt a method applied by Kliorin et al. (1989); Liberman et al. [26, 27].

The obtained mean-field equation contains the correlation function for fluctuations of the absorption coefficient of radiation κ′\kappa' and the radiation intensity I′I', i.e., ⟨κ′I′⟩\langle\kappa'I'\rangle. This correlation is due to fluctuations of temperature and density. This equation also contains the correlation function for fluctuations of the absorption coefficient of radiation κ′\kappa' and the black-body radiation intensity of the gas ⟨κ′Ib′⟩\langle\kappa'I_b'\rangle due to fluctuations of temperature.

To determine the correlation functions, ⟨κ′I′⟩\langle\kappa'I'\rangle and ⟨κ′Ib′⟩\langle\kappa'I_b'\rangle, we derive equation for fluctuations of the radiation intensity I′I' by subtracting the obtained mean-field equation from the radiation transport equation (1). Since the equation for fluctuations of the radiation intensity I′I' is a linear equation, we solve this equation exactly. However, the solution of this equation is nonlinear in fluctuations of κ′\kappa'. This causes appearance of the high-order moments in fluctuations of κ′\kappa' in the expression for the correlation function ⟨κ′I′⟩\langle\kappa'I'\rangle. We assume that fluctuations of κ′\kappa' are essentially less than the mean absorption coefficient of radiation. This allows us to obtain the closed results.

The main expected result of this study is that the derived mean-field equation for the radiation transfer with the effective transport coefficients yields the effective penetration length of radiation. When the effective penetration length of radiation is larger than the mean penetration length of radiation, the absorption coefficient decreases and an observer can see more deeper layers inside the stars. The reasons for the increase of the effective penetration length of radiation in turbulent flows are caused by the correlation between fluctuations of the radiation absorption coefficient κ′\kappa' and fluctuations of the irradiation intensity I′I'. We show below that this correlation function should be negative, because an increase of the absorption of radiation decreases the radiation intensity and wise versa. We also demonstrate in this study that this effect is essential in the vicinity of the solar surface.

Mean-field radiation transport equation

In this section we derive the mean-field radiation transport equation. Ensemble averaging of equation (1) yields the equation for the mean radiation intensity I‾\overline{I}:

(s^⋅∇)I‾=−κ‾(I‾−Ib‾)−⟨κ′I′⟩+⟨κ′Ib′⟩.(3)(\hat{\boldsymbol{s}}\cdot\nabla)\overline{I}=-\overline{\kappa}\left(\overline{I}-\overline{I_b}\right)-\langle\kappa'I'\rangle+\langle\kappa'I_b'\rangle. \tag*{(3)}

This equation contains unknown correlation functions, ⟨κ′I′⟩\langle\kappa'I'\rangle and ⟨κ′Ib′⟩\langle\kappa'I_b'\rangle. To determine these correlation functions, we derive equation for fluctuations of the radiation intensity I′I'. To this end, we subtract the mean-field radiation transport equation (3) from equation (1), so that the equation for fluctuations of I′I' reads:

(s^⋅∇+κ‾+κ′)I′(r,s^)=Isource,(4)\left(\hat{\boldsymbol{s}}\cdot\nabla+\overline{\kappa}+\kappa'\right)I'(r,\hat{\boldsymbol{s}})=I_{\mathrm{source}}, \tag*{(4)}

where the source term IsourceI_{\mathrm{source}} is given by

Isource=−κ′(I‾−Ib‾)+⟨κ′I′⟩+(κ‾+κ′)Ib′−⟨κ′Ib′⟩.(5)I_{\mathrm{source}}=-\kappa'\left(\overline{I}-\overline{I_b}\right)+\langle\kappa'I'\rangle+\left(\overline{\kappa}+\kappa'\right)I_b'-\langle\kappa'I_b'\rangle. \tag*{(5)}

The solution of equation (4) reads

I′(r,s^)=∫−∞∞exp⁡[−∫s′s[κ‾+κ′′(s′′)] ds′′]×Isource(s′) ds′,(6)I'(r,\hat{\boldsymbol{s}})=\int_{-\infty}^{\infty}\exp\left[-\int_{s'}^{s}\left[\overline{\kappa}+\kappa''(s'')\right]\,\mathrm{d}s''\right]\times I_{\mathrm{source}}(s')\,\mathrm{d}s', \tag*{(6)}

where s=r⋅s^s=r\cdot\hat{\boldsymbol{s}}. This solution is nonlinear in fluctuations of κ′\kappa'. The latter causes appearance of the high-order moments in fluctuations of κ′\kappa' in the expression for the correlation function ⟨κ′I′⟩\langle\kappa'I'\rangle. The high-order moments are much less than the lower-order moments, because κ′≪κ‾\kappa'\ll\overline{\kappa}. This allows us to expand the function, exp⁡[−∫s′sκ′(s′′) ds′′]\exp\left[-\int_{s'}^{s}\kappa'(s'')\,\mathrm{d}s''\right], in equation (6) in Taylor series:

exp⁡[−∫s′sκ′(s′′) ds′′]=1−∫s′sκ′(s′′) ds′′+O(κ′2).(7)\exp\left[-\int_{s'}^{s}\kappa'(s'')\,\mathrm{d}s''\right]=1-\int_{s'}^{s}\kappa'(s'')\,\mathrm{d}s''+O\left(\kappa'^{2}\right). \tag*{(7)}

Therefore, equation (6) can be rewritten as:

I′(r,s^)=∫−∞∞Isource(s′)exp⁡(−κ‾∣s−s′∣)(1−∫s′sκ′(s′′) ds′′) ds′+O(κ′2).(8)I'(r,\hat{\boldsymbol{s}})=\int_{-\infty}^{\infty}I_{\mathrm{source}}(s')\exp\left(-\overline{\kappa}|s-s'|\right)\left(1-\int_{s'}^{s}\kappa'(s'')\,\mathrm{d}s''\right)\,\mathrm{d}s'+O\left(\kappa'^{2}\right). \tag*{(8)}

Multiplying equation (8) by κ′\kappa' and averaging over the ensemble, we obtain expression for the one-point correlation function ⟨κ′I′⟩\langle\kappa'I'\rangle:

⟨κ′I′⟩[1+∫−∞∞(∫s′s⟨κ′(s)κ′(s′′)⟩ ds′′)exp⁡(−κ‾∣s−s′∣) ds′]=−[∫−∞∞⟨κ′(s)κ′(s′)⟩exp⁡(−κ‾∣s−s′∣) ds′](I‾−Ib‾).(9)\langle\kappa'I'\rangle\left[1+\int_{-\infty}^{\infty}\left(\int_{s'}^{s}\langle\kappa'(s)\kappa'(s'')\rangle\,\mathrm{d}s''\right)\exp\left(-\overline{\kappa}|s-s'|\right)\,\mathrm{d}s'\right] =-\left[\int_{-\infty}^{\infty}\langle\kappa'(s)\kappa'(s')\rangle\exp\left(-\overline{\kappa}|s-s'|\right)\,\mathrm{d}s'\right]\left(\overline{I}-\overline{I_b}\right). \tag*{(9)}

where we neglect the third-order and higher-order moments in fluctuations of κ′\kappa'. Equation (9) can be rewritten as

⟨κ′I′⟩=−κ‾(I‾−Ib‾)2κ‾J11+2κ‾J2.(10)\langle\kappa'I'\rangle=-\overline{\kappa}\left(\overline{I}-\overline{I_b}\right)\frac{2\overline{\kappa}J_1}{1+2\overline{\kappa}J_2}. \tag*{(10)}

where the integrals J1J_1 and J2J_2 in equation (10) are defined as:

J1=∫0∞Φ(Z)exp⁡(−κ‾Z) dZ,(11)J_1=\int_{0}^{\infty}\Phi(Z)\exp(-\overline{\kappa}Z)\,\mathrm{d}Z, \tag*{(11)}
J2=κ‾∫0∞(∫0ZΦ(Z′) dZ′)exp⁡(−κ‾Z) dZ,(12)J_2=\overline{\kappa}\int_{0}^{\infty}\left(\int_{0}^{Z}\Phi(Z')\,\mathrm{d}Z'\right)\exp(-\overline{\kappa}Z)\,\mathrm{d}Z, \tag*{(12)}

the function Φ(Z)\Phi(Z) is defined as Φ(Z)=⟨κ′(s)κ′(s′)⟩\Phi(Z)=\langle\kappa'(s)\kappa'(s')\rangle and Z=∣s−s′∣Z=|s-s'|.

Substituting equation (10) into the mean-field equation (3), we arrive at the mean-field radiation transport equation as

(s^⋅∇)I‾=−κeff(I‾−Ibeff),(13)(\hat{\boldsymbol{s}}\cdot\nabla)\overline{I}=-\kappa_{\mathrm{eff}}\left(\overline{I}-I_b^{\mathrm{eff}}\right), \tag*{(13)}

where the effective absorption coefficient κeff\kappa_{\mathrm{eff}} is given by

κeff=κ‾(1−2κ‾J11+2κ‾J2),(14)\kappa_{\mathrm{eff}}=\overline{\kappa}\left(1-\frac{2\overline{\kappa}J_1}{1+2\overline{\kappa}J_2}\right), \tag*{(14)}

and the effective radiation intensity is

Ibeff=Ib‾+⟨κ′Ib′⟩κeff.(15)I_b^{\mathrm{eff}}=\overline{I_b}+\frac{\langle\kappa'I_b'\rangle}{\kappa_{\mathrm{eff}}}. \tag*{(15)}

The function Ib‾\overline{I_b} is expanded in Taylor series as

Ib‾=[Ib+⟨θ2⟩2∂2Ib∂T2]T=T‾,(16)\overline{I_b}=\left[I_b+\frac{\langle\theta^2\rangle}{2}\frac{\partial^2 I_b}{\partial T^2}\right]_{T=\overline{T}}, \tag*{(16)}

where the fluid temperature is decomposed into the mean T‾\overline{T} and fluctuating θ\theta parts: T=T‾+θT=\overline{T}+\theta. Solution of the mean-field radiation transport equation (13) for the mean irradiation intensity I‾(r,s^,ω)\overline{I}(r,\hat{\boldsymbol{s}},\omega) is given by

I‾(r,s^,ω)=∫−∞∞Ibeff(r′,ω)exp⁡[−∣τ(r,r′,s^)∣]s^⋅dr′,(17)\overline{I}(r,\hat{\boldsymbol{s}},\omega)=\int_{-\infty}^{\infty}I_b^{\mathrm{eff}}(r',\omega)\exp\left[-|\tau(r,r',\hat{\boldsymbol{s}})|\right]\hat{\boldsymbol{s}}\cdot\mathrm{d}r', \tag*{(17)}

where τ(r,r′,s^)=∫rr′κeff(r′′) s^⋅dr′′\tau(r,r',\hat{\boldsymbol{s}})=\int_{r}^{r'}\kappa_{\mathrm{eff}}(r'')\,\hat{\boldsymbol{s}}\cdot\mathrm{d}r'' is the optical depth.

Effective penetration length of radiation and fluctuations of absorption coefficient

In this section we determine the effective penetration length of radiation in turbulent flows, defined as Leff=κeff−1L_{\mathrm{eff}}=\kappa_{\mathrm{eff}}^{-1}. Since the main contribution to the second moment ⟨κ′(s)κ′(s′)⟩\langle\kappa'(s)\kappa'(s')\rangle of fluctuations of the absorption coefficient is from the integral scale of turbulence ℓ0\ell_0, we assume that this correlation function has the form:

⟨κ′(s)κ′(s′)⟩=⟨κ′2⟩exp⁡(−∣s−s′∣ℓ0).(18)\langle\kappa'(s)\kappa'(s')\rangle=\langle\kappa'^{2}\rangle\exp\left(-\frac{|s-s'|}{\ell_0}\right). \tag*{(18)}

Using equations (11)–(12) and (18), we calculate the integrals J1J_1 and J2J_2 as

J1=J2=⟨κ′2⟩κ‾3(1+Lrℓ0)−1,(19)J_1=J_2=\frac{\langle\kappa'^{2}\rangle}{\overline{\kappa}^{3}}\left(1+\frac{L_r}{\ell_0}\right)^{-1}, \tag*{(19)}

where Lr=κ‾−1L_r=\overline{\kappa}^{-1} characterises the mean penetration length of radiation in turbulent flow. Therefore, equations (14) and (19) allow us to determine the effective penetration length Leff=κeff−1L_{\mathrm{eff}}=\kappa_{\mathrm{eff}}^{-1} of radiation in a turbulent flow as

Leff=Lr[1+2⟨κ′2⟩κˉ2(1+Lrℓ0)−1].(20)L_{\mathrm{eff}}=L_r\left[1+\frac{2\langle\kappa'^{2}\rangle}{\bar{\kappa}^{2}}\left(1+\frac{L_r}{\ell_0}\right)^{-1}\right]. \tag*{(20)}

We consider two limiting cases:

(i) ℓ0≪Lr\ell_0\ll L_r, the effective penetration length LeffL_{\mathrm{eff}} is

Leff=Lr[1+2⟨κ′2⟩κˉ2ℓ0Lr],(21)L_{\mathrm{eff}}=L_r\left[1+\frac{2\langle\kappa'^{2}\rangle}{\bar{\kappa}^{2}}\frac{\ell_0}{L_r}\right], \tag*{(21)}

(ii) ℓ0≫Lr\ell_0\gg L_r, the effective penetration length LeffL_{\mathrm{eff}} is

Leff=Lr[1+2⟨κ′2⟩κˉ2].(22)L_{\mathrm{eff}}=L_r\left[1+\frac{2\langle\kappa'^{2}\rangle}{\bar{\kappa}^{2}}\right]. \tag*{(22)}

(22) implies that for ℓ0≫Lr\ell_0\gg L_r, the effective penetration length LeffL_{\mathrm{eff}} can increase in 3 times in comparison with the mean penetration length LrL_r of radiation due to the turbulence effects when ⟨κ′2⟩∼κˉ2\langle\kappa'^{2}\rangle\sim\bar{\kappa}^{2}.

The mechanism of increase of the effective penetration length LeffL_{\mathrm{eff}} in turbulent flows is related to the correlation between fluctuations of the radiation absorption coefficient κ′\kappa' and fluctuations of the irradiation intensity I′I'. The correlation ⟨κ′I′⟩\langle\kappa'I'\rangle is negative because an increase of the absorption of radiation decreases the radiation intensity and wise versa. Fluctuations of the radiation absorption coefficient are caused by fluctuations of fluid temperature and density in turbulent flow.

Now we determine fluctuations of the radiation absorption coefficient. For simplicity, we assume that the opacity of gas is κop=κ0ρaTb\kappa_{\mathrm{op}}=\kappa_0\rho^a T^b, so that the absorption coefficient of gas is κ=ρκop=κ0ρa+1Tb\kappa=\rho\kappa_{\mathrm{op}}=\kappa_0\rho^{a+1}T^b. According to the Schwarzschild stability criterion, the case a=1a=1 and b=0b=0 corresponds to the marginally stable regime, while the case a=1a=1 and b=1b=1 corresponds to the unstable regime [3]. The equation κ=κ0ρa+1Tb\kappa=\kappa_0\rho^{a+1}T^b allows us to determine the ratio of fluctuations of the absorption coefficient κ′\kappa' to the mean value of κˉ\bar{\kappa} as

κ′κˉ=(a+1)ρ′ρˉ+bθTˉ,(23)\frac{\kappa'}{\bar{\kappa}}=(a+1)\frac{\rho'}{\bar{\rho}}+b\frac{\theta}{\bar{T}}, \tag*{(23)}

where ρ′\rho' are density fluctuations and ρˉ\bar{\rho} is the mean fluid density.

Using equation (23), we determine the correlation function ⟨κ′Ib′⟩\langle\kappa'I'_b\rangle as:

⟨κ′Ib′⟩=κˉ[(a+1)⟨ρ′θ⟩ρˉ+b⟨θ2⟩Tˉ](∂Ib∂T)T=Tˉ,(24)\langle\kappa'I'_b\rangle=\bar{\kappa}\left[(a+1)\frac{\langle\rho'\theta\rangle}{\bar{\rho}}+b\frac{\langle\theta^2\rangle}{\bar{T}}\right]\left(\frac{\partial I_b}{\partial T}\right)_{T=\bar{T}}, \tag*{(24)}

where we take into account that Ib′=θ(∂Ib/∂T)T=TˉI'_b=\theta\left(\partial I_b/\partial T\right)_{T=\bar{T}}. To find the effective penetration length of radiation, we determine the level of fluctuations ⟨κ′2⟩\langle\kappa'^{2}\rangle of the absorption coefficient as

⟨κ′2⟩κˉ2=(a+1)2⟨ρ′2⟩ρˉ2+b2⟨θ2⟩Tˉ2+2b(a+1)⟨θρ′⟩ρˉTˉ.(25)\frac{\langle\kappa'^{2}\rangle}{\bar{\kappa}^{2}}=(a+1)^2\frac{\langle\rho'^{2}\rangle}{\bar{\rho}^{2}}+b^2\frac{\langle\theta^2\rangle}{\bar{T}^{2}}+2b(a+1)\frac{\langle\theta\rho'\rangle}{\bar{\rho}\bar{T}}. \tag*{(25)}

The intensity of temperature fluctuations in a developed compressible turbulence for large Péclet and Reynolds numbers is given by

⟨θ2⟩Tˉ2=8fc(γ−1)2(σc1+σc)3[1−(λℓ0)29]+89ℓ02b2[∇TˉTˉ+(γ−1)λ]2,(26)\frac{\langle\theta^2\rangle}{\bar{T}^{2}}=8f_c(\gamma-1)^2\left(\frac{\sigma_c}{1+\sigma_c}\right)^3\left[1-\frac{(\lambda\ell_0)^2}{9}\right]+\frac{8}{9}\ell_0^2b^2\left[\frac{\nabla\bar{T}}{\bar{T}}+(\gamma-1)\lambda\right]^2, \tag*{(26)}

(see Appendix A), where γ=cp/cv\gamma=c_p/c_v is the ratio of specific heats, λ=−∇ln⁡ρˉ\lambda=-\nabla\ln\bar{\rho} characterizes the inhomogeneity of the mean fluid density, the parameter

σc=⟨(∇⋅u)2⟩⟨(∇×u)2⟩(27)\sigma_c=\frac{\langle(\nabla\cdot\boldsymbol{u})^2\rangle}{\langle(\nabla\times\boldsymbol{u})^2\rangle} \tag*{(27)}

is the degree of compressibility of the turbulent velocity field u\boldsymbol{u} and ℓ0\ell_0 is the integral scale of turbulence. The function fc(q,qc,σc)f_c(q,q_c,\sigma_c) depends on the degree of compressibility and the exponents of spectra qq and qcq_c for the incompressible and compressible parts of velocity fluctuations (see Appendix A):

fc=qc−13qc−5+2(qc−1)σc(q+2qc−5)+qc−1σc2(2q+qc−5).(28)f_c=\frac{q_c-1}{3q_c-5}+\frac{2(q_c-1)}{\sigma_c(q+2q_c-5)}+\frac{q_c-1}{\sigma_c^2(2q+q_c-5)}. \tag*{(28)}

Equation (26) is different from that derived by Rogachevskii & Kleorin [41]. In this study, we take into account a strong density stratification. The latter is important in view of applications to the solar convective zone, where the fluid density in radial direction is changed by seven orders of magnitude. We also neglect the gradient of the turbulent diffusion that is changed very slowly in the solar convective zone. The first term in the right hand side of equation (26) determines a compressibility contribution of velocity fluctuations to temperature fluctuations, while the second term in equation (26) is proportional to the squared gradient of the mean entropy.

The level of fluid density fluctuations in a developed compressible turbulence for large Péclet and Reynolds numbers is given by

⟨ρ′2⟩ρˉ2=8fc(σc1+σc)3[1−(λℓ0)29],(29)\frac{\langle\rho'^{2}\rangle}{\bar{\rho}^{2}}=8f_c\left(\frac{\sigma_c}{1+\sigma_c}\right)^3\left[1-\frac{(\lambda\ell_0)^2}{9}\right], \tag*{(29)}

and the cross correlations ⟨θρ′⟩\langle\theta\rho'\rangle is

⟨θρ′⟩ρˉTˉ=8fc(γ−1)(σc1+σc)3[1−(λℓ0)29].(30)\frac{\langle\theta\rho'\rangle}{\bar{\rho}\bar{T}}=8f_c(\gamma-1)\left(\frac{\sigma_c}{1+\sigma_c}\right)^3\left[1-\frac{(\lambda\ell_0)^2}{9}\right]. \tag*{(30)}

Equations (26), (29) and (30) are valid for small σc<1\sigma_c<1. The latter condition is typical for developed turbulence and turbulent convection. Derivation of equations (26), (29) and (30) is given in Appendix A.

Therefore, the level of fluctuations ⟨κ′2⟩\langle\kappa'^{2}\rangle of the absorption coefficient is

⟨κ′2⟩κˉ2=8fc(σc1+σc)3[1−(λℓ0)29][a+1+b(γ−1)]2+89ℓ02b2[∇TˉTˉ+(γ−1)λ]2.(31)\frac{\langle\kappa'^{2}\rangle}{\bar{\kappa}^{2}}=8f_c\left(\frac{\sigma_c}{1+\sigma_c}\right)^3\left[1-\frac{(\lambda\ell_0)^2}{9}\right]\left[a+1 +b(\gamma-1)\right]^2+\frac{8}{9}\ell_0^2b^2\left[\frac{\nabla\bar{T}}{\bar{T}}+(\gamma-1)\lambda\right]^2. \tag*{(31)}

For a=1a=1 and b=1b=1, the level of fluctuations ⟨κ′2⟩\langle\kappa'^{2}\rangle of the absorption coefficient is given by

⟨κ′2⟩κˉ2=8fc(σc1+σc)3[1−(λℓ0)29](γ+1)2\frac{\langle\kappa'^{2}\rangle}{\bar{\kappa}^{2}}=8f_c\left(\frac{\sigma_c}{1+\sigma_c}\right)^3\left[1-\frac{(\lambda\ell_0)^2}{9}\right](\gamma+1)^2
Image not extracted

Figure 1. The radial profile of the ratio Lr/ℓ0L_r/\ell_0 for the solar convective zone.

Image not extracted

Figure 2. The profile of the ratio ℓm/Hρ\ell_m/H_\rho of the mixing length ℓm\ell_m to the density stratification length HρH_\rho versus r/R⊙r/R_\odot that is based on the model of the solar convective zone.

+89ℓ02b2[∇T‾T‾+(γ−1)∇ρ‾ρ‾]2.(32)+\frac{8}{9}\ell_0^2 b^2\left[\frac{\nabla\overline{T}}{\overline{T}}+(\gamma-1)\frac{\nabla\overline{\rho}}{\overline{\rho}}\right]^2. \tag*{(32)}

Note that for nearly isentropic flows where ∇ln⁡T‾≈(γ−1)∇ln⁡ρ‾\nabla\ln\overline{T}\approx(\gamma-1)\nabla\ln\overline{\rho}, the last term in equations (31)–(32) is small. This term is proportional to the gradient of the mean entropy ∇S‾=cv[∇ln⁡T‾−(γ−1)∇ln⁡ρ‾]\nabla\overline{S}=c_v[\nabla\ln\overline{T}-(\gamma-1)\nabla\ln\overline{\rho}].

Application to the solar convective zone

In this section, we apply the obtained results to the solar convective zone. We use the model of the solar convective zone by Spruit (1974) based on the mixing length theory. According to this model, the mean penetration length of radiation is much less than the turbulent correlation length. Indeed, in Figs. 1–3 we show the radial profiles of the ratio of the mean penetration length of radiation to the integral scale of turbulence Lr(r)/ℓ0L_r(r)/\ell_0, the ratio ℓm/Hρ\ell_m/H_\rho of the mixing length ℓm\ell_m to the density stratification length Hρ=λ−1H_\rho=\lambda^{-1} and the Reynolds number Re⁡(r)=u0ℓ0/ν\operatorname{Re}(r)=u_0\ell_0/\nu for the solar convective zone based on the model by Spruit (1974). The radius rr is measured in units of the solar radius R⊙R_\odot. The mixing length ℓm\ell_m is identified with the size of the solar granulations, while the ratio ℓm/ℓ0=5−7\ell_m/\ell_0=5-7 is justified by the results of analytical study (Elperin et al. 2002, 2006) and laboratory experiments [10], which show that the integral scale ℓ0\ell_0 of the turbulent convection is smaller in 5–7 times in comparison with the size of the large-scale circulations. These turbulent parameters increase towards the solar surface.

To determine the effective penetration length of radiation, we estimate the degree of compressibility of the turbulent fluid flow for small Mach numbers as

Image not extracted

Figure 3. The radial profile of the Reynolds number Re⁡(r)=u0ℓ0/ν\operatorname{Re}(r)=u_0\ell_0/\nu for the solar convective zone.

Rogachevskii & Kleeorin (2021)

σc∼Ma⁡5Re⁡1/4,(33)\sigma_c\sim\operatorname{Ma}^5\operatorname{Re}^{1/4}, \tag*{(33)}

where Ma⁡=u0/cs\operatorname{Ma}=u_0/c_s is the Mach number, u0=⟨u2⟩1/2u_0=\langle u^2\rangle^{1/2} and cs=(γP‾/ρ‾)1/2c_s=(\gamma\overline{P}/\overline{\rho})^{1/2} is the sound speed, Re⁡=u0ℓ0/ν\operatorname{Re}=u_0\ell_0/\nu is the Reynolds number and ν\nu is the kinematic viscosity. The estimate (33) is obtained assuming that the effect of compressibility to the viscous heating Jν(c)J_\nu^{(c)} is of the order of the radiative wave energy density EwE_w. In particular, turbulence can generate acoustic waves, and the rate of the energy radiated by the acoustic waves per unit mass for small Mach numbers is given by [28, 29, 37]

Ew=⟨u2⟩τ0Ma⁡5,(34)E_w=\frac{\langle u^2\rangle}{\tau_0}\operatorname{Ma}^5, \tag*{(34)}

where τ0=ℓ0/u0\tau_0=\ell_0/u_0 is the turbulent correlation time. The compressibility contribution Jν(c)J_\nu^{(c)} to the rate of the viscous heating is (Rogachevskii & Kleeorin 2021)

Jν(c)=⟨u2⟩τ0σc1+σcRe⁡−1/4.(35)J_\nu^{(c)}=\frac{\langle u^2\rangle}{\tau_0}\frac{\sigma_c}{1+\sigma_c}\operatorname{Re}^{-1/4}. \tag*{(35)}

Equations (34)–(35) yield the estimate (33) for the degree of compressibility σc\sigma_c for small Mach numbers. In Figs. 4–5 we show the radial profiles of the Mach number Ma⁡(r)=u0/cs\operatorname{Ma}(r)=u_0/c_s and the degree of compressibility σc\sigma_c of the fluid velocity field for the solar convective zone based on the model by Spruit (1974). The degree of compressibility σc\sigma_c increases to the surface because the decrease of the sound speed in the vicinity of the solar surface. In Fig. 6 we plot the radial profile of the r.m.s. of temperature fluctuations σrms\sigma_{\mathrm{rms}} measured in the units of the mean temperature T‾\overline{T} using the above parameters for the solar convective zone. Temperature fluctuations increase towards to the solar surface due to the compressibility effects.

Similar behaviour is observed for fluid density fluctuations ⟨ρ′2⟩\langle\rho'^2\rangle and the temperature-density correlations ⟨θρ′⟩\langle\theta\rho'\rangle [see equations (29)–(30)]. In particular, these second moments enhance towards to the solar surface, resulting in increase of fluctuations of the radiation absorption coefficient and the effective penetration length of radiation. This seen in Fig. 7, where we show the radial profile of the ratio of the turbulence induced effective penetration length of radiation to the mean radiation penetration length Leff/LrL_{\mathrm{eff}}/L_r for the solar convective zone based on the model by Spruit (1974). The ratio Leff/LrL_{\mathrm{eff}}/L_r increases in 2.5 times in the vicinity of the solar surface.

Image not extracted

Figure 4. The radial profile of the Mach number Ma⁡(r)=u0/cs\operatorname{Ma}(r)=u_0/c_s for the solar convective zone.

Image not extracted

Figure 5. The radial profile of the degree of compressibility σc\sigma_c for the solar convective zone.

Conclusions

We study a turbulent transport of radiation in the solar convective zone. To this end, we derive a mean-field equation for the irradiation intensity and show that due to the turbulent effects the effective penetration length of radiation is increased in several times in comparison with the mean penetration length of radiation that is defined as an inverse mean absorption coefficient. To demonstrate this effect, we adopt a model of the solar convective zone based on the mixing length theory. The mean penetration length of radiation in this model is much smaller than the turbulent integral scale. We have shown that the ratio of the effective penetration length of radiation to the mean penetration length of radiation is increased in 2.5 times in the vicinity of the solar surface.

This effect can be explained by the compressibility effects that become important in the vicinity of the solar surface, so that the level of temperature and density fluctuations is increased towards the solar surface. It causes an increase of fluctuations of the radiation absorption coefficient and the effective penetration length of radiation. Since the effective penetration length of radiation is changed only in

Image not extracted

Figure 6. The radial profile of the r.m.s. of temperature fluctuations θrms\theta_{\mathrm{rms}} measured in the units of the mean temperature TT for the solar convective zone.

Image not extracted

Figure 7. The radial profile of the ratio Leff/LrL_{\mathrm{eff}}/L_r for a=b=1a=b=1 for the solar convective zone.

the vicinity of the solar surface (at the depth ∼2000\sim2000 km), the effect of turbulence on the radiation transport is not strong for the solar type stars. However, this effect can be essential for cold stars (like M3-M5 stars), for which Mach number is larger than that for the sun.

Data availability

There are no new data associated with this article.

Temperature and density fluctuations

In this Appendix we derive expression for the level of temperature fluctuations ⟨θ2⟩\langle\theta^2\rangle, density fluctuations ⟨ρ′2⟩\langle\rho^{\prime2}\rangle as well as temperature–density correlations ⟨θρ′⟩\langle\theta\rho^\prime\rangle using the method described by Rogachevskii et al. [42]; ); Rogachevskii & Kleeorin [41]; ); Rogachevskii [40]. The temperature field T(t,r)T(t,\mathbf{r}) in a compressible fluid velocity field U(t,r)\mathbf{U}(t,\mathbf{r}) is described by Landau & Lifshits [24]

∂T∂t+(U⋅∇)T+(γ−1)T(∇⋅U)=DΔT+Jν(36)\frac{\partial T}{\partial t}+(\mathbf{U}\cdot\nabla)T+(\gamma-1)T(\nabla\cdot\mathbf{U})=D\Delta T+J_\nu \tag*{(36)}

where DD is the molecular thermal conductivity, γ=cp/cv\gamma=c_p/c_v is the ratio of specific heats and JνJ_\nu is the heating source due to a viscous dissipation.

We study turbulent flows with large Reynolds (Re⁡=u0ℓ0/ν≫1\operatorname{Re}=u_0\ell_0/\nu\gg1) and Péclet (Pe⁡=u0ℓ0/D≫1\operatorname{Pe}=u_0\ell_0/D\gg1) numbers, where u0u_0 is the characteristic turbulent velocity in the integral scale ℓ0\ell_0 of turbulence. Equations for the intensity of temperature fluctuations is derived by means of the mean-field approach, where the temperature T=T‾+θT=\overline{T}+\theta, pressure P=P‾+pP=\overline{P}+p, density ρ=ρ‾+ρ′\rho=\overline{\rho}+\rho^\prime and velocity U=U‾+u\mathbf{U}=\overline{\mathbf{U}}+\mathbf{u} are decomposed into mean and fluctuating parts, with T‾=⟨T⟩\overline{T}=\langle T\rangle being the mean fluid temperature, P‾=⟨P⟩\overline{P}=\langle P\rangle being the mean fluid pressure, ρ‾=⟨ρ⟩\overline{\rho}=\langle\rho\rangle being the mean fluid density, and U‾=⟨U⟩\overline{\mathbf{U}}=\langle\mathbf{U}\rangle being the mean fluid velocity. Here θ\theta, pp, ρ′\rho^\prime and u\mathbf{u} are fluctuations of temperature, pressure, density and velocity, respectively, and the angular brackets denote an ensemble averaging. Application of the mean-field approach implies that there is a separation of spatial (ℓ0≪HT\ell_0\ll H_T) and temporal (τ0≪τT\tau_0\ll\tau_T) scales, where HTH_T and τT\tau_T are the characteristic spatial and temporal scales characterizing the variations of the mean temperature field, and τ0=ℓ0/u0\tau_0=\ell_0/u_0.

Ensemble averaging of equation (A1) yields the mean temperature field:

∂T‾∂t+∇⋅⟨θu⟩=−(γ−2)⟨θ(∇⋅u)⟩+DΔT‾+J‾ν(37)\frac{\partial\overline{T}}{\partial t}+\nabla\cdot\langle\theta\mathbf{u}\rangle=-(\gamma-2)\langle\theta(\nabla\cdot\mathbf{u})\rangle+D\Delta\overline{T}+\overline{J}_\nu \tag*{(37)}

where ⟨θu⟩\langle\theta\mathbf{u}\rangle is the turbulent heat flux, and J‾ν\overline{J}_\nu is the mean heating source caused by the viscous dissipation of the turbulent kinetic energy. Here the case U‾=0\overline{\mathbf{U}}=0 is studied for simplicity. By means of equations (A1) and (A2), we obtain equation for temperature fluctuations, θ(x,t)=T−T‾\theta(x,t)=T-\overline{T}:

∂θ∂t+Q−DΔθ=−(u⋅∇)T‾−(γ−1)T‾∇⋅u(38)\frac{\partial\theta}{\partial t}+Q-D\Delta\theta=-(\mathbf{u}\cdot\nabla)\overline{T}-(\gamma-1)\overline{T}\nabla\cdot\mathbf{u} \tag*{(38)}

where Q=∇⋅(θu−⟨θu⟩)+(γ−2)[θ∇⋅u−⟨θ∇⋅u⟩]Q=\nabla\cdot(\theta\mathbf{u}-\langle\theta\mathbf{u}\rangle)+(\gamma-2)\left[\theta\nabla\cdot\mathbf{u}-\langle\theta\nabla\cdot\mathbf{u}\rangle\right] is the nonlinear term and temperature fluctuations are caused by the source −(u⋅∇)T‾−(γ−1)T‾∇⋅u-(\mathbf{u}\cdot\nabla)\overline{T}-(\gamma-1)\overline{T}\nabla\cdot\mathbf{u}. For simplicity we describe the effect of turbulence on the temperature field, and neglect the feedback effect of the temperature on the turbulence.

We use two-point second-order correlation functions taking into account small-scale properties of the turbulence, where the turbulent correlation time and the turbulent kinetic energy spectrum are related via the Kolmogorov scalings (Monin & Yaglom [33, 34]; McComb [31]; Frisch [18]). We adopt the multi-scale approach (Roberts & Soward [39]), and rewrite the two-point second-order correlation functions as:

⟨θ(x,t)θ(y,t)⟩=∫dk1 dk2 ⟨θ(k1,t)θ(k2,t)⟩×exp⁡[i(k1⋅x+k2⋅y)]=∫Θ(II)(k,R,t)exp⁡[ik⋅r] dk(39)\begin{aligned} \langle\theta(\mathbf{x},t)\theta(\mathbf{y},t)\rangle=\int dk_1\,dk_2\,\langle\theta(\mathbf{k}_1,t)\theta(\mathbf{k}_2,t)\rangle\\ &\times\exp\left[i(\mathbf{k}_1\cdot\mathbf{x}+\mathbf{k}_2\cdot\mathbf{y})\right]=\int\Theta^{(II)}(\mathbf{k},\mathbf{R},t)\exp[i\mathbf{k}\cdot\mathbf{r}]\,dk \tag*{(39)} \end{aligned}

where

Θ(II)(k,R,t)=∫⟨θ(k1,t)θ(k2,t)⟩exp⁡[iK⋅R] dK(40)\Theta^{(II)}(\mathbf{k},\mathbf{R},t)=\int\langle\theta(\mathbf{k}_1,t)\theta(\mathbf{k}_2,t)\rangle\exp[i\mathbf{K}\cdot\mathbf{R}]\,d\mathbf{K} \tag*{(40)}

and we use large-scale variables: R=(x+y)/2R=(x+y)/2, K=k1+k2K=k_1+k_2, as well as small-scale variables: r=x−yr=x-y, k=(k1−k2)/2k=(k_1-k_2)/2. Here k1=k+K/2k_1=k+K/2, and k2=−k+K/2k_2=-k+K/2. Mean-fields depend on the large-scale variables, while fluctuations depend on the small-scale variables.

The procedure of the derivations of the expressions for the intensity of temperature fluctuations implies the following steps:

  • derivation of equations for the second-order moments in the Fourier space using the multi-scale approach;

  • application of the spectral τ\tau approach (see below) which relates the deviations of the third-order moments from those of the background turbulence with the corresponding deviations of the second-order moments;

  • solution of the obtained equations for the second-order moments in the Fourier space;

  • inverse transformation to the physical space to derive expressions for the intensity of temperature fluctuations.

By means of equation (A3) for temperature fluctuations θ\theta and the Navier-Stokes equation for velocity fluctuations u\mathbf{u} rewritten in Fourier space, we derive an equation for the second-order moment ⟨θ(k1)θ(k2)⟩\langle\theta(\mathbf{k}_1)\theta(\mathbf{k}_2)\rangle as

∂∂t⟨θ(k1)θ(k2)⟩=−[⟨ui(k1)θ(k2)⟩+⟨θ(k1)ui(k2)⟩]∇iT‾−(γ−1)[⟨(div⁡u)k1θ(k2)⟩+⟨θ(k1)(div⁡u)k2⟩]T‾+M^Θ(III),(41)\begin{aligned} \frac{\partial}{\partial t}\langle\theta(\mathbf{k}_1)\theta(\mathbf{k}_2)\rangle={}&-\left[\langle u_i(\mathbf{k}_1)\theta(\mathbf{k}_2)\rangle+\langle\theta(\mathbf{k}_1)u_i(\mathbf{k}_2)\rangle\right]\nabla_i\overline{T}\\ &-(\gamma-1)\left[\langle(\operatorname{div}\mathbf{u})_{\mathbf{k}_1}\theta(\mathbf{k}_2)\rangle+\langle\theta(\mathbf{k}_1)(\operatorname{div}\mathbf{u})_{\mathbf{k}_2}\rangle\right]\overline{T}+\widehat{M}_{\Theta}^{(\mathrm{III})}, \tag*{(41)} \end{aligned}

where M^Θ(III)\widehat{M}_{\Theta}^{(\mathrm{III})} are the third-order moment terms caused by the nonlinear terms in the equation for temperature fluctuations. Temperature and velocity fluctuations θ(k1,2,t)\theta(\mathbf{k}_{1,2},t) and ui(k1,2,t)u_i(\mathbf{k}_{1,2},t) depend also on tt, and the mean temperature T‾(t,R)\overline{T}(t,R) depend on tt and RR as well. For brevity of notations, we do not show these dependencies hereafter.

Equation (A6) for the second-order moments includes the third-order moments M^Θ(III)\widehat{M}_{\Theta}^{(\mathrm{III})}, and the closure problem arises, i.e., how to express the third-order moments M^Θ(III)\widehat{M}_{\Theta}^{(\mathrm{III})} through the lower-order moments [33, 34, 31]. We adopt the spectral τ\tau approach which postulates that the deviations of the third-moment terms, M^Θ(III)(k)\widehat{M}_{\Theta}^{(\mathrm{III})}(\mathbf{k}), from those afforded by the background turbulence, M^Θ(III,0)(k)\widehat{M}_{\Theta}^{(\mathrm{III,0})}(\mathbf{k}), can be expressed through similar deviations of the second-order moments, Θ(II)(k)−Θ(II,0)(k)\Theta^{(\mathrm{II})}(\mathbf{k})-\Theta^{(\mathrm{II,0})}(\mathbf{k}) as (Orszag 1970; Pouquet et al. 1976; Kleeorin et al. 1990)

M^Θ(III)(k)−M^Θ(III,0)(k)=−Θ(II)(k)−Θ(II,0)(k)τr(k),(42)\widehat{M}_{\Theta}^{(\mathrm{III})}(\mathbf{k})-\widehat{M}_{\Theta}^{(\mathrm{III,0})}(\mathbf{k})=-\frac{\Theta^{(\mathrm{II})}(\mathbf{k})-\Theta^{(\mathrm{II,0})}(\mathbf{k})}{\tau_r(k)}, \tag*{(42)}

where τr(k)\tau_r(k) is the scale-dependent relaxation time which can be identified with the correlation time τ(k)\tau(k) of the turbulent velocity field for large Reynolds and Péclet numbers. Since the functions with superscript (0) describe the background turbulence with a zero turbulent heat flux, equation (A7) is reduced to M^Θ(III)=−⟨θ(k1)θ(k2)⟩/τ(k)\widehat{M}_{\Theta}^{(\mathrm{III})}=-\langle\theta(\mathbf{k}_1)\theta(\mathbf{k}_2)\rangle/\tau(k). We apply the τ\tau approximation only for the deviations from background turbulence, while the background turbulence is assumed to be known (see below). Validation of the τ\tau approximation for different problems has been performed in various numerical simulations (Brandenburg et al. 2004; Brandenburg & Subramanian 2005a,b,c; Brandenburg et al. 2008; Rogachevskii et al. 2011; Rä edler et al. 2011; Haugen et al. 2012; Rogachevskii et al. 2018; Elperin et al. 2017).

Since the characteristic times of variation of the second-order moment Θ(II)\Theta^{(\mathrm{II})} are much larger than the correlation time τ(k)\tau(k) in all turbulence scales, we use the steady-state solution of equation (A6) as

⟨θ(k1)θ(k2)⟩=−τ(k){[⟨ui(k1)θ(k2)⟩+⟨θ(k1)ui(k2)⟩]∇iT‾+(γ−1)[⟨(div⁡u)k1θ(k2)⟩+⟨θ(k1)(div⁡u)k2⟩]T‾}.(43)\langle\theta(\mathbf{k}_1)\theta(\mathbf{k}_2)\rangle=-\tau(k)\left\{\left[\langle u_i(\mathbf{k}_1)\theta(\mathbf{k}_2)\rangle+\langle\theta(\mathbf{k}_1)u_i(\mathbf{k}_2)\rangle\right]\nabla_i\overline{T}+(\gamma-1)\left[\langle(\operatorname{div}\mathbf{u})_{\mathbf{k}_1}\theta(\mathbf{k}_2)\rangle+\langle\theta(\mathbf{k}_1)(\operatorname{div}\mathbf{u})_{\mathbf{k}_2}\rangle\right]\overline{T}\right\}. \tag*{(43)}

Similarly, we derive expression for the second moments entering in equation (A8), i.e., for ⟨ui(k1)θ(k2)⟩\langle u_i(\mathbf{k}_1)\theta(\mathbf{k}_2)\rangle and ⟨θ(k1)uj(k2)⟩\langle\theta(\mathbf{k}_1)u_j(\mathbf{k}_2)\rangle as

⟨ui(k1)θ(k2)⟩=−τ(k)[⟨ui(k1)uj(k2)⟩∇jT‾+(γ−1)⟨ui(k1)(div⁡u)k2⟩T‾].(44)\langle u_i(\mathbf{k}_1)\theta(\mathbf{k}_2)\rangle=-\tau(k)\left[\langle u_i(\mathbf{k}_1)u_j(\mathbf{k}_2)\rangle\nabla_j\overline{T}+(\gamma-1)\langle u_i(\mathbf{k}_1)(\operatorname{div}\mathbf{u})_{\mathbf{k}_2}\rangle\overline{T}\right]. \tag*{(44)}
⟨θ(k1)uj(k2)⟩=−τ(k)[⟨ui(k1)uj(k2)⟩∇iT‾+(γ−1)⟨(div⁡u)k1uj(k2)⟩T‾].(45)\langle\theta(\mathbf{k}_1)u_j(\mathbf{k}_2)\rangle=-\tau(k)\left[\langle u_i(\mathbf{k}_1)u_j(\mathbf{k}_2)\rangle\nabla_i\overline{T}+(\gamma-1)\langle(\operatorname{div}\mathbf{u})_{\mathbf{k}_1}u_j(\mathbf{k}_2)\rangle\overline{T}\right]. \tag*{(45)}

and for the second moments ⟨(div⁡u)k1θ(k2)⟩\langle(\operatorname{div}\mathbf{u})_{\mathbf{k}_1}\theta(\mathbf{k}_2)\rangle and ⟨θ(k1)(div⁡u)k2⟩\langle\theta(\mathbf{k}_1)(\operatorname{div}\mathbf{u})_{\mathbf{k}_2}\rangle:

⟨(div⁡u)k1θ(k2)⟩=−τ(k)[⟨(div⁡u)k1uj(k2)⟩∇jT‾+(γ−1)⟨(div⁡u)k1(div⁡u)k2⟩T‾].(46)\langle(\operatorname{div}\mathbf{u})_{\mathbf{k}_1}\theta(\mathbf{k}_2)\rangle=-\tau(k)\left[\langle(\operatorname{div}\mathbf{u})_{\mathbf{k}_1}u_j(\mathbf{k}_2)\rangle\nabla_j\overline{T}+(\gamma-1)\langle(\operatorname{div}\mathbf{u})_{\mathbf{k}_1}(\operatorname{div}\mathbf{u})_{\mathbf{k}_2}\rangle\overline{T}\right]. \tag*{(46)}
⟨θ(k1)(div⁡u)k2⟩=−τ(k)[⟨ui(k1)(div⁡u)k2⟩∇iT‾+(γ−1)⟨(div⁡u)k1(div⁡u)k2⟩T‾].(47)\langle\theta(\mathbf{k}_1)(\operatorname{div}\mathbf{u})_{\mathbf{k}_2}\rangle=-\tau(k)\left[\langle u_i(\mathbf{k}_1)(\operatorname{div}\mathbf{u})_{\mathbf{k}_2}\rangle\nabla_i\overline{T}+(\gamma-1)\langle(\operatorname{div}\mathbf{u})_{\mathbf{k}_1}(\operatorname{div}\mathbf{u})_{\mathbf{k}_2}\rangle\overline{T}\right]. \tag*{(47)}

Substituting equations (A9)–(A12) into equation (A8), we obtain

⟨θ(k1)θ(k2)⟩=2τ2(k){⟨ui(k1)uj(k2)⟩(∇iT‾)(∇jT‾)+(γ−1)[⟨(div⁡u)k1uj(k2)⟩+⟨uj(k1)(div⁡u)k2⟩]×T‾∇jT‾+(γ−1)2⟨(div⁡u)k1(div⁡u)k2⟩T‾ 2}.(48)\begin{aligned} \langle\theta(\mathbf{k}_1)\theta(\mathbf{k}_2)\rangle=2\tau^2(k)\bigg\{&\langle u_i(\mathbf{k}_1)u_j(\mathbf{k}_2)\rangle(\nabla_i\overline{T})(\nabla_j\overline{T})\\ &+(\gamma-1)\left[\langle(\operatorname{div}\mathbf{u})_{\mathbf{k}_1}u_j(\mathbf{k}_2)\rangle+\langle u_j(\mathbf{k}_1)(\operatorname{div}\mathbf{u})_{\mathbf{k}_2}\rangle\right]\\ &\times\overline{T}\nabla_j\overline{T}+(\gamma-1)^2\langle(\operatorname{div}\mathbf{u})_{\mathbf{k}_1}(\operatorname{div}\mathbf{u})_{\mathbf{k}_2}\rangle\overline{T}^{\,2}\bigg\}. \tag*{(48)} \end{aligned}

Since all terms in equations (A8) and (A13) are proportional to either (∇T‾)2(\nabla\overline{T})^2, or (∇p‾)2(\nabla\overline{p})^2, or (∇T‾)(∇p‾)(\nabla\overline{T})(\nabla\overline{p}) (see below), and we consider homogeneous and isotropic stratified turbulence, we do not need to perform additional Taylor expansions over small parameters ℓ0/HT\ell_0/H_T and ℓ0/Hρ\ell_0/H_\rho in these terms (Rogachevskii & Kleeorin 2021; Rogachevskii 2021), where HTH_T and HρH_\rho are the characteristic scales of variations of the mean temperature and mean density, respectively. In particular, hereafter we neglect small terms ∼\sim O[(ℓ0/HT)3,(ℓ0/Hρ)3]O[(\ell_0/H_T)^3,(\ell_0/H_\rho)^3]. This implies that we replace k1k_1 by kk and k2k_2 by −k-k in all second moments in equation (48).

In equation (48) we take into account a one way coupling, i.e., we neglect the feedback effect of the mean temperature gradients on the turbulent velocity field. This implies that we replace the correlation function fij=⟨ui(k)uj(−k)⟩f_{ij}=\langle u_i(k)u_j(-k)\rangle in equation (48) by fij(0)=⟨ui(k)uj(−k)⟩(0)f_{ij}^{(0)}=\langle u_i(k)u_j(-k)\rangle^{(0)} for the background turbulence with a zero turbulent heat flux. Similarly we replace ⟨(div⁡u)kuj(−k)⟩\langle(\operatorname{div}u)_k u_j(-k)\rangle, ⟨uj(k)(div⁡u)−k⟩\langle u_j(k)(\operatorname{div}u)_{-k}\rangle and ⟨(div⁡u)k(div⁡u)−k⟩\langle(\operatorname{div}u)_k(\operatorname{div}u)_{-k}\rangle in equation (48) by the corresponding correlation functions for the background turbulence with a zero turbulent heat flux.

To find the intensity of temperature fluctuations ⟨θ2⟩\langle\theta^2\rangle for large Péclet numbers, we adopt a model for the background turbulence, fij(0)(k)=⟨ui(k)uj(−k)⟩(0)f_{ij}^{(0)}(k)=\langle u_i(k)u_j(-k)\rangle^{(0)}, that is a statistically stationary density-stratified compressible turbulence given by [13, 1, 40]:

fij(0)(k)=18πk2(1+σc){E(k)[(δij−kij)(1−λ2k2)+λ2k2(δij−λij)]+1k2[E(k)+σcEc(k)](kjλi−kiλj)+2σcEc(k)kij}⟨u2⟩(49)f_{ij}^{(0)}(k)=\frac{1}{8\pi k^2(1+\sigma_c)}\left\{E(k)\left[(\delta_{ij}-k_{ij})\left(1-\frac{\lambda^2}{k^2}\right)+\frac{\lambda^2}{k^2}(\delta_{ij}-\lambda_{ij})\right]+\frac{1}{k^2}[E(k)+\sigma_c E_c(k)](k_j\lambda_i-k_i\lambda_j)+2\sigma_c E_c(k)k_{ij}\right\}\langle u^2\rangle \tag*{(49)}

where kij=kikj/k2k_{ij}=k_i k_j/k^2, λij=λiλj/λ2\lambda_{ij}=\lambda_i\lambda_j/\lambda^2, and λ=−∇ln⁡ρ‾\boldsymbol{\lambda}=-\nabla\ln\overline{\rho} characterizes the fluid density stratification. This model is different from that derived by Rogachevskii et al. [42]; Rogachevskii & Kleeorin (2021). In particular, this model takes into account a strong density stratification. In addition, the turbulent flux ⟨ρ′u⟩\langle\rho^\prime u\rangle is very small [∼O(λℓ0)3\sim O(\lambda\ell_0)^3].

The background turbulence is of Kolmogorov type with a constant energy flux over the spectrum, i.e., the turbulent kinetic energy spectrum for the incompressible part of turbulence in the inertial range k0<k<kνk_0<k<k_\nu is E(k)=−dτ~(k)/dkE(k)=-d\widetilde{\tau}(k)/dk. Here τ~(k)=(k/k0)1−q\widetilde{\tau}(k)=(k/k_0)^{1-q} with 1<q<31<q<3 is the exponent of the turbulent kinetic energy spectrum. Similarly, the turbulent kinetic energy spectrum for the compressible part of turbulence is Ec(k)=−dτ~c(k)/dkE_c(k)=-d\widetilde{\tau}_c(k)/dk, where τ~c(k)=(k/k0)1−qc\widetilde{\tau}_c(k)=(k/k_0)^{1-q_c} with 1<qc<31<q_c<3. For example, the exponent of the incompressible part of the spectrum, q=5/3q=5/3, corresponds to the Kolmogorov spectrum, while the exponent of the compressible part of the spectrum, qc=2q_c=2, describes the Burgers turbulence with shock waves. These exponents of the spectra are observed in numerical simulations in compressible turbulence [23, 17]. The correlation time for a compressible turbulence in the Fourier space is (Rogachevskii & Kleeorin 2021)

τ(k)=2τ01+σc[τ~(k)+σcτ~c(k)].(50)\tau(k)=\frac{2\tau_0}{1+\sigma_c}\left[\widetilde{\tau}(k)+\sigma_c\widetilde{\tau}_c(k)\right]. \tag*{(50)}

To determine the level of temperature fluctuations ⟨θ2⟩=∫τ2(k)⟨θ(k)θ(−k)⟩ dk\langle\theta^2\rangle=\int\tau^2(k)\langle\theta(k)\theta(-k)\rangle\,dk for large Péclet numbers, we use equations (A13)–(A15). To this end, we calculate the following integrals:

∫τ2(k)⟨ui(k)uj(−k)⟩(0) dk=49ℓ02δij,(51)\int\tau^2(k)\langle u_i(k)u_j(-k)\rangle^{(0)}\,dk=\frac{4}{9}\ell_0^2\delta_{ij}, \tag*{(51)}
∫τ2(k)⟨ui(k)(div⁡u)−k⟩(0) dk=49ℓ02λi,(52)\int\tau^2(k)\langle u_i(k)(\operatorname{div}u)_{-k}\rangle^{(0)}\,dk=\frac{4}{9}\ell_0^2\lambda_i, \tag*{(52)}
∫τ2(k)⟨(div⁡u)k(div⁡u)−k⟩(0) dk=49(ℓ0λ)2+4fc(σc1+σc)3[1−19(ℓ0λ)2].(53)\int\tau^2(k)\langle(\operatorname{div}u)_k(\operatorname{div}u)_{-k}\rangle^{(0)}\,dk=\frac{4}{9}(\ell_0\lambda)^2+4f_c\left(\frac{\sigma_c}{1+\sigma_c}\right)^3\left[1-\frac{1}{9}(\ell_0\lambda)^2\right]. \tag*{(53)}

For the integration in kk space in equations (A16)–(A18), we use the following identities:

∫k0kντ2(k)[E(k)+σcEc(k)] dk=43τ02(1+σc),(54)\int_{k_0}^{k_\nu}\tau^2(k)[E(k)+\sigma_c E_c(k)]\,dk=\frac{4}{3}\tau_0^2(1+\sigma_c), \tag*{(54)}
∫k0kντ2(k)k2Ec(k) dk=4fc(τ0ℓ0)2(σc1+σc)2.(55)\int_{k_0}^{k_\nu}\tau^2(k)k^2E_c(k)\,dk=4f_c\left(\frac{\tau_0}{\ell_0}\right)^2\left(\frac{\sigma_c}{1+\sigma_c}\right)^2. \tag*{(55)}

Therefore, equations (A13) and (A16)–(A18) yield the level of temperature fluctuations ⟨θ2⟩\langle\theta^2\rangle for large Péclet numbers given by equation (26).

Derivation of equations (29) and (30) for ⟨ρ′2⟩\langle\rho^{\prime2}\rangle and ⟨θρ′⟩\langle\theta\rho^\prime\rangle is performed in a similar way. In particular, using the continuity equation for the fluid density fluctuations ρ′\rho^\prime written in the Fourier space, we obtain the evolutionary equation for the second moment ⟨ρ′(k1)ρ′(k2)⟩\langle\rho^\prime(k_1)\rho^\prime(k_2)\rangle as

∂∂t⟨ρ′(k1)ρ′(k2)⟩=−[⟨ui(k1)ρ′(k2)⟩+⟨ρ′(k1)ui(k2)⟩]∇iρ‾−[⟨(div⁡u)k1ρ(k2)⟩+⟨ρ(k1)(div⁡u)k2⟩]ρ‾+M^ρ(III),(56)\frac{\partial}{\partial t}\langle\rho^\prime(k_1)\rho^\prime(k_2)\rangle=-\left[\langle u_i(k_1)\rho^\prime(k_2)\rangle+\langle\rho^\prime(k_1)u_i(k_2)\rangle\right]\overline{\nabla_i\rho}-\left[\langle(\operatorname{div}u)_{k_1}\rho(k_2)\rangle+\langle\rho(k_1)(\operatorname{div}u)_{k_2}\rangle\right]\overline{\rho}+\widehat{\mathcal{M}}_\rho^{(III)}, \tag*{(56)}

where M^ρ(III)\widehat{\mathcal{M}}_\rho^{(III)} are the third-order moment terms related to nonlinear terms in the equation for density fluctuations.

Applying the spectral τ\tau approach, we obtain expression for the second moment ⟨ρ′(k1)ρ′(k2)⟩\langle\rho^\prime(k_1)\rho^\prime(k_2)\rangle as

⟨ρ′(k1)ρ′(k2)⟩=−τ(k){[⟨ui(k1)ρ′(k2)⟩+⟨ρ′(k1)ui(k2)⟩]∇iρ‾+[⟨(div⁡u)k1ρ′(k2)⟩+⟨ρ′(k1)(div⁡u)k2⟩]ρ‾}.(57)\langle\rho^\prime(k_1)\rho^\prime(k_2)\rangle=-\tau(k)\left\{\left[\langle u_i(k_1)\rho^\prime(k_2)\rangle+\langle\rho^\prime(k_1)u_i(k_2)\rangle\right]\overline{\nabla_i\rho}+\left[\langle(\operatorname{div}u)_{k_1}\rho^\prime(k_2)\rangle+\langle\rho^\prime(k_1)(\operatorname{div}u)_{k_2}\rangle\right]\overline{\rho}\right\}. \tag*{(57)}

Similarly, we derive expression for the second moments ⟨ui(k1)ρ′(k2)⟩\langle u_i(k_1)\rho^\prime(k_2)\rangle and ⟨ρ′(k1)uj(k2)⟩\langle\rho^\prime(k_1)u_j(k_2)\rangle as

⟨ui(k1)ρ′(k2)⟩=−τ(k)[⟨ui(k1)uj(k2)⟩∇jρ‾+⟨ui(k1)(div⁡u)k2⟩ρ‾],(58)\langle u_i(k_1)\rho^\prime(k_2)\rangle=-\tau(k)\left[\langle u_i(k_1)u_j(k_2)\rangle\overline{\nabla_j\rho}+\langle u_i(k_1)(\operatorname{div}u)_{k_2}\rangle\overline{\rho}\right], \tag*{(58)}
⟨ρ′(k1)uj(k2)⟩=−τ(k)[⟨ui(k1)uj(k2)⟩∇iρ‾+⟨(div⁡u)k1uj(k2)⟩ρ‾],(59)\langle\rho^\prime(k_1)u_j(k_2)\rangle=-\tau(k)\left[\langle u_i(k_1)u_j(k_2)\rangle\overline{\nabla_i\rho}+\langle(\operatorname{div}u)_{k_1}u_j(k_2)\rangle\overline{\rho}\right], \tag*{(59)}

and for ⟨(div⁡u)k1ρ′(k2)⟩\langle(\operatorname{div}u)_{k_1}\rho^\prime(k_2)\rangle and ⟨ρ′(k1)(div⁡u)k2⟩\langle\rho^\prime(k_1)(\operatorname{div}u)_{k_2}\rangle:

⟨(div⁡u)k1ρ′(k2)⟩=−τ(k)[⟨(div⁡u)k1uj(k2)⟩∇jρ‾+⟨(div⁡u)k1(div⁡u)k2⟩ρ‾].(60)\langle(\operatorname{div}u)_{k_1}\rho^\prime(k_2)\rangle=-\tau(k)\left[\langle(\operatorname{div}u)_{k_1}u_j(k_2)\rangle\overline{\nabla_j\rho}+\langle(\operatorname{div}u)_{k_1}(\operatorname{div}u)_{k_2}\rangle\overline{\rho}\right]. \tag*{(60)}
⟨ρ′(k1)(div⁡u)k2⟩=−τ(k)[⟨ui(k1)(div⁡u)k2⟩∇iρ‾+⟨(div⁡u)k1(div⁡u)k2⟩ρ‾].(61)\left\langle\rho'(\boldsymbol{k}_1)(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_2}\right\rangle =-\tau^{(k)}\left[\left\langle u_i(\boldsymbol{k}_1)(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_2}\right\rangle\nabla_i\overline{\rho} +\left\langle(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_1}(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_2}\right\rangle\overline{\rho}\right]. \tag*{(61)}

Substituting equations (A23)–(A26) into equation (A22) we obtain

⟨ρ′(k1)ρ′(k2)⟩=2τ2(k){⟨ui(k1)uj(k2)⟩(∇iρ‾)(∇jρ‾)+[⟨(div⁡u)k1uj(k2)⟩+⟨uj(k1)(div⁡u)k2⟩]ρ‾∇jρ‾+⟨(div⁡u)k1(div⁡u)k2⟩ρ‾ 2}.(62)\begin{aligned} \left\langle\rho'(\boldsymbol{k}_1)\rho'(\boldsymbol{k}_2)\right\rangle ={}&2\tau^2(k)\bigg\{\left\langle u_i(\boldsymbol{k}_1)u_j(\boldsymbol{k}_2)\right\rangle(\nabla_i\overline{\rho})(\nabla_j\overline{\rho})\\ &+\left[\left\langle(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_1}u_j(\boldsymbol{k}_2)\right\rangle +\left\langle u_j(\boldsymbol{k}_1)(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_2}\right\rangle\right]\overline{\rho}\nabla_j\overline{\rho}\\ &+\left\langle(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_1}(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_2}\right\rangle\overline{\rho}^{\,2}\bigg\}. \tag*{(62)} \end{aligned}

We do not take into account small terms ∼O[(ℓ0/Hρ)3]\sim O[(\ell_0/H_\rho)^3] in equations (A22) and (A27). Since all terms in equations (A22) and (A27) are proportional to (∇ρ‾)2(\nabla\overline{\rho})^2 and we consider homogeneous density stratified turbulence, we do not need to perform additional Taylor expansions over small parameter ℓ0/Hρ\ell_0/H_\rho in these terms. Thus, we replace k1\boldsymbol{k}_1 by k\boldsymbol{k} and k2\boldsymbol{k}_2 by −k-\boldsymbol{k} in all second moments in equation (A27). Using equations (A16)–(A18) and (A27), we determine the level of density fluctuations ⟨ρ′ 2⟩=∫τ2(k)⟨ρ′(k)ρ′(−k)⟩ dk\langle\rho'^{\,2}\rangle=\int\tau^2(k)\langle\rho'(\boldsymbol{k})\rho'(-\boldsymbol{k})\rangle\,d\boldsymbol{k} for large Péclet numbers given by equation (29).

Now we derive the evolutionary equation for the second moment ⟨θ(k1)ρ′(k2)⟩\langle\theta(\boldsymbol{k}_1)\rho'(\boldsymbol{k}_2)\rangle using the continuity equation for the fluid density fluctuations and the equation for the temperature fluctuations written in the Fourier space:

∂∂t⟨θ(k1)ρ′(k2)⟩=−⟨ui(k1)ρ′(k2)⟩∇iT‾−⟨θ(k1)uj(k2)⟩∇jρ‾−(γ−1)⟨(div⁡u)k1ρ′(k2)⟩T‾−⟨θ(k1)(div⁡u)k2⟩ρ‾+M^ρ(III).(63)\begin{aligned} \frac{\partial}{\partial t}\left\langle\theta(\boldsymbol{k}_1)\rho'(\boldsymbol{k}_2)\right\rangle ={}&-\left\langle u_i(\boldsymbol{k}_1)\rho'(\boldsymbol{k}_2)\right\rangle\nabla_i\overline{T}\\ &-\left\langle\theta(\boldsymbol{k}_1)u_j(\boldsymbol{k}_2)\right\rangle\nabla_j\overline{\rho} -(\gamma-1)\left\langle(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_1}\rho'(\boldsymbol{k}_2)\right\rangle\overline{T}\\ &-\left\langle\theta(\boldsymbol{k}_1)(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_2}\right\rangle\overline{\rho} +\widehat{\mathcal{M}}_{\rho}^{(III)}. \tag*{(63)} \end{aligned}

where M^ρ(III)\widehat{\mathcal{M}}_{\rho}^{(III)} are the third-order moment terms related to the nonlinear terms in the equations for temperature and density fluctuations. Applying the spectral τ\tau approach, we obtain expression for the second moment ⟨θ(k1)ρ′(k2)⟩\langle\theta(\boldsymbol{k}_1)\rho'(\boldsymbol{k}_2)\rangle as

⟨θ(k1)ρ′(k2)⟩=−τ(k)[⟨ui(k1)ρ′(k2)⟩∇iT‾+⟨θ(k1)uj(k2)⟩∇jρ‾+(γ−1)⟨(div⁡u)k1ρ′(k2)⟩T‾+⟨θ(k1)(div⁡u)k2⟩ρ‾].(64)\left\langle\theta(\boldsymbol{k}_1)\rho'(\boldsymbol{k}_2)\right\rangle =-\tau(k)\left[ \left\langle u_i(\boldsymbol{k}_1)\rho'(\boldsymbol{k}_2)\right\rangle\nabla_i\overline{T} +\left\langle\theta(\boldsymbol{k}_1)u_j(\boldsymbol{k}_2)\right\rangle\nabla_j\overline{\rho} +(\gamma-1)\left\langle(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_1}\rho'(\boldsymbol{k}_2)\right\rangle\overline{T} +\left\langle\theta(\boldsymbol{k}_1)(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_2}\right\rangle\overline{\rho} \right]. \tag*{(64)}

Substituting equations (A10), (A12), (A23), (A25) into equation (A29), we obtain

⟨θ(k1)ρ′(k2)⟩=2τ2(k){⟨ui(k1)uj(k2)⟩(∇iT‾)(∇jρ‾)+(γ−1)[⟨(div⁡u)k1(div⁡u)k2⟩ρ‾+⟨(div⁡u)k1uj(k2)⟩(∇jρ‾)]T‾+⟨ui(k1)(div⁡u)k2⟩ρ‾(∇iT‾)}.(65)\begin{aligned} \left\langle\theta(\boldsymbol{k}_1)\rho'(\boldsymbol{k}_2)\right\rangle =2\tau^2(k)\bigg\{& \left\langle u_i(\boldsymbol{k}_1)u_j(\boldsymbol{k}_2)\right\rangle(\nabla_i\overline{T})(\nabla_j\overline{\rho})\\ &+(\gamma-1)\left[\left\langle(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_1}(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_2}\right\rangle\overline{\rho} +\left\langle(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_1}u_j(\boldsymbol{k}_2)\right\rangle(\nabla_j\overline{\rho})\right]\overline{T}\\ &+\left\langle u_i(\boldsymbol{k}_1)(\operatorname{div}\boldsymbol{u})_{\boldsymbol{k}_2}\right\rangle\overline{\rho}(\nabla_i\overline{T})\bigg\}. \tag*{(65)} \end{aligned}

We neglect small terms ∼O[(ℓ0/HT)3,(ℓ0/Hρ)3]\sim O[(\ell_0/H_T)^3,(\ell_0/H_\rho)^3] in equations (A29) and (A30). Since all terms in equations (A29) and (A30) are proportional to either (∇ρ‾)2(\nabla\overline{\rho})^2 or (∇T‾)(∇ρ‾)(\nabla\overline{T})(\nabla\overline{\rho}), and we consider homogeneous density stratified turbulence, we do not need to perform additional Taylor expansions over small parameters ℓ0/HT\ell_0/H_T and ℓ0/Hρ\ell_0/H_\rho in these terms.

Therefore, we replace k1\boldsymbol{k}_1 by k\boldsymbol{k} and k2\boldsymbol{k}_2 by −k-\boldsymbol{k} in all second moments in equation (A30). Using equations (A16)–(A18) and (A30), we determine ⟨θρ′⟩=∫τ2(k)⟨θ(k)ρ′(−k)⟩ dk\langle\theta\rho'\rangle=\int\tau^2(k)\langle\theta(\boldsymbol{k})\rho'(-\boldsymbol{k})\rangle\,d\boldsymbol{k} for large Péclet numbers given by equation (30).

References

  1. [1]Amir, G., Bar, N., Eidelman, A., Elperin, T., Kleeorin, N., Rogachevskii, I., 2017, Phys. Rev. Fluids, 2, 064605
  2. [2]Apresyan, L. A., Kravtsov, Y. A., 1996, Radiation Transfer. Gordon and Breach Publ., Amsterdam
  3. [3]Barekat, A., Brandenburg, A., 2014, Astron. Astrophys., 571, A68
  4. [4]Brandenburg, A., Das, U., 2021, Turbulent radiative diffusion and turbulent Newtonian cooling, ArXiv: 2010.07046DOI
  5. [10]Bukai, M., Eidelman, A., Elperin, T., Kle​eorin, N., Rogachevskii, I., Sapir-Katiraie, I., 2009, Phys. Rev. E, 79, 066302
  6. [11]Chandrasekhar, S., 1960, Radiative transfer. Dover, New York
  7. [12]Davidson, P. A., 2013, Turbulence in Rotating, Stratified and Electrically Conducting Fluids. Cambridge University Press, CambridgeDOI
  8. [13]Elperin, T., Kleeorin, N., Rogachevskii, I., 1995, Phys. Rev. E, 52, 2617
  9. [17]Federrath, C., 2013, Mon. Not. R. Astron. Soc., 436, 1245
  10. [18]Frisch, U., 1995, Turbulence: the Legacy of A. N. Kolmogorov. Cambridge University Press, Cambridge
  11. [20]Howell, J. R., Menguc, M. P., Siegel, R., 2010, Thermal Radiation Heat Transfer. CRC Press, New York
  12. [23]Kritsuk, A. G., Norman, M. L., Padoan, P., Wagner, R., 2007, Astrophys. J., 665, 416
  13. [24]Landau, L. D., Lifshits, E. M., 1987, Fluid Mechanics. Pergamon, Oxford
  14. [25]Lesieur, M., 2008, Turbulence in Fluids. Springer, Dordrecht
  15. [26]Liberman, M., Kleorin, N., Rogachevskii, I., Haugen, N. E. L., 2017, Phys. Rev. E, 95, 051101
  16. [27]Liberman, M., Kleorin, N., Rogachevskii, I., Haugen, N. E. L., 2018, Combustion Theory and Modelling, 22, 1084
  17. [28]Lighthill, M. J., 1952, Proc. Roy. Soc., A211, No. 1107, 564
  18. [29]Lighthill, M. J., 1954, Proc. Roy. Soc., A222, No. 1148, 1
  19. [30]Liou, K. N., 2002, An Introduction to Atmospheric Radiation. Elsevier Science, New York
  20. [31]McComb, W. D., 1990, The Physics of Fluid Turbulence. Oxford Science Publ., Oxford
  21. [32]Mihalas, D., Mihalas, B. W., 1984, Foundations of Radiation Hydrodynamics. Oxford University Press, New York
  22. [33]Monin, A. S., Yaglom, A. M., 1971, Statistical Fluid Mechanics. MIT Press, Cambridge, Massachusetts, v. 1
  23. [34]Monin, A. S., Yaglom, A. M., 1975, Statistical Fluid Mechanics. MIT Press, Cambridge, Massachusetts, v. 2
  24. [37]Proudman, I., 1952, Proc. Roy. Soc., A214, No. 1116, 119
  25. [39]Roberts, P. H., Soward, A. M., 1975, Astron. Nachr., 296, 49
  26. [40]Rogachevskii, I., 2021, Introduction to Turbulent Transport of Particles, Temperature and Magnetic Fields. Cambridge University Press, Cambridge
  27. [41]Rogachevskii, I., and Kleorin, N., 2021, Phys. Rev. E, 103, 013107
  28. [42]Rogachevskii, I., Kleorin, N., Brandenburg, A., 2018, J. Plasma Phys., 84, 735840502
  29. [44]Spiegel, E. A., 1957, Astrophys. J., 126, 202

Paper details

vlnc:W_gs7tx7cd
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents