Open Access
Issue
A&A
Volume 663, July 2022
Article Number A114
Number of page(s) 6
Section Interstellar and circumstellar matter
DOI https://doi.org/10.1051/0004-6361/202141087
Published online 22 July 2022

© E. Jaupart and G. Chabrier 2022

Licence Creative CommonsOpen Access article, published by EDP Sciences, under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

This article is published in open access under the Subscribe-to-Open model. Subscribe to A&A to support open access publication.

Introduction

In a previous article (Jaupart & Chabrier 2021b, hereafter Paper I) we derived a general framework aimed at assessing the relevance and validity of a statistical approach to characterize various properties of star-forming molecular clouds (MCs) from a limited number of observations or simulations. We calculated the auto-covariance function (ACF) and correlation length of density fields in MCs and provided a way to determine the correct statistical error bars on observed probability density functions (PDFs). Applying these results to two typical starforming clouds, Polaris and Orion B, which display two different types of PDFs, we have shown that the ratio of the correlation length of density fluctuations over the size of the cloud is typically lc/L ≲ 0.1. This justifies the relevance of an approach that uses the hypothesis of statistical homogeneity when exploring star-forming MC properties, notably the evolution of the PDF, as carried out in Jaupart & Chabrier (2020).

In this joint article, we apply the same formalism to a different kind of (observational or numerical) study of MCs. Indeed, as observations cannot fully unravel the entire complexity of the inner structure of star-forming clouds, it is important to know whether global observable estimates, such as the total mass and size of the cloud, can be used to give an accurate estimation of various key global physical quantities that are supposed to characterize the dynamics of the cloud. Of prime importance, for instance, is the correct determination of the total gravitational (binding) energy of a cloud, and its virial parameter, from estimations of its total mass and size only.

2 Correlation Length and Gravitational Binding Energy

In Paper I, we relied on the correlation length to determine confidence intervals for the measured statistical quantities. Here, we follow similar lines of argument to assess the accuracy of estimates of cloud characteristics deduced from global properties without any knowledge of the cloud internal structure. We focus on the cloud potential energy, noted as |eG|, and virial parameter, noted as αvir, which can be deduced from the total mass, M, and size, L. This key issue was raised in particular by Federrath & Klessen (2012, Federrath & Klessen 2013). In their numerical simulations of turbulent star-forming MCs, these authors found a large discrepancy between values of the virial parameter deduced from the cloud’s global (average) characteristics and those measured directly from the numerical results. The authors thus suggested that “comparing simple theoretical estimates of the virial parameter, solely based on the total mass, as a measure for |eG| […], should be considered with great caution because such an estimate ignores the internal structure of the clouds”. We show below that the dependence of the total potential energy and virial parameter on the “internal structure” can be assessed from the knowledge of the correlation length of density fluctuations within the cloud, lc(ρ). We further quantify the difference in |eG| obtained in Federrath & Klessen (2012) in terms of the global properties of the cloud. This difference is primarily due to artificial effects stemming from the resolution of the Poisson equation in a periodic box. Such periodic boundary conditions were known to the authors to result in a significant difference in |eG| but were used to try to mimic the effects of a surrounding medium on the region of interest in their computational boxes.

We first started by deriving the total potential energy, noted as eG 1, of a statistically homogeneous cloud in a domain, Ω, in any geometry, and isolate the contribution of the internal structure from the rest. For the sake of simplicity, we assumed that the cloud possesses a center of symmetry, which we take as the origin, such that y ∈ Ω, −y ∈ Ω: eG = eG(ρ) =1| Ω |Ωρ(x)ΦG(x)dx$\left\langle {{e_G}} \right\rangle = \left\langle {{e_G}\left( \rho \right)} \right\rangle = {1 \over {\left| \Omega \right|}}\,\int_\Omega {\rho \left( x \right){\Phi _G}\left( x \right){\rm{d}}x}$(1) =1| Ω |Ωρ(x)ΦGreen(xx)dx dx= eG( ρ ) +Ω2δρ(x)δρ(x)| Ω |ΦGreen(xx) dx dx$\matrix{ { = {1 \over {\left| \Omega \right|}}\,\int_\Omega {\rho \left( {x'} \right)} \,{\Phi _{{\rm{Green}}}}\left( {x - x'} \right){\rm{d}}x'\,{\rm{d}}x} \hfill \cr { = \left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle + \int_{{\Omega ^2}} {{{\delta \rho \left( x \right)\delta \rho \left( {x'} \right)} \over {\left| \Omega \right|}}} {\Phi _{{\rm{Green}}}}\left( {x - x'} \right)\,{\rm{d}}x'\,{\rm{d}}x} \hfill \cr }$(2) +2 ρ   Ω2δρ(x)| Ω |ΦGreen(xx) dx dx$ + 2\left\langle \rho \right\rangle \,\,\int_{{\Omega ^2}} {{{\delta \rho \left( x \right)} \over {\left| \Omega \right|}}{\Phi _{{\rm{Green}}}}\left( {x - x'} \right)\,{\rm{d}}x'\,{\rm{d}}x}$(3) = eG( ρ ) +IC(ρ)+2 ρ Iδ(ρ),$ = \left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle + {I_C}\left( \rho \right) + 2\left\langle \rho \right\rangle {I_\delta }\left( \rho \right),$(4)

where |Ω| = L3 is the volume of the Ω domain, ΦGreen is Green’s function of the gravitation potential, ΦG, which is parity invariant, 〈ρ〉 is the (volumetric) average density of the cloud, ρ =1| Ω |Ωρ(x)dx=M/| Ω |,$\left\langle \rho \right\rangle = {1 \over {\left| \Omega \right|}}\int_\Omega {\rho \left( x \right)} {\rm{d}}x = {M \mathord{\left/ {\vphantom {M {\left| \Omega \right|}}} \right. \kern-\nulldelimiterspace} {\left| \Omega \right|}},$(5)

δρ = ρ − 〈ρ〉, and 〈 eG(〈ρ〉)〉 is the potential energy of the cloud if it were strictly homogeneous: eG( ρ ) =1| Ω |Ω ρ ΦG(x)dx.$\left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle = {1 \over {\left| \Omega \right|}}\int_\Omega {\left\langle \rho \right\rangle {\Phi _G}\left( x \right){\rm{d}}x.}$(6)

Using the change of fields (u, u) = φ(x, x2) = (xx2, x + x2) and using Ĉ ρ,L to denote the biased ergodic estimator of the ρ ACF (see Paper I), we obtained IC(ρ)=φ1(Ω)duΦGreen(u)φ2u(Ω)dυ8| Ω |δρ(u+υ2)δρ(uυ2)=φ1(Ω)duΦGreen(u)C^ρ,L(u),$\matrix{ {{I_C}\left( \rho \right)} \hfill & { = \,\int_{{\varphi _1}\left( \Omega \right)} {{\rm{d}}u{\Phi _{{\rm{Green}}}}\left( u \right)} \,\int_{\varphi _2^u\left( \Omega \right)} {{{{\rm{d}}{\bf{\upsilon }}} \over {8\left| \Omega \right|}}\delta \rho \left( {{{u + {\bf{\upsilon }}} \over 2}} \right)} \,\delta \rho \,\left( {{{u - {\bf{\upsilon }}} \over 2}} \right)} \hfill \cr {} \hfill & { = \,\int_{{\varphi _1}\left( \Omega \right)} {{\rm{d}}u{\Phi _{{\rm{Green}}}}\left( u \right)\,{{\hat C}_{\rho ,L}}\left( u \right),} } \hfill \cr }$(7) Iδ(ρ)=  φ1(Ω)duΦGreen(u)  φ2u(Ω)dυ8| Ω |δρ(u+υ2)   eG( ρ ) 2 ρ ,$\matrix{ {{I_\delta }\left( \rho \right)} \hfill & { = \,\,\int_{{\varphi _1}\left( \Omega \right)} {{\rm{d}}u} {\Phi _{{\rm{Green}}}}\left( u \right)\,\,\int_{\varphi _2^u\left( \Omega \right)} {{{{\rm{d}}{\bf{\upsilon }}} \over {8\left| \Omega \right|}}\delta \rho \left( {{{u + {\bf{\upsilon }}} \over 2}} \right)} } \hfill \cr {} \hfill & { \ll \,\,{{\left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle } \over {2\left\langle \rho \right\rangle }},} \hfill \cr }$(8)

where (φ1(Ω),φ2u(Ω))$\left( {{\varphi _1}\left( \Omega \right),\varphi _2^{\bf{u}}\left( \Omega \right)} \right)$ is a parameterisation of φ2) (see Appendix A). For example, if  Ω =[ L2,L2 ]3,${\rm{if}}\,\,{\rm{\Omega }}\, = {\left[ { - {L \over 2},{L \over 2}} \right]^3},$(9) φ1(Ω)=[ L,L ]3${\varphi _1}\left( {\rm{\Omega }} \right) = {\left[ { - L,L} \right]^3}$(10)

and φ2u(Ω)=[ L+| ui |,L| ui | ].$\varphi _2^u\left( \Omega \right) = \left[ { - L + \left| {{u_i}} \right|,L - \left| {{u_i}} \right|} \right].$(11)

We then write C^ρ=Var(ρ)×C^ρ${\hat C_\rho } = {\rm{Var}}\left( \rho \right) \times {\hat C_\rho }$ such that C^ρ(0)=1${\hat C_\rho }\left( 0 \right) = 1$ to finally obtain: eG eG( ρ ) +Var(ρ)φ1(Ω)duΦGreen(u)C˜ρ,L(u).$\left\langle {{e_G}} \right\rangle \simeq \left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle + {\rm{Var}}\left( \rho \right)\,\int_{{\varphi _1}\left( \Omega \right)} {{\rm{d}}u{\Phi _{{\rm{Green}}}}\left( u \right)} \,{\tilde C_{\rho ,L}}\left( u \right).$(12)

In this expression, the contributions from the global (average) observables and the internal structure are separated explicitly. We now compare the case of two geometrical configurations, the real space ℝ3, relevant to observations, and the periodic simulation box T3${{\mathbb{T}}^3}$.

2.1 Isolated Cloud

In ℝ3 we have ΦGreen(x) = −G/|x|; hence, eG 3= eG( ρ ) 3GVar(ρ)  φ1(Ω)duC˜ρ,L(u)| u |,${\left\langle {{e_G}} \right\rangle _{{^3}}} = {\left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle _{{^3}}} - G{\rm{Var}}\left( \rho \right)\,\,\int_{{\varphi _1}\left( \Omega \right)} {{\rm{d}}u{{{{\tilde C}_{\rho ,L}}\left( u \right)} \over {\left| u \right|}}} ,$(13)

with eG( ρ ) 3=2GMcg ρ L,${\left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle _{{^3}}} = - 2G\,M{c_g}{{\left\langle \rho \right\rangle } \over L},$(14)

where cg is a geometric factor of order unity if the cloud is roughly of the same dimension L in the three directions. For example if Ω = B(R) is a ball2 of radius R = L/2, cg = 1.2 while if Ω=[ L2,L2 ]3$\Omega = {\left[ { - {L \over 2},{L \over 2}} \right]^3}$ is a cuboid of size L, cg ≃ 1.9/2 ≃ 0.95. Then, φ1(Ω)duC˜ρ,L(u)| u |=8c˜glc(ρ)2f(R/lc(ρ)),$\int_{{\varphi _1}\left( \Omega \right)} {{\rm{d}}u} {{{{\tilde C}_{\rho ,L}}\left( u \right)} \over {\left| u \right|}} = 8{\tilde c_g}{l_c}{\left( \rho \right)^2}f\,\left( {R/{l_c}\left( \rho \right)} \right),$(15)

where c˜g${\widetilde c_{\rm{g}}}$ is a geometric factor of order unity and f some function that converges rapidly towards 1. For an exponential ACF (see Paper I), c˜g=π1/3/30.73${\widetilde c_{\rm{g}}} = {{{\pi ^{{1 \mathord{\left/ {\vphantom {1 3}} \right. \kern-\nulldelimiterspace} 3}}}} \mathord{\left/ {\vphantom {{{\pi ^{{1 \mathord{\left/ {\vphantom {1 3}} \right. \kern-\nulldelimiterspace} 3}}}} 3}} \right. \kern-\nulldelimiterspace} 3} \simeq 0.73$ and f(x) ≃ 1 − (1 + x)e−x. As we have lc/R ≳ 0.1 (see Paper I), we can write f (R/lc(ρ)) ≃ 1 and eG 32GMcg ρ L(1+4Var(ρ)L ρ Mc˜gcglc(ρ)2)${\left\langle {{e_G}} \right\rangle _{{^3}}} \simeq - 2G\,M{c_g}{{\left\langle \rho \right\rangle } \over L}\left( {1 + {{4{\rm{Var}}\left( \rho \right)L\,} \over {\left\langle \rho \right\rangle M}}{{{{\tilde c}_g}} \over {{c_g}}}{l_c}{{\left( \rho \right)}^2}} \right)$(16) 2GMcg ρ L(1+2c˜gcgξg Var(ρ ρ )(lc(ρ)R)2),$\simeq - 2G\,M{c_g}{{\left\langle \rho \right\rangle } \over L}\left( {1 + 2{{{{\tilde c}_g}} \over {{c_g}\,{\xi _g}}}\,{\rm{Var}}\left( {{\rho \over {\left\langle \rho \right\rangle }}} \right){{\left( {{{{l_c}\left( \rho \right)} \over R}} \right)}^2}} \right),$(17)

where, ξg is of order unity (for a ball ξg = π/3 and for a cube ξg = 2). In Paper I we derived a useful relation between the variance of the density field, ρ, the column density, Σ, and the correlation length, lc(ρ), namely: Var( ) Var(ρ ρ )lc(ρ)R,${\rm{Var}}\left( {{\sum \over {\left\langle \sum \right\rangle }}} \right) \simeq \,{\rm{Var}}\left( {{\rho \over {\left\langle \rho \right\rangle }}} \right)\,{{{l_c}\left( \rho \right)} \over R},$(18)

providing that lc/R ≪ 1 (see Eq. (44) in Paper I). Thus, we have eG32GMcgρL(1+2c˜gcgξgvar()(lc(ρ)R)).${\left\langle {{e_G}} \right\rangle _{{^3}}} \simeq - 2G\,M{c_g}{{\left\langle \rho \right\rangle } \over L}\left( {1 + 2{{{{\tilde c}_g}} \over {{c_g}{\xi _g}}}{\mathop{\rm var}} \left( {{\sum \over {\left\langle \sum \right\rangle }}} \right)\left( {{{{l_c}\left( \rho \right)} \over R}} \right)} \right).$(19)

Equation (19) enables us to infer the influence of the internal structure of the cloud from observations of column densities. We thus see that, if the product Var (ρ/ 〈ρ〉) × (lc (ρ)/R)2 ≪ 1 for volume densities, or if the product Var (Σ/ 〈Σ〉) × (lc(ρ)/R) ≪ 1 for column densities, the correction of the cloud’s internal structure contribution to the average gravitational energy is negligible.

2.1.1 Isothermal Compressible Turbulent Conditions

For isothermal turbulent conditions, argued to be representative of initial conditions in star-forming clouds (see, e.g., McKee & Ostriker 2007 and reference therein), Var(ρ)(b)2E(ρ)2(b)2 ρ 2${\rm{Var}}\left( \rho \right) \simeq {\left( {b{\cal M}} \right)^2}{\mathbb{E}}{\left( \rho \right)^2} \simeq {\left( {b{\cal M}} \right)^2}{\left\langle \rho \right\rangle ^2}$ and we get Var(ρρ)(lc(ρ)R)2((bM)lc(ρ)R)2,${\rm{Var}}\left( {{\rho \over {\left\langle \rho \right\rangle }}} \right){\left( {{{{l_c}\left( \rho \right)} \over R}} \right)^2}\, \simeq {\left( {{{\left( {b{\cal M}} \right){l_c}\left( \rho \right)} \over R}} \right)^2},$(20)

where b is the coefficient reflecting the driving mode contributions of the turbulence (Federrath et al. 2008). For typical MC conditions in the Milky Way, (bℳ) < 5 (ℳ ~ 10 and b ≃ 0.5).

Moreover, in Jaupart & Chabrier (2021a), we show that for compressible turbulence without gravity, within a factor of order unity, lc(ρ) ~ λsL/2, where λs is the sonic scale that is found to be close to the average width of filamentary structures in isothermal turbulence (Federrath 2016). However, as mentioned in the above article, while in case of pure gravitationless turbulence the correlation length should be about the sonic length, this is not necessarily the case if gravity initially plays a non-negligible role.

In any case, since as shown in Paper I, lc(ρ)/R ≲ 0.1 (and more likely lc(ρ)/R ~ 10−2) with Var(Σ/ 〈Σ〉) ≲ 1 in these clouds, we conclude that eG32GMcgρL(1+ξg),${\left\langle {{e_G}} \right\rangle _{{^3}}} \simeq - 2G\,M{c_g}{{\left\langle \rho \right\rangle } \over L}\left( {1 + {\xi _g}} \right),$(21)

where ξ˜gξ˜(b,lc(ρ)/R)${\widetilde \xi _{\rm{g}}} \equiv \widetilde \xi \left( {b{\cal M},{l_{\rm{c}}}{{\left( \rho \right)} \mathord{\left/ {\vphantom {{\left( \rho \right)} R}} \right. \kern-\nulldelimiterspace} R}} \right)$ is again at most of order unity for large values of bℳ ~ 10 but more generally is on the order of a few percent.

2.1.2 Late-stage Evolution of Star-forming Clouds

However, if gravity has already started to affect the PDF of the cloud, yielding two extended power-law tails, the variance of p can become very large (see Jaupart & Chabrier 2020 and Sect. 6.3.3 of Paper I). In that case, the product Var(ρ/ 〈ρ〉) × (lc(ρ)/R)2 can become sizeable and can affect the estimate of eG.

Indeed in Jaupart & Chabrier (2021a) the authors show that for a statistically homogeneous density field ρ, E(ρ) Var(ρE(ρ))lc(ρ)3ρ Var(ρρ)lc(ρ)3$\left( \rho \right)\,{\rm{Var}}\left( {{\rho \over {\left( \rho \right)}}} \right){l_c}{\left( \rho \right)^3} \simeq \left\langle \rho \right\rangle \,{\rm{Var}}\left( {{\rho \over {\left\langle \rho \right\rangle }}} \right){l_c}{\left( \rho \right)^3}$(22)

is an invariant of the dynamics of star-forming MCs. Moreover, they show that since this increase in variance due to gravity occurs on a short (local) timescale compared with the typical timescale of variation of E(ρ) ρ ${\mathbb{E}}\left( \rho \right) \simeq \left\langle \rho \right\rangle$ the invariant essentially yields Var(ρρ)lc(ρ)3=const.,${\rm{Var}}\left( {{\rho \over {\left\langle \rho \right\rangle }}} \right)\,{l_c}{\left( \rho \right)^3} = {\rm{const}}{\rm{.,}}$(23)

at least in the initial phase of gravitational collapse. In that case, as gravity proceeds to concentrate gas in smaller clumpier structures, the product Var(ρρ)lc(ρ)2(t)Var(ρρ)1/3(t),${\rm{Var}}\left( {{\rho \over {\left\langle \rho \right\rangle }}} \right)\,{l_c}{\left( \rho \right)^2}\left( t \right) \propto {\rm{Var}}{\left( {{\rho \over {\left\langle \rho \right\rangle }}} \right)^{1/3}}\left( t \right),$(24)

and as such the contribution of internal structures, can become sizeable since Var (ρ/ 〈ρ〉) becomes very large. Physically speaking, this occurs when gravity has started to break the cloud into small condensed and isolated regions.

2.1.3 Concluding Remarks and the Case of Polaris and Orion B

For typical initial star-forming conditions, however, we still expect Var(Σ/(Σ)) × (lc(ρ)/R) ≪ 1, as found in Paper I for Polaris and Orion B, where Var(Σ/(Σ)) × (lc(p)/R) < 10−1 (see Sect. 6 of Paper I).

Therefore, we conclude that for typical initial star-forming cloud conditions, the gravitational potential energy can be correctly estimated from the total mass and size of the cloud. The (observationally undetermined) internal structure of the cloud yields only a small correction to this average value. Hence, most of the uncertainties comes from the geometrical factors in Eq. (14). As seen in the next section, however, this is no longer true for simulations in a periodic box of volume VBox = L3.

2.2 Simulations in a Periodic Box

In a periodic box (topology T3${{\mathbb{T}}^3}$) of volume VBox = L3, the gravitation potential, ΦG, satisfies the modified Poisson equation: ΔΦGL(x)=4πG(ρ(x)MVbox),${\rm{\Delta }}\Phi _G^L\left( x \right) = 4\pi G\left( {\rho \left( x \right) - {M \over {{V_{{\rm{box}}}}}}} \right),$(25)

where M is the total mass in the box (Ricker 2008; Guillet & Teyssier 2011). This is due to the ill-posed problem of the standard Poisson equation in T3${{\mathbb{T}}^3}$ and is a well-known feature of statistical mechanics of Coulomb systems. Then, the Green function of the gravitational potential satisfies ΔxΦGreenL(x)=4πG(δ(x)1Vbox).${{\rm{\Delta }}_x}\Phi _{{\rm{Green}}}^L\left( x \right) = 4\pi G\left( {\delta \left( x \right) - {1 \over {{V_{{\rm{box}}}}}}} \right).$(26)

Rescaling the various fields, y = x/L, one ends up with ΦGreenL(x)=1LΦGreen1(xL)=1LΦGreen1(y),$\Phi _{{\rm{Green}}}^L\left( x \right) = {1 \over L}\Phi _{{\rm{Green}}}^1\left( {{x \over L}} \right) = {1 \over L}\Phi _{{\rm{Green}}}^1\left( y \right),$(27) ΔyΦGreen1(y)=4πG(δ(y)1).${\Delta _y}\Phi _{{\rm{Green}}}^1\left( y \right) = 4\pi G\left( {\delta \left( y \right) - 1} \right).$(28)

We note that ΦGreen1$\Phi _{{\rm{Green}}}^1$ is periodic (of period 1) and defined up to a constant which is usually chosen to be such that the average of ΦGreen1$\Phi _{{\rm{Green}}}^1$ or the zero mode of its Fourier transform, is 0. Then eG(ρ)T3=0,${\left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle _{{^{\rm{3}}}}} = 0,$(29)

and eGT3=L2[12,12]3C˜ρ,L(Ly)ΦGreen1(y)dy.${\left\langle {{e_G}} \right\rangle _{{^3}}} = {L^2}\int_{{{\left[ { - {1 \over 2},{1 \over 2}} \right]}^3}} {{{\tilde C}_{\rho ,L}}\left( {L{\bf{y}}} \right)\Phi _{{\rm{Green}}}^1\left( y \right){\rm{d}}y.}$(30)

Then, in turn, eGT3=L2Var(ρ)[12,12]3C˜ρ,L(Ly)ΦGreen1(y)dy.${\left\langle {{e_G}} \right\rangle _{{^3}}} = {L^2}{\rm{Var}}\left( \rho \right)\int_{{{\left[ { - {1 \over 2},{1 \over 2}} \right]}^3}} {{{\tilde C}_{\rho ,L}}\left( {L{\bf{y}}} \right)\Phi _{{\rm{Green}}}^1\left( y \right){\rm{d}}y.}$(31)

In this case, therefore, the gravitational potential is only a measure of the internal, fluctuating density structure within the box.

In the following, we examine the case of pure turbulent (initial) conditions (without gravity), as in Federrath & Klessen (2012, Federrath & Klessen 2013). We thus write eGT3=2G Var(ρ)lc2gb,M(L/lc),${\left\langle {{e_G}} \right\rangle _{{^3}}} = 2G\,{\rm{Var}}\left( \rho \right)l_{\rm{c}}^2{g_{b,{\cal M}}}\left( {L/{l_{\rm{c}}}} \right),$(32)

where gb,ℳ is some bounded dimensionless function that may depend on the Mach number ℳ and the type of turbulence forcing, as measured by coefficient b. Noting that Var(ρ) ≃ (bℳ)2ρ2, we obtain eGT3=2G(bM)2ρ2lc2gb,M(L/lc)${\left\langle {{e_G}} \right\rangle _{{^3}}} = 2G\,{\left( {b{\cal M}} \right)^2}{\left\langle \rho \right\rangle ^2}l_{\rm{c}}^2{g_{b,{\cal M}}}\left( {L/{l_{\rm{c}}}} \right)$(33) =2GMρL((bM)lcL)2gb,M(L/lc).$ = 2G\,M{{\left\langle \rho \right\rangle } \over L}{\left( {{{\left( {b\,{\cal M}} \right){l_c}} \over L}} \right)^2}{g_{b,{\cal M}}}\left( {L/{l_{\rm{c}}}} \right).$(34)

3 Virial Parameter

An important quantity in the study of MCs is the virial parameter, which is defined as the ratio of twice the kinetic energy over the gravitational energy, αvir = 2 〈eK/|<eG〉| (see e.g. McKee & Zweibel 1992 for a more complete discussion). For clarity purposes, we again use the simulations of Federrath & Klessen (2012) for comparisons in T3${{\mathbb{T}}^3}$ and restrict ourselves to isothermal turbulence conditions. In these conditions, the following conditional expectation holds (Kritsuk et al. 2007; Federrath et al. 2010): E(M|ρ)E(M),$\left( {{\cal M}|\rho } \right) \simeq \left( {\cal M} \right),$(35)

and we get eKρσV2E(ρ)σV2,$\left\langle {{e_{\rm{K}}}} \right\rangle \simeq \left\langle \rho \right\rangle \sigma _{\rm{V}}^2 \simeq \left( \rho \right)\sigma _{\rm{V}}^2,$(36)

where σV is the 3D velocity dispersion. This yields the virial parameters αvir,3=σv2L2GM(1+ξ˜g)cg=σv2L2GM1(1+ξ˜g)cg,${\alpha _{{\rm{vir,}}{^3}}} = {{\sigma _{\rm{v}}^2\,L} \over {2G\,M\left( {1 + {{\tilde \xi }_{\rm{g}}}} \right){c_{\rm{g}}}}} = {{\sigma _{\rm{v}}^2L} \over {2G\,M}}{1 \over {\left( {1 + {{\tilde \xi }_g}} \right){c_{\rm{g}}}}},$(37)

where 2ξ˜gcg$2{\widetilde \xi _{\rm{g}}}{c_{\rm{g}}}$ is usually taken to be equal to 1 in order to match the virial parameter with that of a homogeneous sphere, and αvir,T3=σv2L2GMgb,M(L/lc)(L(bM)lc)2.${\alpha _{{\rm{vir,}}{^3}}} = {{\sigma _{\rm{v}}^2\,L} \over {2G\,M{g_{b,{\cal M}}}\left( {L/{l_{\rm{c}}}} \right)}}\,{\left( {{L \over {\left( {b{\cal M}} \right){l_{\rm{c}}}}}} \right)^2}.$(38)

The ratio of the two virial parameters is therefore αvir,T3αvir,31gb,M(L/lc)(L(bM)lc)2.${{{\alpha _{{\rm{vir,}}{^3}}}} \over {{\alpha _{{\rm{vir,}}{^3}}}}} \simeq {1 \over {{g_{b,{\cal M}}}\left( {L/{l_{\rm{c}}}} \right)}}\,{\left( {{L \over {\left( {b{\cal M}} \right){l_{\rm{c}}}}}} \right)^2}.$(39)

We can test the validity of this equation. Assuming that the type of turbulence forcing has only a moderate influence on C˜ρ,L${\widetilde C_{\rho ,L}}$ and thus on gb,(L/lc)lc2${g_{b,{\cal M}}}\left( {{L \mathord{\left/ {\vphantom {L {{l_{\rm{c}}}}}} \right. \kern-\nulldelimiterspace} {{l_{\rm{c}}}}}} \right)l_{\rm{c}}^2$ (see Eq. (32)), we have, for a given largescale Mach number, ℳ, and size L: αvir,T3/αvir,3b2.${\alpha _{{\rm{vir,}}{^3}}}/{\alpha _{{\rm{vir,}}{^3}}}\, \propto \,{b^{ - 2}}.$(40)

Even though the forcing parameter, b, may have some moderate influence on gb,(L/lc)lc2${g_{b,{\cal M}}}\left( {{L \mathord{\left/ {\vphantom {L {{l_{\rm{c}}}}}} \right. \kern-\nulldelimiterspace} {{l_{\rm{c}}}}}} \right)l_{\rm{c}}^2$ we still expect the ratio in Eq. (39) to decrease when b increases, with a scaling close to b−2. Indeed, we find a good agreement, within one order of magnitude, between this scaling and the results obtained in Federrath & Klessen (2012, Federrath & Klessen 2013), even when we consider variations in the Mach number and virial parameters (Fig. 1).

These calculations demonstrate the origin of the large differences between values of the gravitational potential and the virial parameter that are calculated in a periodic simulation domain (T3)$\left( {{{\mathbb{T}}^3}} \right)$ and those that are derived from observations in space (ℝ3) from the size and mass of the cloud for pure (initial) turbulent conditions. This mismatch is a numerical artifact due to the numerical resolution of the Poisson equation in a torus geometry. It should be emphasized, however, that this does not imply that the numerical results of Federrath & Klessen (2012, Federrath & Klessen 2013) are erroneous, but it does call for a reexamination of interpretations that involves the estimation of αvir from the gravitational potential returned by the simulations.

thumbnail Fig. 1

Ratio of virial parameters from the simulations of Federrath & Klessen (2012, Federrath & Klessen 2013) for three different types of forcing. Each point corresponds to a Mach number, ℳ, and size L. Triangles (sol/comp) show the ratio between virial parameters for solenoidal (b 0.3) and compressive forcing (b 1) and circles (sol/mix) show the ratio between solenoidal and mixed forcing (b 0.4). The horizontal dashed lines give the expected value of th e ratios for a scaling in b−2.

4 Conclusion

In this article we have applied the statistical formalism developed in Paper I to the determination of the total gravitational energy and virial parameter of a cloud. We have demonstrated that the contribution of the (undetermined) internal structure of the clouds has only a small impact on these determinations in clouds that are not in a too advanced stage of star formation. This is the case, for example, for the clouds Polaris and Orion B studied in Paper I. In that case, the cloud gravitational energy and virial parameter can thus be safely estimated from the observed total mass and size, with no knowledge of their internal structures. We, note, however that for clouds within which star formation has already been ongoing for a significant amount of time, that is when gravity has started to break the clouds into small isolated condensed regions, contributions from the internal structure can become sizeable and affect the estimate of the potential energy, eG (see Sect. 2.1). This is an important result because the virial parameter determines the dynamics of a cloud, equilibrium, expansion or gravitational contraction.

Examining the same problem in a torus geometrical configuration, which is characteristic of numerical simulations in a periodic box, we have shown that, in contrast to real space, only the inner structure of the density fluctuations in the box contributes to the determination of the gravitational potential and the virial parameter (see Sect. 2.2). In that case, the (dominant) average contribution is lacking, a consequence of the ill-posed problem of solving the Poisson equation in T3${{\mathbb{T}}^3}$. We have demonstrated that, for pure (initial) turbulent conditions, the ratio of the viral parameter values in the box over the ones in real geometry is proportional to ((lc/L) × (bℳ))−2 (Eq. (39)). Thus, for a given large-scale Mach number and size of the simulation box, this decreases approximately as ~b−2, where b ∈ [1/3,1] denotes the (solenoidal vs compressive) turbulence forcing parameter. This explains the puzzling large discrepancy found in Federrath & Klessen (2012, Federrath & Klessen 2013) between the gravitational potential and virial parameter values inferred from the global characteristics of the simulation box and those inferred from the numerical results (see Sect. 3). We note that this does not affect the validity of the simulations of Federrath & Klessen (2012, Federrath & Klessen 2013). However, it does call for a reexamination of the interpretations that involve the estimation of αvir from the gravitation potential returned by the simulations.

Finally, we remark that these findings only hold for clouds that are statistically homogeneous (see Paper I) and large enough compared to the correlation length of ρ. They notably do not hold for (small-scale) collapsed subregions.

These calculations highlight again the power of the statistical formalism developed in Paper I to explore the general statistical properties of star-forming MCs from a limited number of observations or simulations. As explored in the present paper, its power is notably significant when estimating its global gravitational energy and virial parameter, and thus the level of binding, of such MCs. These results will be used in a forthcoming paper aimed at exploring the evolution of the PDF in star-forming clouds.

Acknowledgements

The authors are grateful to Christoph Federrath for always providing data from his numerical simulations upon request.

Appendix A Computation of the total potential energy on a control volume Ω

We derive the gravitational binding energy of a cloud covering a domain Ω in Sect. (2), and divided it into three contributions to isolate the effects of the internal structure (deviation from the average): eG(ρ)=Ω2ρ2|Ω|ΦGreen(xx)dx dx,$\left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle = \int_{{\Omega ^2}} {{{{{\left\langle \rho \right\rangle }^2}} \over {\left| \Omega \right|}}{\Phi _{{\rm{Green}}}}\left( {x - x'} \right){\rm{d}}x'} \,{\rm{d}}x,$(A.1) IC(ρ)=Ω2δρ(x)δρ(x)|Ω|ΦGreen(xx)dx dx,${I_{C\left( \rho \right)}} = \int_{{\Omega ^2}} {{{\delta \rho \left( x \right)\delta \rho \left( {x'} \right)} \over {\left| \Omega \right|}}{\Phi _{{\rm{Green}}}}\left( {x - x'} \right){\rm{d}}x'} \,{\rm{d}}x,$(A.2) 2ρIδ(ρ)=2ρΩ2δρ(x)|Ω|ΦGreen(xx) dx dx.$2\left\langle \rho \right\rangle \,{I_\delta }\left( \rho \right) = 2\left\langle \rho \right\rangle \int_{{\Omega ^2}} {{{\delta \rho \left( x \right)} \over {\left| \Omega \right|}}{\Phi _{{\rm{Green}}}}\left( {x - x'} \right)} \,{\rm{d}}x'\,{\rm{d}}x.$(A.3)

Then, using the change in variables (u, v) = φ(x, x2) = (x −x2, x + x2), we obtain eG(ρ)=ρφ1(Ω)duΦGreen(u)  φ2u(Ω)ρdυ8|Ω|,$\left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle = \left\langle \rho \right\rangle \,\int_{{\varphi _1}\left( \Omega \right)} {{\rm{d}}{\bf{u}}{\Phi _{{\rm{Green}}}}\left( {\bf{u}} \right)} \,\,\int_{\varphi _2^{\bf{u}}\left( \Omega \right)} {\left\langle \rho \right\rangle } {{{\rm{d}}{\bf{\upsilon }}} \over {8\left| \Omega \right|}},$(A.4) IC(ρ)=φ1(Ω)duΦGreen(u)  φ2u(Ω)dυ8|Ω|δρ(u+υ2)δρ(uυ2)=φ1(Ω)duΦGreen(u)C^ρ,L(u),$\matrix{ {{I_C}\left( \rho \right)} \hfill &amp; = \hfill &amp; {\,\int_{{\varphi _1}\left( \Omega \right)} {{\rm{d}}{\bf{u}}{\Phi _{{\rm{Green}}}}\left( {\bf{u}} \right)} \,\,\int_{\varphi _2^{\bf{u}}\left( \Omega \right)} {\,{{{\rm{d}}{\bf{\upsilon }}} \over {8\left| \Omega \right|}}} \,\delta \rho \left( {{{{\bf{u}} + {\bf{\upsilon }}} \over 2}} \right)\delta \rho \left( {{{{\bf{u}} - {\bf{\upsilon }}} \over 2}} \right)} \hfill \cr {} \hfill &amp; = \hfill &amp; {\int_{{\varphi _1}\left( \Omega \right)} {{\rm{d}}{\bf{u}}{\Phi _{{\rm{Green}}}}\left( {\bf{u}} \right)} \,{{\hat C}_{\rho ,L}}\left( {\bf{u}} \right),} \hfill \cr }$(A.5) Iδ(ρ)=φ2u(Ω)duΦGreen(u)φ2u(Ω)dυ8|Ω|δρ(u+υ2),$I\delta \left( \rho \right) = \int_{\varphi _2^u\left( \Omega \right)} {{\rm{d}}{\bf{u}}{\Phi _{{\rm{Green}}}}\left( {\bf{u}} \right)} \,\int_{\varphi _2^u\left( \Omega \right)} {{{{\rm{d}}{\bf{\upsilon }}} \over {8\left| \Omega \right|}}\delta \rho \left( {{{{\bf{u}} + {\bf{\upsilon }}} \over 2}} \right),}$(A.6)

due to the fact that Ω possess a center of symmetry and where φ1(Ω)=2Ω,${\varphi _1}\left( \Omega \right) = 2\Omega ,$(A.7) φ2u(Ω)=2((Ωu)Ω)+u,$\varphi _2^{\bf{u}}\left( \Omega \right) = 2\left( {\left( {\Omega - {\bf{u}}} \right)\, \cap \Omega } \right) + {\bf{u}},$(A.8)

and C^ρ,L${\hat C_{\rho ,L}}$ is the biased ergodic estimator of the ACF of ρ (see Sect. (2.1.2) and Appendix (A) of Paper I). For example, if Ω=[ L2,L2 ]3$\Omega = {\left[ { - {L \over 2},{L \over 2}} \right]^3}$ then φ2u(Ω)=[ L+| ui |,L| ui | ] ]$\varphi _2^{\bf{u}}\left( \Omega \right) = \left. {\left[ { - L + \left| {{u_i}} \right|,L - \left| {{u_i}} \right|} \right]} \right]$.

Furthermore, denoting I(u)=φ2u(Ω)dυ8|Ω|δρ(u+υ2),$I\left( {\bf{u}} \right) = \int_{\varphi _2^{\bf{u}}\left( \Omega \right)} {{{{\rm{d}}{\bf{\upsilon }}} \over {8\left| \Omega \right|}}\delta \rho \left( {{{{\bf{u}} + {\bf{\upsilon }}} \over 2}} \right)} ,$(A.9)

we see that I(u) is, modulo the factor 1/|Ω|, the average of the density deviations, δρ, in the sub-volume ((Ω − u) ∩ Ω) + u/2. This is easier to see in the pedagogical case where Ω=[ L2,L2 ]3$\Omega = {\left[ { - {L \over 2},{L \over 2}} \right]^3}$ as I(u)=L+|ui|L|ui|dυ8|Ω|δρ(u+υ2)=1|Ω|L+|ui|+ui2L|ui|+ui2δρ(x) dx.$\matrix{ {I\left( {\bf{u}} \right)} \hfill &amp; { = \int {\int {\int_{ - L + \left| {{u_i}} \right|}^{L - \left| {{u_i}} \right|} {{{{\rm{d}}{\bf{\upsilon }}} \over {8\left| \Omega \right|}}\delta \rho \left( {{{{\bf{u}} + {\bf{\upsilon }}} \over 2}} \right)} } } } \hfill \cr {} \hfill &amp; { = {1 \over {\left| \Omega \right|}}\int {\int {\int_{{{ - L + \left| {{u_i}} \right| + {u_i}} \over 2}}^{{{L - \left| {{u_i}} \right| + {u_i}} \over 2}} {\delta \rho \left( {\bf{x}} \right)} \,{\rm{d}}{\bf{x}}} .} } \hfill \cr }$(A.10)

Then, if the volume of ((Ω − u) ∩ Ω) is sufficiently large, for example | ((Ω − u) n Ω) | ≫ lc(ρ)3, I(u) ≃ | ((Ω − u) ∩ Ω) | 〈δρ〉 = 0 and I(u)φ2u(Ω)ρdυ8|Ω|=ρ|((Ωu)Ω)||Ω|.$I\left( {\bf{u}} \right) \ll \int_{\varphi _2^u\left( \Omega \right)} {\left\langle \rho \right\rangle } {{{\rm{d}}{\bf{\upsilon }}} \over {8\left| \Omega \right|}} = \left\langle \rho \right\rangle {{\left| {\left( {\left( {\Omega - {\bf{u}}} \right) \cap \Omega } \right)} \right|} \over {\left| \Omega \right|}}.$(A.11)

The integral I(u) thus only gives non-negligible contributions for u in a small volume of order lc(ρ)3 near the border ∂(2Ω), such as | ((Ω − u) ∩ Ω) | ≲ lc(p)3.

Therefore, providing that L = |Ω|1/3lc(ρ), we can neglect 2 〈 ρIδ(ρ) with respect to 〈eG(〈ρ 〉)〉. This leaves eGeG(ρ)+φ1(Ω)duΦGreen(u)C^ρ,L(u).$\left\langle {{e_G}} \right\rangle \simeq \left\langle {{e_G}\left( {\left\langle \rho \right\rangle } \right)} \right\rangle + \int_{{\varphi _1}\left( \Omega \right)} {{\rm{d}}u{\Phi _{{\rm{Green}}}}\left( u \right)} \,{\hat C_{\rho ,L}}\left( u \right).$(A.12)

References

  1. Federrath, C. 2016, MNRAS, 457 375 [Google Scholar]
  2. Federrath, C., & Klessen, R. S. 2012, ApJ, 761 156 [Google Scholar]
  3. Federrath, C., & Klessen, R. S. 2013, ApJ, 763 51 [Google Scholar]
  4. Federrath, C., Klessen, R. S., & Schmidt, W. 2008, ApJ, 688, L79 [NASA ADS] [CrossRef] [Google Scholar]
  5. Federrath, C., Roman-Duval, J., Klessen, R., Schmidt, W., & Mac Low, M.-M. 2010, A&A, 512, A81 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  6. Guillet, T., & Teyssier, R. 2011, J. Comput. Phys., 230 4756 [Google Scholar]
  7. Jaupart, E., & Chabrier, G. 2020, ApJ, 903, L2 [NASA ADS] [CrossRef] [Google Scholar]
  8. Jaupart, E., & Chabrier, G. 2021a, ApJ, 922, L36 [NASA ADS] [CrossRef] [Google Scholar]
  9. Jaupart, E., & Chabrier, G. 2021b, A&A, 663, A113 (Paper I) [Google Scholar]
  10. Kritsuk, A. G., Norman, M. L., Padoan, P., & Wagner, R. 2007, ApJ, 665 416 [NASA ADS] [CrossRef] [Google Scholar]
  11. McKee, C. F., & Ostriker, E. C. 2007, ARA&A, 45 565 [Google Scholar]
  12. McKee, C. F., & Zweibel, E. G. 1992, ApJ, 399 551 [CrossRef] [Google Scholar]
  13. Ricker, P. 2008, ApJS, 176 293 [NASA ADS] [CrossRef] [Google Scholar]

1

The usual definition of the binding energy involves a 1/2 multiplying factor in order to account for summations on interacting pairs, which we omit here for the sake of clarity.

2

We recall that we omitted the usual factor 1/2 in the definition of the binding energy.

All Figures

thumbnail Fig. 1

Ratio of virial parameters from the simulations of Federrath & Klessen (2012, Federrath & Klessen 2013) for three different types of forcing. Each point corresponds to a Mach number, ℳ, and size L. Triangles (sol/comp) show the ratio between virial parameters for solenoidal (b 0.3) and compressive forcing (b 1) and circles (sol/mix) show the ratio between solenoidal and mixed forcing (b 0.4). The horizontal dashed lines give the expected value of th e ratios for a scaling in b−2.

In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.