Notes on β0

The goal of β0 is to capture the expansion of a displaced fluid element due to the stratification of the atmosphere. MAESTROeX computes β0 as:

(360)β0(r,t)=ρ0exp(0r1Γ1p0p0rdr)

Constant Composition

Consider an isentropically stratified atmosphere, with a constant composition as a function of r. If you displace a parcel of fluid upwards, it will expand adabatically and continue to rise until its density matches the background density. Even if Γ1 is not constant in r, following from the definition of β0,

(361)1β0dβ0dr=1Γ1p0dp0dr

and the definition of Γ1,

(362)Γ1=dlogpdlogρ|s

So, at constant entropy, from the definition of Γ1, it must hold that

(363)1ρdρdr=1Γ1pdpdz.

Comparing to the definition of β0 then

(364)1β0β0dr=1Γ1p0dp0dr=1ρ0dρ0dr.

Therefore, β0=ρ0.

This means that if we have a constant composition and an isentropically stratified atmosphere, as we displace a fluid element, it will always remain neutrally buoyant.

Composition Gradient

If there is a change in composition with r, the situation is more complicated. Consider again an isentropically stratified atmosphere, now with a composition gradient. If you displace a parcel of fluid upwards, it will rise. If there are no processes that change the composition (e.g. reactions), then the composition in the fluid element will remain fixed. As it rises, it will the ambient medium will have a different composition that it has. In this case, what is the path to equilibrium?

On the Effect of Chemical Potential

In MAESTRO, we do things in an operator split fashion — the hydro is de-coupled from the burning. This means that during the hydro parts of the algorithm (where β0 is used), the system is fixed in chemical equilibrium. For completeness, however, here we describe the effects of the species’ chemical potentials, which were neglected in the original derivation of β0. Note that similar terms appear in the calculation of things such as specific heats, which may be important in the burning step — there appears to be very little about this in the literature, but everyone seems to assume it makes little difference.

Derivation of α

In paper I, α is defined as

(365)α((1ρhp)pTρcpρ2cppρ)

where

(366)hp(hp)T,X,cp(hT)p,X,pT(pT)ρ,X,pρ(pρ)T,X

where the subscript X means holding all Xi constant. In the absence of reactions, the X subscript can be dropped from all derivatives and with the use of the equation of state p=p(ρ,T), α can be written as α=α(ρ,T). Such a system without reactions and in thermal equilibrium could be either a pure system of one species, or a system of many species in chemical (and therefore thermodynamic) equilibrium. Cox and Giuli (hereafter CG) call the former type of system a “simple system” and therefore the latter a “non-simple system” in chemical equilibrium. The analysis in paper I that reduced Eq.365 to

(367)α=1Γ1p0

used CG’s discussion of the various adiabatic Γ’s. However, their discussion only pertains to “simple systems” or “non-simple systems” in chemical equilibrium. In general, nuclear reactions will be important and therefore this analysis needs to be reformed.

Even in the presence of reactions, Eq.365 can be rewritten as was done in paper I:

(368)α=1pχρcp[(1ρχρρeρpχρ)pχTTcp],

where

(369)χρ(lnplnρ)T,XχT(lnplnT)ρ,X.

Following the results of paper I, we want to find a relation between pχρ and Γ1.

For an equation of state p=p(ρ,T,X) we have

(370)dlnp=(lnplnρ)T,Xdlnρ+(lnplnT)ρ,XdlnT+i(lnplnXi)ρ,T,(Xj,ji)dlnXi.

We define another logarithmic derivative

(371)χXi(lnplnXi)ρ,T,(Xj,ji)

and therefore

(372)dlnp=χρ dlnρ+χT dlnT+iχXi dlnXi.

From here we get the general statement

(373)lnplnρ=χρ+χTlnTlnρ+iχXilnXilnρ

which must hold for an adiabatic process as well, and therefore we have

(374)Γ1=χρ+χT(Γ31)+iχXiΓ4,i

where we use CG’s definition of Γ1 and Γ3 and introduce a fourth gamma function:

(375)Γ1(lnplnρ)AD,Γ31(lnTlnρ)AD,Γ4,i(lnXilnρ)AD,

where the subscript AD means along an adiabat. We now derive an expression for Γ3.

The first law of thermodynamics can be written as

(376)dQ=dE+pdViμidNi

where μi=(ENi)AD,ρ,(Nj,ji) is the chemical potential; or per unit mass we have

(377)dq=depρ2dρiμid(niρ)=depρ2dρi(eXi)ρ,AD,(Xj,ji)dXi

where we have used Xiρi/ρ=Aini/ρNA and the chemical potential has been replaced with μi=AiNA(eXi)ρ,AD,(Xj,ji). Using this and expressing the specific internal energy as e=e(ρ,T,X) we then have

(378)dq=cvdT+[(eρ)T,Xpρ2]dρ+i[(eXi)ρ,T,(Xj,ji)(eXi)ρ,AD,(Xj,ji)]dXi

and

(379)(dlnTdlnρ)ADΓ31=1cvT[pρ(elnρ)T,X+i[(eXi)ρ,AD,(Xj,ji)(eXi)ρ,T,(Xj,ji)]XiΓ4,i]

Now we need to evaluate (e/lnρ)T,X. Again using the first law and the fact that ds=dq/T is an exact differential (i.e. mixed derivatives are equal) we have

(380)(ρ[cvT])T,X=(T[1T(eρ)T,XpTρ2])ρ,X1T(ρ(eT)ρ,X)T,X=1T2(eρ)T,X+1T(T(eρ)T,X)ρ,X+pT2ρ21Tρ2(pT)ρ,X(elnρ)T,X=pρ(1χT),

exactly the same result if we were to exclude species information. Similarly, we can find an expression for the derivative of energy with respect to composition

(381)(Xi[cvT])ρ,T,(Xj,ji)=(T[1T(eXi)ρ,T,(Xj,ji)1T(eXi)ρ,AD,(Xj,ji)])ρ,X1T(Xi(eT)ρ,X)ρ,T,(Xj,ji)=1T2[(eXi)ρ,AD,(Xj,ji)(eXi)ρ,T,(Xj,ji)]+     1T[(T(eXi)ρ,T,(Xj,ji))ρ,X(T(eXi)ρ,AD,(Xj,ji))ρ,X](eXi)ρ,T,(Xj,ji)=(eXi)ρ,AD,(Xj,ji)(lnT(eXi)ρ,AD,(Xj,ji))ρ,X.

Plugging these back into Eq.379 we have

(382)Γ31=1cvT[pρχT+i(lnT(eXi)ρ,AD,(Xj,ji))ρ,XXiΓ4,i],

or

(383)cv=1T(Γ31)[pρχT+i(lnT(eXi)ρ,AD,(Xj,ji))ρ,XXiΓ4,i].

We can obtain an expression for the specific heat at constant pressure from the enthalpy

(384)cp(hT)p,X=(eT)p,Xpρ2(ρT)p,X=(eT)p,X+pρ2(pT)ρ,X(ρp)T,X=(eT)p,X+pρTχtχρ.

The first term on the rhs can be obtained from writing e=e(p,T,X) and p=p(ρ,T,X):

(385)de=(ep)T,Xdp+(eT)p,XdT+i(eXi)p,T,(Xj,ji)dXidp=(pρ)T,Xdρ+(pT)ρ,XdT+i(pXi)ρ,T,(Xj,ji)dXi (eT)ρ,X=(ep)T,X(pT)ρ,X+(eT)p,X (eT)p,X=cv(eρ)T,X(ρp)T,X(pT)ρ,X=cvpχTρTχρ(1χT)

and

(386)cp=pρTχT2χρ+cv

Dividing this by Eq.383 and using the relation between the Γ ’s, Eq.374, we then have

(387)γcpcv=1+p(Γ31)ρχT2χρ[pρχT+i(lnT(eXi)ρ,AD,(Xj,ji))ρ,XXiΓ4,i]1=1+pχT(Γ1χρiχXiΓ4,i)pχρχT+ρχρi(lnT(eXi)ρ,AD,(Xj,ji))ρ,XXiΓ4,i=pχTΓ1+i[ρχρ(lnT(eXi)ρ,AD,(Xj,ji))ρ,XXipχTχXi]Γ4,ipχρχT+ρχρi(lnT(eXi)ρ,AD,(Xj,ji))ρ,XXiΓ4,ipχρ=1χTγ[pχTΓ1+i[ρχρ(1γ)(lnT(eXi)ρ,AD,(Xj,ji))ρ,XXipχTχXi]Γ4,i].

Plugging [eq:pchirho] into Eq.368 and rewriting the partial derivative of e with the help of [eq:dedlnrho] we have

(388)α=1pχρcp[(1ρχρρeρpχρ)pχTTcp]=γcpcpχT+(ρ(elnρ)T,Xp)χT2TρχρpχTΓ1+i[ρχρ(1γ)(lnT(eXi)ρ,AD,(Xj,ji))ρ,XXipχTχXi]Γ4,i=γΓ1pcp[cppχT2Tρχρ1+i[ρχρpχT(1γ)(lnT(eXi)ρ,AD,(Xj,ji))ρ,XXiχXi]Γ4,iΓ1]=(1Γ1p)[1+i[ρχρpχT(1γ)(lnT(eXi)ρ,AD,(Xj,ji))ρ,XXiχXi]Γ4,iΓ1]1
(389)α=1Γ1p[1+i[ρ2pρppT(1γ)NAAi(μiT)ρ,XXiχXi]Γ4,iΓ1]1

Recalling Derivation of β0

Recall from paper I that β0 was derived from the equation

(390)U+αUp0=S~

in such a fashion that we ended up with an equation of the form

(391)(β0(r)U)=β0S~.

The derivation in Appendix B of paper I for a β0 that satisfies [eq:beta constraint] automatically assumed α=(Γ10p0)1. This would have to be modified with the above derivation of α to be correct in a non-operator split fashion.