Wrong Explanations for the Luminosity of Main-Sequence Stars

  • Thread starter Thread starter Ken G
  • Start date Start date
  • Tags Tags
    Luminosity Stars
AI Thread Summary
Many introductory astronomy texts incorrectly explain why high-mass main-sequence stars are more luminous than low-mass stars, attributing it to stronger gravity and higher core temperatures. The correct explanation is that high-mass stars require less contraction to reach fusion temperatures, resulting in lower core densities and pressures, making them "leakier" and thus more luminous. High-mass stars' luminosity is primarily determined by the rate at which light escapes, rather than their fusion rates, which self-regulate to maintain equilibrium. This misunderstanding persists across various authoritative sources, leading to confusion among students seeking accurate information. The discussion highlights the need for clearer explanations of the mass-luminosity relationship and the underlying physics of stellar structure.
Astronomy news on Phys.org
  • #52
TEFLing said:
I thought that was what I was saying

I guess I'm confused

A diffusion equation with no net diffusion would imply a zero second derivative, yes?
In an idealized plane-parallel geometry, then zero second derivative means it is time steady diffusion, but there is still diffusion. The spherical symmetry changes things a bit.
 
  • #53
TEFLing said:
http://en.wikipedia.org/wiki/Kramers'_opacity_law

Doesn't free free scattering dominate electron scattering given the large coefficients?
Kramers type opacity can certainly enter, especially for cooler stars. For hotter stars, electron scattering is at least close to dominant, if not completely dominant. So no doubt, assuming constant opacity is a bit of a stretch, but it's simple. There is no general expression, because electron scattering has no T dependence if H is ionized, Kramers opacity has a steeply dropping dependence with T, and H-minus opacity, important in the envelopes of very cool stars, actually rises with T at low T because you need some free electrons to make it work. So the opacity is all over the map, and you'd be totally wrong using electron scattering for stars less massive than the Sun, but you can get order-unity kinds of results for most of the main sequence. It's definitely the weakest assumption, some advanced authors try to include some opacity effects but there are subjective choices in the tradeoffs between complexity and accuracy.
 
  • #54
I think I need to clarify vocabulary terms

I want one term for 'shell luminosity ' = 4pi r2 T4
Which is what you would see if you could magically excavate the stellar envelope down to that radius

And I want another word for Transported through luminosity ' which is the net diffusion of energy outwards...
And which is constant outside the core like you said for the energy conservation reason rationale you stated
 
  • #55
I want to try a calculation

From radiative equilibrium

d2L/ds2 = 0

Where L = 4 pi r2 T4 is the shell luminosity, and ds = rho k dr is the differential optical depth.

So

dL/ds = constant

L = Lcore - A s

At the surface

L* = Lcore - A rho k R

For ES, k ~ constant, rho R ~ M/R2, Lcore ~ rho2 Tn x Vcore ~ M2 / R3 Tn

Using the VT, T~ M/R, So

L* = (T/R) x ( B Tn+2 - A )

If the LHS ~ 4 pi R2 T4, and for n=4...

R2 T4 ~ (T/R) x ( B T6 - A )

R3 T3 ~ ( B T6 - A )

M3 + A ~ B T6

That is vaguely like T ~ sqrtM, which the VT converts to R ~ sqrtM ~ T

As we said above, and implies L ~ M^3
 
  • #56
For Kramers opacity

Lcore ~ M2 / R3 Tn

rho k r ~ rho2 T-3.5 ~ M2 / R5 T-3.5

L* ~ R2 T4 ~ M2 T2

Cancel out the M2, multiply by R5, and use VT

M2 R3 ~ M2 Tn-2 - T-3.5

Again ignoring the second term on the RHS, and for n=4

T2 ~ R3

So

M ~ R2.5
L ~ R8 ~ M3.2

Which is basically the above also
 
  • #57
Lcore ~ M2 / R3 Tn

L* ~ R2 T4 ~ M2 T2

R3 ~ Tn-2

n = 4
R3 ~ T2

n=18
R3 ~ T16
T ~ R3/16
M ~ RT ~ R19/16
L ~ R2 T4 ~ R2.75 ~ M2.3

Seems like all calculation paths lead to the vicinity of L ~ M3
 
  • #58
L ~ M2 / R3 Tn
L ~ R2 T4

Those are the scaling relations I was looking for
They capture the general gist of the balance between core fusion and surface luminosity
And they give informatively useful and accurate results

Precise details are much more complicated
 
  • #59
TEFLing said:
I think I need to clarify vocabulary terms

I want one term for 'shell luminosity ' = 4pi r2 T4
Which is what you would see if you could magically excavate the stellar envelope down to that radius

And I want another word for Transported through luminosity ' which is the net diffusion of energy outwards...
And which is constant outside the core like you said for the energy conservation reason rationale you stated
OK, then all is fine. Note that in standard lexicon, "transported through luminosity" is just "luminosity", and "shell luminosity" is just energy density times surface area, with the appropriate constant. But these are just the ways the words are normally used, when one is deep in the interior of the star it's pretty arbitrary what we might want to call the luminosity, and I agree with you that yours is a useful scenario for understanding how diffusion works.
 
Last edited:
  • #60
TEFLing said:
L ~ M2 / R3 Tn
L ~ R2 T4

Those are the scaling relations I was looking for
They capture the general gist of the balance between core fusion and surface luminosity
And they give informatively useful and accurate results

Precise details are much more complicated
There is indeed some latitude in what you do if you are not exactly solving the equations of stellar structure, giving the basic explanation for L a kind of subjective character. Personally, I'm not in favor of using the core and surface T as though they were the same thing, because they are set by very different physical processes. The core T is set by the need to self-regulate fusion to replace the heat that is diffusing out, and the surface T is set by the need for the blackbody emission at the surface to match that luminosity, so the former is fusion physics and the latter is Stefan-Boltzmann physics, and there is no reason the two should be the same or even proportional. So I wouldn't use L ~ R2 T4 for anything but a way to infer the surface T. Also, if you use diffusion physics to get L, you never need L ~ M2 / R3 Tn, you can just use L ~ R4 T4 / M from diffusion with constant opacity (or modify that for some other opacity like Kramers, but it won't be more accurate), and T ~ M/R from VT, and get L without even saying whether fusion is happening or not. However, if you want to get T in the core, and R, and then T at the surface, you would need to do what you are doing, since then you do need to know that fusion is happening.

The bottom line is, it is possible to get simple scaling relations to help understand, at a rough level of approximation, all the basic properties of a main-sequence star that is not completely convective and does not have strong radiative forces. These general scaling laws have been around for a long time, and there is really no excuse for any astronomy textbook to claim that high-mass main-sequence stars are more luminous because their stronger gravity compresses the core to higher pressure and temperature. None of the scaling laws have that character, it's really just completely wrong.
 
  • Like
Likes TEFLing
  • #61
Ken G said:
...

Also, if you use diffusion physics to get L, you never need L ~ M2 / R3 Tn, you can just use L ~ R4 T4 / M from diffusion with constant opacity (or modify that for some other opacity like Kramers, but it won't be more accurate), and T ~ M/R from VT, and get L without even saying whether fusion is happening or not...

M2 / R3 Tn ~ L ~ R4 T4 / M

M3 Tn ~ R7 T4 ~ R3 M4

Tn ~ R3 M ~ R4 T

Tn-1 ~ R4 ~ M/R ~ M/T(n-1)/4

T5(n-1)/4 ~ M ~ R5

L ~ M2/5 M16/5(n-1)

n=4
-----
L ~ M1.5

Seems like the two equations for L are not compatible...(??)
 
  • #62
I want to ask about the VT...

Isn't the VT based on the IGL and an assumption of HSE?

For, the VT is consistent with the one zone scaling relation form of the HSE equation

dP/dr = -rho g

P/R ~ M2 / R5

rho T ~ rho g R

T ~ M/RBUT what if pressure support derives from RADIATION pressure, not gas pressure??

Wouldn't wht

dP/dr = -rho g

P/R ~ M2 / R5

T4 ~ M2 / R4

T ~ sqrt(M) / R
 
  • #63
TEFLing said:
Tn-1 ~ R4 ~ M/R ~ M/T(n-1)/4
Wasn't there a problem in the middle there?
 
  • Like
Likes TEFLing
  • #64
TEFLing said:
I want to ask about the VT...

Isn't the VT based on the IGL and an assumption of HSE?
Yes exactly.
BUT what if pressure support derives from RADIATION pressure, not gas pressure??
Then you need a different form for the VT, but it's only an issue for the highest mass stars.
T4 ~ M2 / R4

T ~ sqrt(M) / R
Radiation P is not T4, it is ~ T4/R3. You should ultimately find the Eddington limit: L ~ M, if you take constant opacity.
 
  • #65
If we have scaling relations...

L ~ R2 T4
L ~ M2 Tn / R3
T ~ M/R or M1/2 / R

Gas support
==========
R2 T4 ~ (RT)2 Tn / R3 ~ Tn+2 / R

R3 ~ Tn-2

M3 ~ Tn+1

L ~ R2 T4 ~ T2(n-2)/3 M12/(n+1) ~ M2(n-2)/(n+1) M12/(n+1) ~ M(2n+8)/(n+1)

L ~ M3.2 (n=4)
Radiation support
===============
R2 T4 ~ (RT)4 Tn / R3

R ~ Tn

M1/2 ~ Tn+1

L ~ R2 T4 ~ T2n M2/(n+1) ~ Mn/(n+1) M2/(n+1) ~ M(n+2)/(n+1)

L ~ M1.1 (n=18)Isn't this qualitatively correct? Doesn't L ~ M for high mass stars?
 
  • #66
Take x2...
========

M2 / R3 Tn ~ L ~ R4 T4 / M

M3 Tn ~ R7 T4 ~ R3 M4

Tn ~ R3 M ~ R4 T

Tn-1 ~ R4

T ~ R4/(n-1) ~ M/R

M ~ R(n+3)/(n-1)

R ~ M(n-1)/(n+3)

T ~ M4/(n+3) L ~ R2 T4 ~ M(2n+14)/(n+3)

n=4
L ~ M22/7
 
  • #67
TEFLing said:
Radiation support
===============
R2 T4 ~ (RT)4 Tn / R3

R ~ Tn

M1/2 ~ Tn+1

L ~ R2 T4 ~ T2n M2/(n+1) ~ Mn/(n+1) M2/(n+1) ~ M(n+2)/(n+1)

L ~ M1.1 (n=18)Isn't this qualitatively correct? Doesn't L ~ M for high mass stars?
It is qualitatively correct, but it's kind of a coincidence. What actually happens is L ~ M exactly, if we use constant opacity, and the value of n makes no difference at all. The flaw in this analysis is that it assumes the core T and the surface T are either the same, or at least proportional, and there's no physical basis for that assumption. All the same, since it works out to be approximately true that they are proportional, you can get away with asserting it in the derivation. But doing so puts the cart before the horse-- we know L ~ M for other reasons (the Eddington limit), so your scaling analysis should actually be used to derive the ratio of surface T to interior T. When you find that the two are approximately proportional, you have demonstrated something important, but it's not the same thing being demonstrated as if you just assume that from the start.
 
  • #68
TEFLing said:
Take x2...
========

M2 / R3 Tn ~ L ~ R4 T4 / M

M3 Tn ~ R7 T4 ~ R3 M4

Tn ~ R3 M ~ R4 T

Tn-1 ~ R4

T ~ R4/(n-1) ~ M/R

M ~ R(n+3)/(n-1)

R ~ M(n-1)/(n+3)

T ~ M4/(n+3)L ~ R2 T4 ~ M(2n+14)/(n+3)

n=4
L ~ M22/7
Yes, that's bang on. So we see we have some subjective freedom as to how we will derive L. We can take an n for nuclear burning, even though that will limit our range of applicability (p-p fusion doesn't dominate past a solar mass), and give exactly the analysis you have. You get a slightly complicated power in the mass-luminosity relation, but it is quite useful all the same, and works fine. Alternatively, we can ignore any need to assert an n for fusion, and just say that fusion induces a thermostatic effect that regulates T to be something like 20 million K. Unfortunately this will not be at all exact, but we do get a very bare bones kind of derivation that gives L ~ M3. A simpler power, simpler physics, but not more justifiable-- just a different set of things to idealize. So it's a subjective choice at this point.
 
  • Like
Likes TEFLing
  • #69
http://www.physics.rutgers.edu/~pryor/ph442/lecture12.pdf

I found the above link for the index of refraction of a plasma...

In slide 1 I think they mean to say n is IMAGINARY at low frequencies, not negative

I think at the high temperatures inside stars, n~1 is a very accurate approximation
 
Last edited by a moderator:
  • #70
Ken G said:
Yes, that's bang on. So we see we have some subjective freedom as to how we will derive L. We can take an n for nuclear burning, even though that will limit our range of applicability (p-p fusion doesn't dominate past a solar mass), and give exactly the analysis you have. You get a slightly complicated power in the mass-luminosity relation, but it is quite useful all the same, and works fine. Alternatively, we can ignore any need to assert an n for fusion, and just say that fusion induces a thermostatic effect that regulates T to be something like 20 million K. Unfortunately this will not be at all exact, but we do get a very bare bones kind of derivation that gives L ~ M3. A simpler power, simpler physics, but not more justifiable-- just a different set of things to idealize. So it's a subjective choice at this point.
WOW

I think I almost understand what you are saying

Roughly speaking, we start with a diffusion equation for the starlight, which is time invariant, d/dt -> 0

=> constant Laplacian double derivative on the other side...

Implying a constant heat flow

L ~ c l d/dr(4 pi r2 T4)

Where l~1/(rho sigma) is a mean free path measure

So the scaling relation simplification is

L ~ L / ( rho sigma T )

Rho sigma R ~ constant

Which says that if a star becomes more opaque per particle, the column density of particles must decrease to let the same light luminosity out through the plasma to the surface

Assuming ES is dominant, sigma ~ constant, and rho R must be constant too

==============
Using FF opacity

Sigma ~ rho T^-3.5

Rho sigma R ~ rho^2 T^-3.5 R ~ 1

M^2/R^5 ~ T^3.5

-----------------------
Using VT , RT ~ M

M^2 ~ R^1.5 M^3.5

1 ~ (RM)^1.5

M ~ 1/R

(??)-----------------------
Using VT , RT ~ sqrt(M)

M^2/R^5 ~ T^3.5

M^2 ~ R^1.5 M^1.75

M^.25 ~ R^1.5

M ~ R^6

M^2/R^5 ~ T^3.5

M^7/6 ~ T^7/2

M ~ T^3

L ~ R2T4 ~ M^.333 M^1.333 ~ M^5/3
 
  • #71
TEFLing said:
Implying a constant heat flow

L ~ c l d/dr(4 pi r2 T4)

Where l~1/(rho sigma) is a mean free path measure
Yes, this is correct.
So the scaling relation simplification is

L ~ L / ( rho sigma T )

Rho sigma R ~ constant
I wasn't sure what you were getting at here. We both get L ~ d/dtau(r2 T4), where dtau = sigma*rho*dr is the optical depth across dr, and r2 T4 is like the radiative energy per shell of unit radial length. This is like setting dtau=1 and using the shells you were talking about above. Where I would go with that is just insert the characteristic proportionality tau ~ M/R2 for constant sigma (for simplicity, or as a benchmark for comparison), yielding L ~ (RT)4/M. So that's all we can get from radiative diffusion, and if we take M as given, then the L depends only on RT if sigma is constant. Then the VT shows that RT is also constrained by M, since RT ~ M. Thus L depends only on M. You can use different opacity laws, which will introduce some independent dependence on R and T, which will in turn bring in the fusion physics (and the power n) because T will now matter, but if constant-sigma dominates, that won't occur, and if n is large (as for CNO cycle), then the T physics is highly thermostatic and it also won't matter much how the fusion changes as M changes, i.e., the explicit value of n won't matter if n is large. But you are right that for M ~ 0.5 solar, we have p-p fusion with n = 4, and it only goes up to about n = 5 for solar M, so those values of n are small enough that some fusion physics will end up showing up in L, as will some Kramers opacity effects. You only get the super-simple result L ~ M3 if you ignore those details, but then we are also ignoring convection and radiative forces, so we are already going to be stuck with some pretty rough results.

You are also looking at the scaling laws for R(M), so they will start with R ~ M/T and have to use some fusion physics to get T, but for higher mass stars, again T will be fairly thermostatic, around 20 MK to 30 MK, so you won't do too badly with R ~ M. For lower M stars, the lower n of p-p fusion makes R(M) drop less rapidly as M drops, and convection also becomes more of an issue, so again it's a matter of what subjective tradeoffs you want to make in your idealizations.

I wouldn't try to get surface T until the end-- once I have a handle on both L and R, I would then say the surface T is ~ L1/4/R1/2. The surface T rarely has much input into either L or R, so that last formula is not very useful as an input to either the L or the R determination. However, you have seen some nice scaling relations about the ratio of surface to core T.
 
  • Like
Likes TEFLing
  • #72
Ken G said:
Yes, this is correct.
I wasn't sure what you were getting at here. We both get L ~ d/dtau(r2 T4), where dtau = sigma*rho*dr is the optical depth across dr, and r2 T4 is like the radiative energy per shell of unit radial length. This is like setting dtau=1 and using the shells you were talking about above. Where I would go with that is just insert the characteristic proportionality tau ~ M/R2 for constant sigma (for simplicity, or as a benchmark for comparison), yielding L ~ (RT)4/M. So that's all we can get from radiative diffusion, and if we take M as given, then the L depends only on RT if sigma is constant. Then the VT shows that RT is also constrained by M, since RT ~ M. Thus L depends only on M. You can use different opacity laws, which will introduce some independent dependence on R and T, which will in turn bring in the fusion physics (and the power n) because T will now matter, but if constant-sigma dominates, that won't occur, and if n is large (as for CNO cycle), then the T physics is highly thermostatic and it also won't matter much how the fusion changes as M changes, i.e., the explicit value of n won't matter if n is large. But you are right that for M ~ 0.5 solar, we have p-p fusion with n = 4, and it only goes up to about n = 5 for solar M, so those values of n are small enough that some fusion physics will end up showing up in L, as will some Kramers opacity effects. You only get the super-simple result L ~ M3 if you ignore those details, but then we are also ignoring convection and radiative forces, so we are already going to be stuck with some pretty rough results.

You are also looking at the scaling laws for R(M), so they will start with R ~ M/T and have to use some fusion physics to get T, but for higher mass stars, again T will be fairly thermostatic, around 20 MK to 30 MK, so you won't do too badly with R ~ M. For lower M stars, the lower n of p-p fusion makes R(M) drop less rapidly as M drops, and convection also becomes more of an issue, so again it's a matter of what subjective tradeoffs you want to make in your idealizations.

I wouldn't try to get surface T until the end-- once I have a handle on both L and R, I would then say the surface T is ~ L1/4/R1/2. The surface T rarely has much input into either L or R, so that last formula is not very useful as an input to either the L or the R determination. However, you have seen some nice scaling relations about the ratio of surface to core T.

Conceptually I think you have an equation like

L ~ d/dtau ( r2 T4) ~ ( Lcore - Lsurfsce)/( M/R2)

I think you can then use scaling relations like

Lcore ~ M rho T^n

And ignore Lsurface on the RHS since it's smaller than Lcore

Seemingly you're saying that the L seen at the surface is the original core luminosity stepped down like a voltage transformer , by the number of MFP i.e. optical depth

Assuming constant opacity, and using n=4,18 and RT ~ M, sqrt(M)...

You seem to get

L ~ M3 for low mass stars
L ~ M2 for intermediate mass stars
L ~ M for high mass stars
 
  • #73
I get L ~ M3 for all stars, independent of n, if I assume gas pressure dominates over radiation pressure, ideal gas T, ignore convection, and assume fixed opacity. I don't even need to say if fusion is even happening to get this, indeed look at the tracks of pre-main-sequence stars in an H-R diagram to see just how unnecessary it is to say if fusion is happening to get L.

If we instead use radiation pressure, instead of gas pressure, we get the Eddington limit value for L, which obeys L ~ M for constant opacity and no convection, and doesn't need to know anything about T or n. You'd actually get a lot of convection, but it might only carry some of the L, so the L ~ M is probably still not too bad.
 
Last edited:

Similar threads

Back
Top