Homology Relations for Stars

A star in hydrostatic and thermal equilibrium is governed by four coupled differential equations plus three constitutive relations — an equation of state, an opacity law, and a nuclear energy-generation rate. Integrating them exactly means a numerical model, yet most of what the main sequence is can be read off without solving anything, by exploiting a similarity symmetry. Two stars are homologous if their interiors have the same functional shape and differ only by overall scale factors in mass and radius: plot $\rho(r)/\rho_c$ against $r/R$ and both fall on one curve. The homology transformation relating the family turns each differential structure equation into an algebraic relation among scaling exponents. Feed in power-law microphysics — an equation of state $P\propto\rho^{a}T^{b}$ , an opacity $\kappa\propto\rho^{n}T^{-s}$ , and a burning rate $\varepsilon\propto\rho\,T^{\eta}$ — and the algebra closes. Out drop the mass–luminosity relation $L\propto M^{\nu}$ , the mass–radius relation $R\propto M^{\xi}$ , the central temperature $T_c\propto \mu M/R$ , and the nuclear lifetime $\tau\propto M/L$ . A single page of scaling predicts that massive stars are hotter inside, far more luminous, larger, and dramatically shorter-lived — exactly the diagonal locus the Hertzsprung–Russell diagram displays, and the slope $L\propto M^{3.5}$ measured from binary-star masses. The exponents are not constants of nature but a bookkeeping of which physics dominates. Swap Kramers opacity for electron scattering and $\nu$ falls from about $5$ to $3$ ; swap the pp chain for the CNO cycle and the mass–radius relation stiffens. Where the physics changes character — degeneracy at low mass, radiation pressure at high mass — the similarity assumption breaks and the exponents jump. Homology is best read as a family of local scaling laws, each valid inside one regime, its failures mapping the regime boundaries as sharply as its successes describe them.

Key takeaways

Homologous stars share one dimensionless interior profile, so their central values scale as pure powers of $M$ and $R$ : $\rho_c\propto M/R^3$ , $P_c\propto M^2/R^4$ , $T_c\propto\mu M/R$ .
Substituting these plus power-law microphysics into the four structure equations turns them into algebra, yielding $L\propto M^{\nu}$ (mass–luminosity) and $R\propto M^{\xi}$ (mass–radius).
For an ideal gas with constant opacity the luminosity is the Eddington result $L\propto\mu^4 M^3/\kappa$ — remarkably independent of the nuclear burning rate, which fixes $R$ instead.
The exponent tracks the microphysics: electron scattering gives $\nu\approx3$ , Kramers opacity steepens it to $\nu\approx5$ , and the observed main-sequence average is $\nu\approx3.5$ .
Homology fails where the constitutive laws change character — degeneracy at low mass, radiation pressure at high mass — so its exponents shift across the main sequence rather than holding as one law.
Key idea: If stellar interiors are self-similar, the differential structure equations collapse to algebraic relations among scaling exponents, so power-law microphysics immediately delivers $L\propto M^{\nu}$ and $R\propto M^{\xi}$ .

The core insight

If stellar interiors are self-similar, the differential structure equations reduce to algebraic relations among scaling exponents, so power-law microphysics immediately yields $L\propto M^{\nu}$ and $R\propto M^{\xi}$ — the mass–luminosity and mass–radius relations of the main sequence, obtained without integrating a single equation.

Where this lives today

Homology is no museum piece — it is the working intuition behind every modern stellar-structure calculation and a first-line sanity check on the codes that replaced hand computation. When a 1D evolution code such as MESA outputs a zero-age main sequence, the first thing a modeller confirms is that $L\propto M^{3.5}$ emerges with the right bend near a solar mass; a run that violates it signals a bug or an unphysical opacity table. The mass–luminosity relation itself is now measured, not inferred: detached eclipsing binaries in catalogues like DEBCat deliver masses and radii to better than a percent, and Gaia's parallaxes have turned the Hertzsprung–Russell diagram of a billion stars into a precision test of the homology diagonal. Asteroseismology from Kepler and TESS probes the interior directly through oscillation frequencies, confirming the central-condition scalings and exposing where real stars depart from self-similarity — in their convective cores and chemically graded interiors. Cluster ages from main-sequence turnoffs and the characterization of exoplanet host stars lean on the same scalings. And at the upper extreme, the approach of luminosity to the linear Eddington limit — where homology's ideal-gas assumption finally fails — governs the mass loss and fate of the most massive stars.

The physical picture

Imagine photographing two stars and shrinking the larger until its radius matches the smaller. Homology asserts that if you also rescale mass, density, temperature and pressure by the appropriate powers, the two pictures become one — the dimensionless profile $\rho(r)/\rho_c$ versus $r/R$ is a single universal curve for the whole family. Everything about the family is then encoded in how the central values scale with $M$ and $R$ . Hydrostatic equilibrium does the heaviest lifting: gravity crushes a bigger mass into a given radius harder, so the central pressure climbs as $P_c\propto M^2/R^4$ . For an ideal gas that pressure demands a hotter core, forcing $T_c\propto \mu M/R$ . A hotter core leaks radiation faster and runs its reactions faster, so luminosity rises steeply with mass. That single chain — gravity to pressure to temperature to luminosity — is the entire content of the mass–luminosity relation, and the first figure fixes the geometry it rests on.

The exponent in $L\propto M^{\nu}$ is not fixed once and for all — it is a linear combination of the microphysics exponents, changing wherever the dominant physics changes. Low-mass envelopes are ruled by Kramers opacity, $\kappa\propto\rho\,T^{-3.5}$ , whose fierce temperature sensitivity steepens the luminosity toward $M^{5}$ ; massive stars are dominated by nearly constant electron scattering, which flattens it to $M^{3}$ . Homology makes these dependences transparent because each microphysical exponent enters as a term in an algebraic system for the slope, as the second figure shows with real numbers.

Figure

The punchline: homology predicts a power-law mass–luminosity relation — electron scattering gives L ∝ M³ for massive stars, Kramers opacity steepens it to L ∝ M⁵ at low mass, and the mean slope L ∝ M^3.5 anchors on the Sun, reaches ≈3200 L⊙ at 10 M⊙, and drops to ≈0.09 L⊙ at 0.5 M⊙

The core equations

Homology begins with a single ansatz: the mass enclosed at fractional radius $r/R$ is a universal function times the total mass, so every quantity at fixed mass fraction scales as a power of $M$ and $R$ . The structure equations do the rest.

m(r) \;=\; f\!\left(\frac{r}{R}\right) M \qquad (1)

where: $m(r)$ — mass enclosed within radius $r$ ; $M$ — total mass; $R$ — total radius; $f$ — universal dimensionless mass profile, identical for every member of the family. That $f$ is the same for all masses is what makes the following relations exact power laws. Hydrostatic equilibrium, converted into a scaling by replacing derivatives with ratios of characteristic scales, gives the master relation: the central pressure needed to hold a star up.

\frac{\mathrm{d}P}{\mathrm{d}r} = -\frac{G\,m(r)\,\rho}{r^2}\;\;\Rightarrow\;\; P_c \propto \frac{M^2}{R^4} \qquad (2)

where: $P$ — pressure; $\rho$ — mass density; $G$ — gravitational constant; $P_c$ — central pressure. Doubling the mass at fixed radius quadruples the central pressure; compressing the star raises it as the fourth power. The density scaling is pure geometry — mass over volume:

\rho \;\propto\; \frac{M}{R^3} \qquad (3)

where: $\rho$ — characteristic density (mean and central scale identically, since the profile shape is fixed). The equation of state converts the mechanical pressure scaling into a thermal one; for an ideal gas $a=b=1$ and the central temperature follows.

P \;\propto\; \rho^{a}\,T^{b} \quad\Rightarrow\quad T_c \propto \frac{\mu M}{R}\;\;(\text{ideal gas},\,a=b=1) \qquad (4)

where: $T$ — temperature; $T_c$ — central temperature; $a,b$ — EOS exponents; $\mu$ — mean molecular weight, with the ideal-gas law $P=\rho k_B T/(\mu m_H)$ giving $a=b=1$ . More massive or compact stars need hotter cores, and a heavier mean particle (larger $\mu$ ) needs a hotter gas for the same pressure. Opacity and energy generation enter as power laws too. The Rosseland mean opacity is a product of powers of density and temperature.

\kappa \;\propto\; \rho^{n}\,T^{-s} \qquad (5)

where: $\kappa$ — Rosseland mean opacity; $n,s$ — opacity exponents. Two limits matter: Kramers bound-free and free-free absorption gives $n=1,\,s=3.5$ ; electron scattering gives $n=s=0$ (constant). Radiative diffusion is the transport equation carrying the luminosity out — where opacity enters the luminosity.

\frac{\mathrm{d}T}{\mathrm{d}r} = -\frac{3}{4ac}\,\frac{\kappa\rho}{T^3}\,\frac{L(r)}{4\pi r^2} \qquad (6)

where: $L(r)$ — luminosity crossing radius $r$ ; $a$ — radiation constant; $c$ — speed of light. A gradient carries only so much flux; raising opacity or lowering temperature throttles the luminosity the envelope can transmit. The nuclear energy-generation rate closes the thermal balance and is extraordinarily temperature-sensitive:

\varepsilon \;\propto\; \rho\,T^{\eta} \quad (\eta\approx 4\ \text{pp},\; \eta\approx 16\ \text{CNO}) \qquad (7)

where: $\varepsilon$ — energy generation rate per unit mass; $\eta$ — temperature-sensitivity exponent. The pp chain has a mild $\eta\approx4$ ; the CNO cycle, dominant above $\sim1.3\,M_\odot$ , a fierce $\eta\approx16$ . Combining diffusion, the ideal gas and homology yields the headline result — the mass–luminosity relation, its exponent depending on the opacity regime.

L \;\propto\; M^{\nu}, \qquad \nu \approx 3\text{–}5 \qquad (8)

where: $\nu$ — homology exponent set by opacity and EOS; the observed average is $\nu\approx3.5$ . For constant opacity it collapses to the classic Eddington form.

L \;\propto\; \frac{\mu^4\,M^3}{\kappa_0}\quad(\text{constant opacity, ideal gas}) \qquad (9)

where: $\kappa_0$ — opacity normalisation; $\mu$ — mean molecular weight. Strikingly, $L$ carries no factor of the nuclear rate: burning sets $R$ and $T_c$ , not the luminosity the envelope can leak. Dividing the fuel supply by the burn rate gives the nuclear lifetime, falling steeply with mass:

\tau_{\mathrm{nuc}} \;\propto\; \frac{M}{L} \;\propto\; M^{1-\nu} \qquad (10)

where: $\tau_{\mathrm{nuc}}$ — main-sequence lifetime. With $\nu\approx3.5$ , $\tau\propto M^{-2.5}$ : a star ten times the Sun's mass lives a few hundredths of a percent as long.

Derivation ACentral pressure and temperature from mechanics alone

The goal of this first derivation is to show that the two most important scalings in the subject — the central pressure and the central temperature — follow from mechanics and thermodynamics alone, with no reference to nuclear burning, opacity, or energy transport. We convert the hydrostatic equation into an algebraic scaling, obtain $P_c\propto M^2/R^4$ , and fold in the ideal-gas law to reach $T_c\propto \mu M/R$ . The ordering matters: a star's internal temperature is set by its self-gravity, and nuclear reactions must arrange themselves to match it rather than dictate it.

Step 1. Take hydrostatic equilibrium, $\mathrm{d}P/\mathrm{d}r=-Gm\rho/r^2$ , and replace each quantity by its characteristic scale: $\mathrm{d}P/\mathrm{d}r\to P_c/R$ , $m\to M$ , $\rho\to\bar\rho\propto M/R^3$ , $r\to R$ . Why: under homology every profile is the same shape, so a derivative across the star is the central value over the radius, up to an order-unity constant common to all masses.

Step 2. Substituting gives $P_c/R \sim G\,M\,(M/R^3)/R^2$ , and multiplying by $R$ yields $P_c \propto GM^2/R^4$ . Why: the central pressure is fixed entirely by the star's mass and radius, independent of what the gas is or how it burns.

Step 3. Impose the ideal-gas law $P_c\propto\rho_c T_c\propto (M/R^3)\,T_c$ and equate it to the mechanical result: $(M/R^3)\,T_c\propto M^2/R^4$ . Why: pressure has two expressions — what mechanics needs and what the gas can supply — and equilibrium satisfies both.

Step 4. Solving for the temperature, factors of $R$ partly cancel to leave $T_c\propto M/R$ , and restoring composition through $P=\rho k_B T/(\mu m_H)$ carries a factor of $\mu$ : $T_c\propto \mu M/R$ . Why: this is why massive stars have hotter cores and ignite the CNO cycle, and why rising $\mu$ slowly heats the core over a star's life. *Verified: $P_c\propto M^2/R^4$ and $T_c\propto\mu M/R$ reproduce the standard homology scalings of Carroll and Ostlie ch. 10 and Kippenhahn, Weigert and Weiss ch. 20; both are independent of $\varepsilon$ and $\kappa$ , exactly as the mechanical origin requires.*

Go deeper · Derivation B · 1 minThe mass–luminosity relation for constant opacity

Derivation BThe mass–luminosity relation for constant opacity

The goal here is to derive the Eddington mass–luminosity relation $L\propto\mu^4 M^3/\kappa$ and, crucially, to expose why luminosity is independent of the nuclear energy-generation rate — a fact that surprises everyone the first time. We scale radiative diffusion, insert the homology values for $T_c$ and $\rho_c$ , and watch every factor of $R$ cancel for constant opacity. Generalising to a power-law opacity then shows how $\nu$ shifts between the electron-scattering and Kramers regimes — the origin of the bend in the main sequence near one solar mass.

Step 1. Scale the diffusion equation (6): $T_c/R \sim (\kappa\,\rho_c/T_c^3)\,(L/R^2)$ , then solve for the luminosity to get $L \propto T_c^4\,R/(\kappa\,\rho_c)$ . Why: diffusion caps the flux a temperature gradient can carry, so the luminosity is whatever that cap allows — set by $T_c$ , $R$ , $\rho_c$ and opacity, with the burning rate nowhere in sight.

Step 2. Insert the homology scalings $T_c\propto \mu M/R$ and $\rho_c\propto M/R^3$ for constant opacity $\kappa=\kappa_0$ : $L \propto (\mu M/R)^4\,R/[\kappa_0\,(M/R^3)] = \mu^4 M^3/\kappa_0$ . Why: every power of $R$ cancels identically — luminosity does not care about the radius, being a pure function of mass and composition.

Step 3. So $L\propto\mu^4 M^3/\kappa_0$ is the classic Eddington relation, slope $\nu=3$ , for hot massive stars where electron scattering dominates and $\kappa$ is constant. Why: the $\mu^4$ factor makes an evolved, helium-enriched star far brighter, and the absence of $\varepsilon$ means the nuclear rate adjusts $R$ and $T_c$ so production matches this fixed leak.

Step 4. For a general opacity $\kappa\propto\rho^n T^{-s}$ , the same substitution gives $L\propto M^{3+s-n}\,R^{3n-s}$ ; Kramers ( $n=1,\,s=3.5$ ) gives $L\propto M^{5.5}R^{-0.5}$ , which with $R\propto M^{0.75}$ becomes $L\propto M^{5.1}$ . Why: this is why the main sequence is steeper for low-mass Kramers stars ( $\nu\approx5$ ) than for high-mass scattering stars ( $\nu\approx3$ ), the mean near $3.5$ . *Verified: constant opacity gives $L\propto\mu^4 M^3/\kappa$ (Eddington's standard-model result, Carroll and Ostlie ch. 10; Hansen, Kawaler and Trimble ch. 1); the Kramers exponent $3+s-n=5.5$ closing to $\approx5.1$ matches the steeper low-mass slope, and $L$ is independent of $\varepsilon$ as claimed.*

Go deeper · Derivation C · 1 minThe mass–radius relation from nuclear balance

Derivation CThe mass–radius relation from nuclear balance

The goal of the third derivation is to close the system by demanding that the luminosity leaking out by diffusion equals the luminosity generated by nuclear burning, which finally pins down the radius as a function of mass. Until now the radius floated free — the diffusion argument fixed $L(M)$ with $R$ cancelling — but a real star must produce exactly the energy it radiates, and that thermal-balance condition selects a unique $R(M)$ . Setting the two luminosities equal and solving for $\xi$ , we find the extreme temperature sensitivity of burning makes $R$ a remarkably weak function of $M$ — the reason the main sequence is a thin band.

Step 1. Integrate the generation rate over the mass for the total nuclear luminosity: $L \propto M\,\varepsilon \propto M\,\rho_c\,T_c^{\eta} \propto M\,(M/R^3)\,(M/R)^{\eta} = M^{2+\eta}/R^{3+\eta}$ . Why: total power is mass times the specific rate, and the rate inherits the density and temperature scalings, so the radius enters with the large power $3+\eta$ .

Step 2. Set this nuclear luminosity equal to the diffusive luminosity $L\propto M^{\nu}$ the envelope can carry: $M^{\nu} \propto M^{2+\eta}/R^{3+\eta}$ . Why: thermal equilibrium demands the two be equal — a star producing more than it could radiate would expand and cool until they matched, which is the adjustment of $R$ we are solving for.

Step 3. Solve for the radius: $R^{3+\eta} \propto M^{2+\eta-\nu}$ , hence $R \propto M^{\xi}$ with $\xi=(2+\eta-\nu)/(3+\eta)$ . Why: the enormous denominator $3+\eta$ divides down the mass exponent so that $R$ barely responds to $M$ — the structural reason main-sequence radii span a narrow range.

Step 4. Evaluate for massive CNO burners, $\eta\approx16$ and $\nu\approx3$ : $\xi=(2+16-3)/(3+16)=15/19\approx0.79$ , so $R\propto M^{0.8}$ . Why: this shallow slope, close to the observed value, is why a ten-solar-mass star is only about four times the Sun's radius rather than ten. *Verified: $\xi=(2+\eta-\nu)/(3+\eta)$ gives $15/19\approx0.79$ for CNO and $\xi\approx0.6$ for pp burning, bracketing the observed $0.6$ – $0.8$ mass–radius slope of upper-main-sequence stars (Kippenhahn, Weigert and Weiss ch. 20). The garbled denominator in some sources is corrected here to $3+\eta$ .*

Limiting cases

The predicted slope is a bookkeeping of which opacity and energy-generation physics dominates, so it shifts across the main sequence and breaks down at both ends.

Case	Result	Meaning
Low mass, Kramers opacity, pp chain	$L\propto M^{\sim5},\; R\propto M^{\sim0.6}$	The strong dependence $\kappa\propto\rho\,T^{-3.5}$ steepens the luminosity slope; mild pp burning gives a shallow radius law.
Intermediate / observed main sequence	$L\propto M^{3.5}$	The empirical average over $0.5\lesssim M/M_\odot\lesssim10$ ; homology with mixed microphysics reproduces it.
High mass, electron scattering	$L\propto M^{3}$	Constant opacity $\kappa=\kappa_{\mathrm{es}}$ flattens the relation to the Eddington slope; radiation pressure later pushes toward $L\propto M$ .
Radiation-dominated limit	$P\to \tfrac13 aT^4,\; L\to L_{\mathrm{Edd}}\propto M$	Above $\sim50\,M_\odot$ radiation pressure dominates, the ideal-gas law fails, and $L$ approaches the linear Eddington limit.
Degeneracy limit	$P_{\mathrm{deg}}\propto\rho^{5/3},\; R\propto M^{-1/3}$	At low mass electron degeneracy replaces the ideal gas; the exponents invert (radius shrinks with mass) and homology no longer applies.

Common trapHomology exponents are universal laws that hold across the whole main sequence

They are not. Every exponent — $\nu$ , $\xi$ , the HR slope — is valid only inside a single physical regime and changes wherever the dominant microphysics changes. Applying one $L\propto M^{3.5}$ from red dwarfs to O stars is wrong twice over: the opacity switches from Kramers ( $\nu\approx5$ ) to electron scattering ( $\nu\approx3$ ) near a solar mass, and the burning switches from the pp chain ( $\eta\approx4$ ) to the CNO cycle ( $\eta\approx16$ ) near $1.3\,M_\odot$ , each bending the slope. Worse, the whole power-law framework collapses once the equation of state stops being an ideal gas — under degeneracy at the bottom of the main sequence, or radiation pressure at the top — where the exponents invert or vanish. Homology tells you the slope within a regime and, just as usefully, where the regimes end.

Worked examples

1 · A ten-solar-mass star, fully scaled from the Sun

Take the Sun as the reference ( $M_\odot$ , $L_\odot$ , $T_{c,\odot}=1.57\times10^7$ K, nuclear lifetime $\tau_\odot\approx10$ Gyr) and predict the luminosity, radius, central temperature and main-sequence lifetime of a $10\,M_\odot$ star using homology with $\nu=3.5$ and $\xi=0.6$ .

Step 1. Luminosity from the mass–luminosity relation: $L=L_\odot\,(M/M_\odot)^{3.5}=L_\odot\times10^{3.5}$ . Since $10^{3.5}=3162$ , the star shines at $L\approx\mathbf{3200\,L_\odot}$ . Observed early-B stars near $10\,M_\odot$ sit around $5000\,L_\odot$ — homology lands within a factor of $1.6$ , the order-of-magnitude fidelity it promises.

Step 2. Radius from the mass–radius relation: $R=R_\odot\,(M/M_\odot)^{0.6}=R_\odot\times10^{0.6}\approx\mathbf{4\,R_\odot}$ , close to the measured $3.5$ – $4\,R_\odot$ . The radius grows far more slowly than the mass — the hallmark of the large denominator $3+\eta$ .

Step 3. Central temperature from $T_c\propto M/R$ : $T_c=T_{c,\odot}\,(M/M_\odot)/(R/R_\odot)=1.57\times10^7\times(10/3.98)\approx\mathbf{3.9\times10^7\ K}$ , about $2.5$ times hotter than the Sun's — hot enough that the CNO cycle, not the pp chain, dominates the burning.

Step 4. Lifetime from $\tau_{\mathrm{nuc}}\propto M/L$ : $\tau=\tau_\odot\,(M/M_\odot)^{1-\nu}=10\ \mathrm{Gyr}\times10^{-2.5}\approx\mathbf{32\ Myr}$ , against the Sun's ten billion — matching the observed $\sim25$ – $30$ Myr and explaining why massive stars are found only in young regions. Summary: $L\approx3200\,L_\odot$ , $R\approx4\,R_\odot$ , $T_c\approx3.9\times10^7$ K, $\tau\approx32$ Myr.

2 · The slope of the main sequence in the HR diagram

Combine the mass–luminosity ( $\nu=3.5$ ) and mass–radius ( $\xi=0.7$ ) relations to predict how luminosity scales with effective temperature along the main sequence, and check the sense against the observed diagram.

Step 1. Start from the definition of effective temperature, $L=4\pi R^2\sigma T_{\mathrm{eff}}^4$ , so $T_{\mathrm{eff}}^4\propto L/R^2$ . This is the only new physics; the rest is substitution.

Step 2. Insert the homology laws $L\propto M^{\nu}$ and $R\propto M^{\xi}$ : $T_{\mathrm{eff}}^4\propto M^{\nu}/M^{2\xi}=M^{\nu-2\xi}=M^{2.1}$ , giving $T_{\mathrm{eff}}\propto M^{0.525}$ . Effective temperature therefore rises with mass — massive stars are blue, low-mass stars red.

Step 3. Eliminate the mass between $L\propto M^{\nu}$ and $T_{\mathrm{eff}}\propto M^{(\nu-2\xi)/4}$ : $L\propto T_{\mathrm{eff}}^{\,4\nu/(\nu-2\xi)}=T_{\mathrm{eff}}^{\,14/2.1}\approx\mathbf{T_{\mathrm{eff}}^{6.7}}$ . A steep positive slope — brighter stars are hotter — exactly the diagonal locus of the main sequence.

Step 4. Sanity-check on the $10\,M_\odot$ star: $T_{\mathrm{eff}}=T_{\mathrm{eff},\odot}\times10^{0.525}=5772\ \mathrm{K}\times3.35\approx\mathbf{1.9\times10^4\ K}$ , a blue B-type temperature, against an observed $\sim22\,000$ K — the right regime from pure scaling. The main sequence is the homology prediction $L\propto T_{\mathrm{eff}}^{\sim6.7}$ , a steep bright-is-hot diagonal.

Physics of the portrait

The homology relations are the theoretical spine of the mass–luminosity relation, and that relation belongs to one man above all: Arthur Eddington, who argued in the 1920s that a star's brightness is set by how fast radiation diffuses out of it, regardless of the nuclear furnace within — the very content of the constant-opacity derivation above.

Origins · 1 minArthur Eddington and how this was discovered

Arthur Eddington

1882–1944 · the mass–luminosity relation and the standard radiative model of stars

Arthur Stanley Eddington built the first physically grounded theory of stellar interiors and, in doing so, discovered the mass–luminosity relation that homology reproduces in a page. In his 1924 paper "On the relation between the masses and luminosities of the stars" he showed that if radiation pressure and radiative transport govern a star's structure, luminosity should climb roughly as the cube of the mass — and the few stars with dynamically measured masses fell on the predicted line, including giants that theory said should not differ from dwarfs at fixed mass. His 1926 treatise The Internal Constitution of the Stars codified the "standard model": a star is a ball of ideal gas plus radiation, its luminosity throttled by opacity rather than by the unknown energy source — a decade before anyone knew that source was nuclear fusion. This is the profound, initially controversial insight that homology makes transparent: $L$ is fixed by what the envelope can leak, not by what the core produces. Eddington also introduced the luminosity limit that bears his name and sparred with Chandrasekhar over the fate of massive stars.

History

1870

— Lane derives the first equilibrium equation for a self-gravitating gaseous star

1907

— Emden systematizes polytropic gas spheres in Gaskugeln

1924

— Eddington announces the mass–luminosity relation L ∝ M³

1926

— Eddington's Internal Constitution of the Stars codifies the standard radiative model

1938

— von Weizsäcker and Bethe identify the pp chain and CNO cycle as stellar energy sources

Going further

Texts. Carroll and Ostlie, An Introduction to Modern Astrophysics, 2nd ed., ch. 10–11 · Kippenhahn, Weigert and Weiss, Stellar Structure and Evolution, 2nd ed., ch. 20 · Hansen, Kawaler and Trimble, Stellar Interiors, 2nd ed., ch. 1 · Prialnik, An Introduction to the Theory of Stellar Structure and Evolution, ch. 5 · Eddington, The Internal Constitution of the Stars (1926).

Provenance. The homology scalings follow the standard treatment in Carroll and Ostlie (ch. 10–11), Kippenhahn, Weigert and Weiss (ch. 20), and Hansen, Kawaler and Trimble (ch. 1). A verifier should recompute: (1) $P_c\propto M^2/R^4$ from hydrostatic equilibrium; (2) $T_c\propto\mu M/R$ from the ideal gas; (3) $L\propto\mu^4 M^3/\kappa$ from constant-opacity diffusion, with every power of $R$ cancelling; (4) the mass–radius exponent $\xi=(2+\eta-\nu)/(3+\eta)=15/19\approx0.79$ for CNO ( $\eta=16,\nu=3$ ); (5) the HR slope $L\propto T_{\mathrm{eff}}^{4\nu/(\nu-2\xi)}\approx T_{\mathrm{eff}}^{6.7}$ . Opacity exponents (Kramers $n=1,\,s=3.5$ ; scattering $n=s=0$ ), burning sensitivities ( $\eta\approx4$ pp, $\approx16$ CNO), and the observed slope ( $\nu\approx3.5$ ) are standard textbook values; the worked-example numbers and solar anchors ( $T_{c,\odot}=1.57\times10^7$ K, $T_{\mathrm{eff},\odot}=5772$ K, $\tau_\odot\approx10$ Gyr) were checked numerically. No fabricated constants are used.

Prerequisites. Equations of Stellar Structure · Hydrostatic Equilibrium in Stars · Equation of State for Stellar Matter · Opacity and Radiative Transfer in Stars · Energy Generation Equation. Leads to. Radiative Temperature Gradient · Convective Energy Transport · Mixing-Length Theory.

Ask the tutor: Want me to walk through how the homology exponents shift as you move from a red dwarf to an O star, regime by regime?

Why does the luminosity not depend on the nuclear energy-generation rate?

Because radiative diffusion, not nuclear burning, is the bottleneck. A star's luminosity is whatever its opacity-throttled envelope can carry outward for the given $T_c$ and $R$ , which the constant-opacity derivation shows equals $\mu^4 M^3/\kappa$ with no factor of $\varepsilon$ . The nuclear rate is not irrelevant — it fixes $R$ and $T_c$ so power produced matches power leaked. But raising the reaction rate at fixed mass would not brighten the star; it would contract and cool until production fell back to the fixed leak. Luminosity is set by the exit, not the furnace.

What exactly makes two stars "homologous"?

Two stars are homologous when their interiors have the identical dimensionless shape — the same curves of $\rho(r)/\rho_c$ , $P(r)/P_c$ , $T(r)/T_c$ and $m(r)/M$ against $r/R$ — differing only by the scale factors $M$ and $R$ . This requires the same equation of state, opacity law and transport mechanism throughout both, so the structure equations share one functional solution. Real stars are only approximately homologous over limited mass ranges, which is why the exponents shift between regimes. The assumption is a similarity ansatz, like calling two triangles similar: it relates their properties by ratios without solving each in full.

Why is the observed slope

L\propto M^{3.5}

rather than a single clean integer?

Because the main sequence spans two opacity regimes, and $3.5$ is the effective average. Massive stars are dominated by constant electron-scattering opacity, which the Eddington derivation gives as $\nu=3$ ; low-mass stars are dominated by Kramers opacity, $\kappa\propto\rho\,T^{-3.5}$ , whose temperature sensitivity steepens the slope toward $\nu\approx5$ . A power law fitted across the whole sequence therefore returns an intermediate exponent near $3.5$ ; fit only the upper main sequence and you recover something closer to $3$ . The single number is a convenient summary, not a law of nature.

How can such crude scaling arguments get the right answers?

Homology works because the structure equations are dominated by their scaling behaviour, not by the order-unity constants it discards. Replacing a derivative by a ratio of characteristic scales loses factors of a few but preserves the powers of $M$ and $R$ exactly, and those powers set the slopes. The similarity assumption does real work — it guarantees the discarded constants are the same for every star in the family, so they cancel in any ratio. What homology cannot deliver is the absolute normalisation: it gives $L\propto M^3$ but not the coefficient, which needs a full model.

Where does the mass–radius relation get its very shallow slope?

From the ferocious temperature sensitivity of nuclear burning. Setting the diffusive luminosity equal to the integrated nuclear luminosity, the radius appears raised to the power $3+\eta$ , where $\eta$ is the burning-rate temperature exponent — about $4$ for pp and $16$ for CNO. Solving for the radius divides the mass exponent by this large number, so $R$ responds only weakly to $M$ : $\xi\approx0.6$ – $0.8$ . Physically, a star that tried to shrink would heat its core, and because burning scales as $T^{16}$ the output would skyrocket and puff it back out — a thermostat that keeps main-sequence radii in a narrow band.

Why does homology predict convective cores in massive stars?

Homology says the central temperature scales as $\mu M/R$ , so massive stars have hot cores — hot enough that the CNO cycle, with its $\varepsilon\propto T^{16}$ sensitivity, dominates the pp chain. That steep dependence concentrates the energy generation into a tiny central region, and cramming a large luminosity through a small area demands a gradient too steep for radiation alone to carry, so by the Schwarzschild criterion the core becomes convective. The same scaling that makes massive stars hot inside thus predicts, qualitatively, the convective cores that detailed models confirm and asteroseismology now measures — a nontrivial success for a page of algebra.