---
title: Wind-Momentum-Luminosity Relation
url: https://www.emergentmind.com/topics/wind-momentum-luminosity-relation
type: topic
---

# Wind-Momentum-Luminosity Relation

Searching arXiv for recent and foundational papers on the Wind–Momentum–Luminosity Relation.
The **Wind-Momentum-Luminosity Relation** (WLR) is the empirical and theoretical relation linking the luminosity of a hot star to a wind-momentum diagnostic constructed from its mass-loss rate, terminal wind velocity, and radius. In its standard form for line-driven winds of hot luminous stars, the relevant quantity is the **modified wind momentum**, commonly written as \(D_{\rm mom}=\dot{M}v_\infty \sqrt{R_\ast}\) or equivalently with an explicit \(R_\odot\) normalization depending on units [2008.06066]. The WLR emerged as a consequence of CAK-type radiative driving, in which the explicit stellar-mass dependence is largely removed from the momentum diagnostic, leaving luminosity as the dominant control parameter [2008.06066]. Across the literature, the WLR functions both as a scaling law for OB-star winds and as a stress test for wind theory: it is well behaved in some regimes, steepens or develops curvature in others, and can cease to be a unique relation when the wind becomes weak, composition-sensitive, or transitions away from the classical line-driven OB-star regime [1010.2178] [1406.1288] [2202.07811].

## 1. Canonical formulation and physical basis

The standard WLR uses the modified wind momentum
\[
D_{\rm mom}=\dot{M}\,v_\infty \left(\frac{R}{R_\odot}\right)^{1/2},
\]
or, in equivalent notation, \(D_{\rm mom} \equiv \dot{M} v_\infty \sqrt{R_\ast}\) when units are specified accordingly [1406.1288] [2008.06066]. Here \(\dot M\) is the mass-loss rate, \(v_\infty\) the terminal wind velocity, and \(R\) the stellar radius. The relation is usually cast in log-linear form,
\[
\log D_{\rm mom} = \log D_0 + \frac{1}{\alpha}\log L,
\]
or with fitted slope and intercept notation such as
\[
\log D = \beta\,\log L/L_\odot + \alpha,
\]
depending on the paper’s convention [2503.23932] [2202.07811].

The physical rationale comes from radiatively driven wind theory. In the CAK-style picture quoted in several papers, the mass-loss rate depends on luminosity and effective mass, while terminal velocity scales with the escape speed, so that the combination \(\dot M v_\infty \sqrt{R_\ast}\) is, to first order, mainly a function of luminosity [2008.06066]. One formulation states
\[
\dot{M} \propto L_\ast^{1/\alpha'} M_{\rm eff}^{\,1-\frac{1}{\alpha'}},
\qquad
v_\infty \propto v_{\rm esc} \propto \sqrt{\frac{M_{\rm eff}}{R_\ast}},
\]
with \(M_{\rm eff}=M_\ast(1-\Gamma_e)\), implying \(D_{\rm mom}\propto L_\ast^{1/\alpha'}\) [2008.06066]. A closely related formulation for empirical interpretation writes
\[
g_{rad}^{lines} \propto N_{eff}L \left( \frac{dv/dr}{\rho} \right)^\alpha,
\]
leading to
\[
\dot{M} \propto N_{eff}^{1/\alpha'}L^{1/\alpha'}(M(1-\Gamma))^{1-1/\alpha'},
\]
with \(\alpha'=\alpha+\delta\), so that the WLR slope is approximately \(1/\alpha'\) and the intercept depends on the effective number of driving lines \(N_{eff}\), hence on metallicity [2202.07811].

This theoretical structure also explains why the WLR has historically been attractive as a unifying representation of OB-star winds. It packages the dynamical wind output into a quantity that should correlate tightly with luminosity if the wind is driven by line absorption in metal lines [2008.06066] [1204.1200]. At the same time, several later studies show that the apparent simplicity of the canonical relation obscures significant dependencies on metallicity, wind ionization, and the weak-wind regime [1010.2178] [1406.1288] [2202.07811] [2503.23932].

## 2. Empirical calibration for OB stars and its revision

Modern empirical WLR work has emphasized homogeneous multiwavelength analyses, especially those combining UV and optical diagnostics. A major CMFGEN-based compilation for Milky Way and Small Magellanic Cloud massive stars defines the WLR in the standard form \(D_{mom}=\dot M\,V_\infty \sqrt{R_\star}\), converts clumped mass-loss rates to homogeneous equivalents via
\[
\dot{M}_{uncl} = \frac{\dot{M}_{cl}}{\sqrt{f}},
\]
and fits the empirical relation as
\[
\log D = \beta\,\log L/L_\odot + \alpha
\]
[2202.07811].

For the preferred “consistent” UV+optical CMFGEN sample, the paper derives
\[
\log D_{MW} = (4.16 \pm 0.23)\,\log L/L_\odot + (5.43 \pm 1.28),
\]
\[
\log D_{SMC} = (3.85 \pm 0.29)\,\log L/L_\odot + (6.67 \pm 1.52),
\]
over a broad luminosity range extending roughly from \(\log L/L_\odot \sim 4.6\) to \(6.1\) [2202.07811]. These slopes are much steeper than the empirical Mokiem et al. relations quoted in the same study,
\[
\log D_{MW} = (1.84 \pm 0.17)\,\log L/L_\odot + (18.87 \pm 0.98),
\]
\[
\log D_{SMC} = (1.84 \pm 0.19)\,\log L/L_\odot + (18.20 \pm 1.09),
\]
because the newer compilation includes low-luminosity stars with UV-derived weak winds that were not part of the classic luminous-star calibration [2202.07811].

This empirical revision is not merely a refit of the bright-star relation. The paper explicitly argues that the high-luminosity behavior remains broadly compatible with earlier work, whereas the low-luminosity extension fundamentally changes the relation. When only bright stars are fit, much shallower slopes are recovered, about \(\beta_{MW}\sim 1.7\) and \(\beta_{SMC}\sim 2.2\), which the authors state are in reasonable agreement with older calibrations [2202.07811]. This suggests that the canonical WLR survives in the luminous regime but becomes much steeper once weak-wind stars are included.

A complementary theoretical grid of hydrodynamically consistent O-star models in the Galaxy, LMC, and SMC yields a new predicted WLR of the form
\[
\begin{split}
\log\left(D_{\rm mom}\right) & =  -1.55 + 0.46\log\left(\frac{Z_{\ast}}{Z_{\odot}}\right) + \\
&\left[2.07 - 0.73\log\left(\frac{Z_{\ast}}{Z_{\odot}}\right)\right]\log\left(\frac{L_{\ast}}{10^{6}L_{\odot}}\right),
\end{split}
\]
with \(D_{\rm mom}\) in \(M_\odot\,{\rm yr}^{-1}\,{\rm km\,s}^{-1}\,R_\odot^{0.5}\) [2008.06066]. Environment-by-environment fits give WLR slopes
\[
x_{\rm Gal}=2.07\pm0.32,\qquad x_{\rm LMC}=2.12\pm0.34,\qquad x_{\rm SMC}=2.56\pm0.44
\]
[2008.06066]. These predicted slopes are steeper than classical empirical values, particularly in the SMC, and the authors note visible curvature, especially for low-luminosity dwarfs [2008.06066].

Taken together, these results show that “the WLR” is no longer best understood as a single universal straight line for all hot stars. Even within the O-star domain, the inferred slope depends on the luminosity range sampled, the treatment of weak winds, and whether the relation is empirical or theoretical [2008.06066] [2202.07811].

## 3. Metallicity dependence and luminosity-dependent steepening

Metallicity enters the WLR because line driving depends on metal-line opacity. In the theoretical O-star grid, the mean metallicity exponent for wind momentum is \(n=0.85\) with \(1\sigma\) scatter \(0.29\), under the assumption \(D_{\rm mom}\propto Z_\ast^n\), and the luminosity-dependent approximation
\[
n(L_\ast) = -0.73\log(L_\ast/10^6L_\odot) + 0.46
\]
is built directly into the global WLR fit [2008.06066]. The same work finds \(\dot M \sim L_\ast^{2.2}\) and \(\dot M \sim Z_\ast^{0.95}\) on average, with a somewhat stronger dependence on metallicity at lower luminosity [2008.06066].

A distinct theoretical study of NLTE wind models with varying metallicity and CNO-processed surface composition reaches a more severe conclusion at very low metallicity. It defines the same modified wind momentum and argues that for \(Z/Z_\odot \gtrsim 0.1\), \(D_{\rm mom}\) depends mainly on luminosity, but for
\[
Z/Z_\odot \lesssim 0.1
\]
the relation shows large scatter and ceases to be unique [1406.1288]. The paper’s metallicity fit for mass loss is
\[
\log \frac{\dot M(Z)}{\dot M(Z_\odot)} = 0.46 \log\left(\frac{Z}{Z_\odot}\right) -0.31 \left[\log\left(\frac{Z}{Z_\odot}\right)\right]^2,
\]
with local exponent
\[
\zeta = 0.46 - 0.62\log\left(\frac{Z}{Z_\odot}\right),
\]
and terminal velocity scaling
\[
\frac{v_\infty}{v_{\mathrm{esc}}} \sim \left(\frac{Z}{Z_\odot}\right)^{0.10}
\]
[1406.1288]. The key claim is not a new fitted WLR line, but that below \(0.1\,Z_\odot\) the relation develops strong scatter because wind driving becomes too sensitive to detailed abundances, ionization balance, and stellar parameters [1406.1288].

Empirical work reaches a different but related conclusion. The CMFGEN-based MW/SMC study infers that for bright stars, roughly \(\log L/L_\odot \gtrsim 5.4\), the mass-loss metallicity dependence is
\[
\dot M \propto Z^{0.5-0.8},
\]
while terminal speed scales as
\[
V_\infty \propto Z^{0.1-0.2},
\]
but toward lower luminosity the metallicity dependence weakens and may even vanish [2202.07811]. The authors explicitly state that MW and SMC stars separate clearly in the WLR only at high luminosity, whereas at lower luminosity the two samples begin to overlap [2202.07811].

A 2025 study of O stars at sub-SMC metallicity extends this luminosity-dependent picture. For stars in dwarf galaxies with \(Z<0.2\,Z_\odot\), the modified wind momentum is found to be consistent with an extrapolated empirical \(D_{\rm mom}(L,Z)\) relation, but current theoretical prescriptions fail at low luminosity and low metallicity, predicting winds stronger by an order of magnitude or more [2503.23932]. For the \(Z\sim0.14\,Z_\odot\) subsample, the fitted WLR parameter is
\[
\alpha = 0.21 \pm 0.04
\]
for the classical fit
\[
\log D_{\rm mom} = \log D_0 + \frac{1}{\alpha}\log L,
\]
and
\[
\alpha = 0.18 \pm 0.06
\]
when the \(M_{\rm eff}\) term is retained [2503.23932]. Since the slope is \(1/\alpha\), these correspond to empirical slopes of about \(4.8\) and \(5.6\), much steeper than the canonical O-star expectation based on \(\alpha\sim2/3\) [2503.23932].

The combined implication is that metallicity does not simply shift the WLR vertically by a fixed amount. The data and models instead indicate luminosity-dependent metallicity behavior, steepening at low \(L\), and—in some empirical analyses—a near-collapse of MW–SMC separation at the faint end [2202.07811] [2503.23932].

## 4. Weak winds and the low-luminosity breakdown

The **weak-wind problem** is the central modern complication in WLR work. A conference study centered on late O and early B dwarfs states that there is “an apparent break-down” of radiation-driven wind theory at
\[
\log L/L_\odot < 5.2,
\]
where the observed wind momentum is smaller, beyond error bars, than predicted by theory [1010.2178]. In that work, the WLR is used diagnostically rather than recalibrated: the observed wind strengths are compared against the theoretical relation of Vink et al. (2000), and the stars are found systematically below it [1010.2178].

The observational basis is twofold. First, UV spectra of these weak-wind objects show almost no classical wind lines, with only C IV \(\lambda1550\) and perhaps N V \(\lambda1240\) visible in some cases. Second, \(H_\alpha\) does not exhibit a wind profile, so optical work yields only upper limits [1010.2178]. The optical wind-strength invariant is explicitly defined as
\[
Q=\dot M (v_{\infty} R_{\star})^{-1.5},
\]
but because \(H_\alpha\) is insensitive, only upper limits on \(Q\) can be derived, and these propagate into upper limits on wind momentum and hence on WLR placement [1010.2178].

The same luminosity threshold recurs in later work. The 2025 sub-SMC study states that theoretical prescriptions do not match the results or other recent analyses at low luminosity,
\[
L \lesssim 10^{5.2}\,L_{\odot},
\]
and low \(Z\), where they predict winds stronger by an order of magnitude or more [2503.23932]. The CMFGEN empirical MW/SMC study likewise ties the steepened low-luminosity WLR to UV-derived weak winds of order \(10^{-9}\,M_\odot\,{\rm yr}^{-1}\), much smaller than classical recipes [2202.07811].

Methodology is decisive here. Several papers insist that UV diagnostics are essential because optical recombination lines lose sensitivity once \(\dot M\) falls below about \(10^{-7}\,M_\odot\,{\rm yr}^{-1}\) [2503.23932] [2202.07811]. The sub-SMC analysis concludes that if only optical spectra are used, mass-loss rates may be overestimated “by as much as an order of magnitude,” which directly inflates \(D_{\rm mom}\) [2503.23932]. The earlier weak-wind study similarly found that standard WM-basic mass-loss rates overproduced UV P Cygni features and that only extremely small \(\dot M\) values could match the data [1010.2178].

A common misconception is that the low-\(L\) WLR anomaly is purely a metallicity effect seen only in metal-poor galaxies. The 2010 weak-wind study explicitly rejects that interpretation by emphasizing Galactic O9–B0.5 dwarfs in Orion, plus 10 Lac and \(\tau\) Sco, as weak-wind objects that also fall below the canonical relation [1010.2178]. The empirical MW/SMC study reaches a related conclusion from the opposite direction: at low luminosity, MW and SMC stars can overlap in wind strength, so the expected metallicity separation may be greatly reduced [2202.07811].

This suggests that the faint-end WLR is not merely the classical relation with lower normalization. A plausible implication is that it constitutes a physically distinct weak-wind branch, requiring additional physics or a reformulation of the line-driving statistics in very low-density winds [1010.2178] [2503.23932].

## 5. Diagnostic methodologies and model dependence

The WLR is only as reliable as the wind parameters used to place stars in the \(D_{\rm mom}\)–\(L\) plane. Across the literature, three methodological themes recur: self-consistency, wind structure assumptions, and the diagnostic domain.

Hydrodynamically consistent NLTE/CMF models are repeatedly presented as the most reliable basis for theoretical WLR work. The 2020 O-star grid computes winds by iterative solution of the equation of motion with full NLTE radiative transfer in the co-moving frame, rather than prescribing a velocity law or inferring \(\dot M\) from a global energy argument [2008.06066]. A separate study of central stars of planetary nebulae stresses that only such consistent modeling can truly test wind theory, because \(v_\infty\) and \(\dot M\) are explicit functions of the stellar parameters rather than independent fit knobs [1204.1200]. That paper supports the physical basis of the WLR for O-type CSPNs with strong winds, while arguing that some adopted stellar masses and luminosities from evolutionary tracks are inconsistent with the observed wind dynamics [1204.1200].

Empirical work likewise stresses simultaneous UV+optical analysis. The weak-wind study contrasts an earlier UV-only WM-basic analysis, which used \(T_{\rm eff}\) and \(\log g\) from separate FASTWIND fits, with a newer CMFGEN treatment of UV and optical spectra together [1010.2178]. The improved modeling of HD 37020 still yields very weak UV wind signatures and no clear \(H_\alpha\) wind feature, supporting the conclusion that low wind momentum is not merely an artifact of inconsistent photospheric parameters [1010.2178].

Clumping complicates both empirical and theoretical WLR placement. In the sub-SMC study, the main analysis assumes optically thin clumping with a void interclump medium,
\[
\rho_{\rm cl} = f_{\rm cl}\,\langle\rho\rangle,\qquad \rho_{\rm ic}=0,
\]
but optically thick clumping tests show that individual stars can shift substantially in inferred \(\dot M\) [2503.23932]. For one star, allowing optically thick clumping raises \(\dot M\) by a factor of \(\sim 6\); for another, by a factor of \(\sim 10\) [2503.23932]. The empirical MW/SMC WLR standardizes all mass-loss rates to unclumped values using \(\dot M_{uncl}=\dot M_{cl}/\sqrt f\), but residual heterogeneity remains because the source papers used different atomic models and detailed assumptions [2202.07811].

X-rays can also bias WLR inference. The weak-wind paper notes that if X-rays are produced within \(\lesssim 1\,R_\ast\) of the photosphere, they may affect effective-temperature diagnostics as well as wind ionization, thereby perturbing both luminosity and wind-parameter estimates [1010.2178]. This is important because WLR placement depends on both axes.

These methodological caveats do not invalidate the WLR, but they do mean that the relation is not a purely observational construct. It is a derived scaling whose apparent slope, intercept, and scatter can change when the analysis moves from optical-only to UV+optical, from prescribed winds to self-consistent hydrodynamics, or from optically thin to more realistic clumping treatments [1010.2178] [2202.07811] [2503.23932] [1204.1200].

## 6. Extensions, alternative regimes, and breakdown of universality

The WLR is most secure in the domain of classical line-driven OB-star winds. Outside that regime, related momentum–luminosity diagnostics remain useful, but they no longer map cleanly onto a single canonical relation.

For **A-type supergiants**, standard fast m-CAK solutions predict mass-loss rates and terminal velocities that are too high. A 2011 study identifies a new slow solution at low rotation and high \(\delta\), yielding \(V_\infty \sim 150{-}250\ {\rm km\,s^{-1}}\), \(V_\infty < V_{\rm esc}\), and modified wind momenta that align well with the empirical A-supergiant WMLR of Kudritzki et al. (1999) [1105.5576]. The paper interprets the A-supergiant WLR discrepancy not as a failure of radiation driving per se, but as evidence that these stars inhabit a different hydrodynamic branch controlled by ionization stratification [1105.5576].

For **central stars of planetary nebulae**, the WLR appears to extend from massive O stars to O-type CSPNs with pronounced winds. The modified wind momentum is explicitly written as
\[
D = \dot{M} v_\infty \sqrt{R},
\]
and the authors state that observed and computed CSPN wind momenta lie along the extrapolated O-star relation [1204.1200]. Yet the same paper emphasizes that wind momenta are “essentially independent of the stellar masses,” so WLR agreement does not by itself validate the assumed mass–luminosity relation of CSPNs [1204.1200]. A different diagnostic, \(v_\infty/v_{\rm esc}\), is needed to expose inconsistencies in the adopted masses and radii [1204.1200].

For **luminous blue variables and Of/late-WN transition objects**, the standard OB-star WLR is often inappropriate. The study of Var 2 in M33 does not construct a canonical modified-wind-momentum relation, but instead uses the wind efficiency parameter
\[
\eta \equiv \frac{\dot{M} v_\infty}{L/c},
\]
and argues that Var 2 has a wind-momentum-to-luminosity ratio consistent with late-WN stars even though its velocity-law shape resembles that of OB supergiants [2510.11802]. This places the star in a transitional regime rather than in a clean WLR calibration sample [2510.11802]. A separate LBV study in NGC 1156 provides enough parameters for a WLR-style estimate but explicitly does not compare the star against published WLR calibrations [2209.06012]. These cases show that once dense, optically thick, or extended atmospheres dominate, the choice of radius and the meaning of the momentum diagnostic become ambiguous.

For **continuum-driven, super-Eddington, rapidly rotating winds**, the classical WLR breaks down more fundamentally. A theoretical paper on continuum-driven winds from rotating stars argues that such outflows do not obey the standard CAK-type WLR. Instead, the mass-loss rate scales with the excess above a critical or Eddington luminosity,
\[
-\dot M \propto \Gamma -1 + \frac{2}{3}\Omega^2,
\]
and the flow becomes strongly latitude dependent and photon-tiring limited [1206.6078]. In this regime, a single scalar modified wind momentum is insufficient to characterize the outflow [1206.6078].

Finally, work on **partially stripped stars** suggests another route by which the WLR may need reinterpretation. A 2025 study does not provide a formal WLR, but shows that partially stripped stars can be substantially more luminous than pure-He stars at the same mass, with corresponding hydrodynamically consistent atmosphere models yielding a 1.5 dex increase in mass-loss rate and about a 1 dex decrease in terminal velocity for a 0.34 dex increase in luminosity [2508.14161]. This suggests that any WLR analysis for stripped stars must account for internal structure and partial stripping, not just mass or surface composition [2508.14161].

The common pattern is that the WLR remains a powerful concept, but its universality is limited. Once the wind becomes weak, composition-sensitive, optically thick, continuum-driven, or structurally anisotropic, the relation must either be modified or replaced by a different momentum–luminosity diagnostic [1010.2178] [1406.1288] [2510.11802] [1206.6078].

## 7. Interpretation, applications, and current status

The WLR continues to serve three distinct functions in contemporary research. First, it is a compact empirical summary of radiatively driven wind behavior in hot stars. Second, it is a discriminant between competing theoretical prescriptions. Third, it is a diagnostic of missing physics when stars deviate systematically from canonical expectations.

As a theoretical benchmark, the WLR has been used to compare prescriptions such as Vink et al. (2001), Björklund et al. (2021), and Krtička & Kubát (2018). In the sub-SMC analysis, all three yield shallower \(D_{\rm mom}(L)\) relations than observed at low \(Z\); Björklund performs best overall, Krtička somewhat worse, and Vink worst of all, with the latter overpredicting mass-loss rates of most stars by more than a factor of ten [2503.23932]. The 2020 CMF/NLTE O-star grid likewise finds a WLR normalization lower by about \(0.5\) dex relative to Vink et al. and mass-loss rates reduced by factors of 2 or more [2008.06066].

As an empirical framework, the WLR remains strongest for bright O stars and related wind-strong objects. The MW/SMC CMFGEN analysis states that for bright stars the classical picture survives: higher luminosity implies stronger winds and lower metallicity shifts the WLR downward [2202.07811]. The CSPN analysis similarly supports the physical applicability of the O-star WLR to O-type central stars with strong winds [1204.1200]. A plausible implication is that in these regimes the WLR remains a meaningful calibration tool for stellar evolution and feedback studies.

At low luminosity and/or low metallicity, however, the WLR is increasingly a probe of failure modes rather than a stable scaling law. Below \(\log L/L_\odot \sim 5.2\), empirical wind momenta fall systematically below canonical theory in both Galactic and metal-poor environments [1010.2178] [2503.23932]. Below \(Z/Z_\odot \lesssim 0.1\), theoretical models indicate large scatter and even no-wind solutions, making any unique WLR questionable [1406.1288]. In the empirical MW/SMC compilation, the metallicity dependence appears to weaken dramatically at low luminosity, further undermining the notion of a single metallicity-shifted relation [2202.07811].

There is also no consensus that all deviations reflect the same missing physics. Proposed explanations include metallicity-dependent wind initiation thresholds, early evolutionary effects, decoupling of driving ions from the bulk plasma, unresolved hot shocked wind components, and magnetic or X-ray overionization [1010.2178] [2202.07811] [2503.23932]. The 2010 weak-wind study is especially explicit that magnetism is presented as a promising ingredient rather than an established solution [1010.2178].

The present state of the subject is therefore best described as **stratified** rather than uniform. The WLR is robust as a first-order consequence of line driving in luminous hot stars, but it is not a single immutable law across all temperatures, luminosities, metallicities, and wind regimes. Its slope can steepen, its zero point can shift, its curvature can become important, and in some regimes the relation can lose uniqueness altogether [2008.06066] [1406.1288] [2202.07811] [2503.23932]. This suggests that future WLR work will likely proceed not by searching for one universal calibration, but by mapping the boundaries between the classical OB-star relation and the physically distinct regimes in which it breaks, bends, or must be reformulated.

Source: https://www.emergentmind.com/topics/wind-momentum-luminosity-relation