TL;DR
Radial velocity (RV) exoplanet surveys typically report $M_T \sin(i)$, the product of a planet's true mass and the sine of its orbital inclination. This paper demonstrates, using Bayesian analysis, that the distribution of $\sin(i)$ is not uniform, but depends crucially on the underlying distribution of exoplanet masses. This means that estimates of planet masses derived from RV data alone are significantly more uncertain than commonly assumed.
What Problem Does This Paper Address?
When detecting exoplanets using the radial velocity (RV) method, we measure the wobble of a star caused by an orbiting planet. This wobble allows us to determine $M_T \sin(i)$, where $M_T$ is the planet's true mass and $i$ is the inclination of its orbit relative to our line of sight. Since we can't directly measure $i$ from RV data alone, $\sin(i)$ is treated as a random variable. The standard assumption is that $\sin(i)$ is uniformly distributed between 0 and 1, implying an isotropic distribution of orbital inclinations. This assumption simplifies mass estimations, allowing us to derive an upper bound on the planet's true mass $M_T$ by multiplying $M_T \sin(i)$ by the inverse of the average $\sin(i)$ value (typically 2). This paper challenges this assumption, arguing that the distribution of $\sin(i)$ is not independent of the planet's true mass distribution, which is currently unknown. This dependence can lead to significant underestimation of the uncertainty in true mass estimates.
Methods
The authors employ a Bayesian framework to derive the posterior distribution of $\sin(i)$ given an observed $M_T \sin(i)$ and an assumed intrinsic distribution of exoplanet masses, $P(M_T)$. The core of their method involves applying Bayes' theorem:
$P(M_T | M_o) = \frac{P(M_o | M_T) P(M_T)}{P(M_o)}$
where $M_o = M_T \sin(i)$. They then explore various "toy models" for $P(M_T)$, including:
- Power-law mass distributions: $P(M_T) \propto M_T^\alpha$.
- Power-law plus a delta function: Incorporating a characteristic mass scale ($M_c$).
- Broken power-law mass distributions: Mimicking distinct populations (e.g., rocky vs. gas giants).
- Solar System-like mass distribution: A simplified model based on our own solar system.
For each assumed $P(M_T)$, they derive the corresponding posterior distributions for $M_T$ and $\sin(i)$ and analyze their implications.
Datasets
Not discussed in the paper. The analysis relies on theoretical "toy models" for exoplanet mass distributions rather than specific observational datasets.
Key Insights and Ideas
- The $\sin(i)$ distribution is not uniform: The most crucial insight is that the probability distribution of $\sin(i)$ for a planet with a measured $M_T \sin(i)$ is not uniform between 0 and 1. It is intrinsically linked to the underlying distribution of exoplanet masses.
- Mass distribution dictates $\sin(i)$: If the intrinsic mass distribution $P(M_T)$ is heavily weighted towards low-mass planets, then for a given $M_o$, the planet is more likely to have a large $\sin(i)$ (i.e., be viewed face-on). Conversely, if $P(M_T)$ favors high-mass planets, then for a given $M_o$, the planet is more likely to have a small $\sin(i)$ (i.e., be viewed edge-on).
- Significant mass uncertainty: The uncertainty in true mass ($M_T$) derived from $M_o$ can be much larger than typically assumed. For plausible power-law mass distributions, the 95% confidence upper bound on $M_T$ can be 4.5 to 400 times larger than $M_o$.
- Characteristic mass scales matter: If exoplanet populations have characteristic mass scales (e.g., a transition between rocky and giant planets), the $\sin(i)$ distribution becomes more complex and can exhibit non-trivial dependencies on the ratio of $M_o$ to $M_c$.
Key Results
- Median $\sin(i)$ varies widely: For power-law mass distributions with indices between -2 and -0.5, the median value of $\sin(i)$ can range from 0.25 to 0.71, deviating significantly from the assumed 0.5 for a uniform distribution.
- Upper bounds on $M_T$ are substantially larger: The 95% confidence upper bound on $M_T$ can be 4.5 to 400 times larger than $M_o$ for power-law indices between -2 and -0.5, respectively.
- Illustrative plots show complex distributions: Figures 1-4 visually demonstrate how different assumed mass distributions lead to varied posterior probabilities for $\sin(i)$ and $M_T$. For instance, a power-law with a steep negative index (many low-mass planets) leads to a higher median $\sin(i)$, while a shallow or positive index (more high-mass planets) leads to a lower median $\sin(i)$.
Limitations and Future Work
- Unknown true mass distribution: The primary limitation is our current ignorance of the true intrinsic distribution of exoplanet masses ($P(M_T)$). The paper uses "toy models" to illustrate the effect of this distribution, but a definitive answer requires better empirical or theoretical constraints on $P(M_T)$.
- Observational selection effects: The analysis assumes that RV surveys uniformly sample exoplanet populations. The authors acknowledge that this is unrealistic and that selection biases (e.g., favoring planets with certain masses or orbital periods) will influence detected samples. They suggest that a parallel analysis incorporating selection functions could be performed.
- Future work: The authors propose investigating more realistic selection functions that depend jointly on $M_T$ and $\sin(i)$. They also advocate for RV studies to report upper bounds on $M_T$ based on explicit assumptions about the underlying mass distribution, rather than solely relying on $M_T \sin(i)$.
Summary updated