Exoplanet Detection Lab · Guide · glidepathscience.org
GuideExoplanet Detection Lab
Contents
This Guide explains the ideas and equations behind every Walkthrough step, Concept Challenge question and Predict round, in the order a textbook takes them. Part 1 prepares the Intro track and Part 2 the Advanced track; each takes about twelve minutes to read. Each section ends with a worked example that is not one of the sim’s own problems.
Part 1 · The essentials
Math you’ll need: algebra, powers and roots, and scientific notation; no calculus.
A star you cannot resolve
Thousands of planets are now known around other stars, yet almost none has ever been photographed. The reason is contrast, not distance alone. A planet shines only by the starlight it reflects (and a faint infrared glow of its own), and a Jupiter-sized planet reflects about a billionth of its star’s light. Seen from a few tens of light-years, it sits less than an arcsecond from that star, inside the glare that every telescope spreads around a bright point of light. Picking it out is like spotting a firefly beside a searchlight from the other side of a city.
So astronomers find planets indirectly, from what a planet does to its star’s light. Two effects do almost all the work. If the planet’s orbit happens to be lined up with our line of sight, the planet passes in front of its star once each orbit and blocks a little of its light: a transit. And whether or not it transits, the planet tugs its star around in a small orbit of its own, so the star moves alternately toward us and away from us, and its light is shifted back and forth in wavelength: the radial-velocity (RV) method. This sim shows both signals coming from one orbit, on one clock.
Transits: the light curve, its depth and its period
A space telescope pointed at one star for weeks records its brightness every few minutes. Plotted against time, this is the star’s light curve. Most of the time it is flat, apart from noise. When a planet crosses the star’s disk, it blocks part of the disk and the curve dips; when it moves off, the curve recovers. Because the planet repeats its orbit, the dips repeat, evenly spaced.
How deep is a dip? The planet is a small dark disk in front of a large bright one, so the fraction of light it blocks is the fraction of the star’s area it covers. A disk’s area is $πR^2$, so the transit depth is the ratio of the two areas:
$δ = (R_p/R_s)^2$
where $R_p$ is the planet’s radius and $R_s$ the star’s. Because the radii are squared, small planets make very shallow dips. Jupiter is about a tenth of the Sun’s radius and would dim it by about 1 %; Earth, a hundredth of the Sun’s radius, by less than one part in ten thousand, 84 parts per million (ppm). The most common slip is to take the radius ratio itself for the depth, which overstates a 1 % dip tenfold.
Turning the depth around gives the planet’s size, in units of the star’s:
$R_p = R_s\sqrt δ$
The star’s radius comes from separate measurements of its temperature and brightness, so a planet is only ever as well measured as its star. (This uniform-disk depth is the Intro model; real stars are a little brighter in the middle, which the Advanced track adds.)
The time from one dip to the next is the planet’s orbital period, $P$. With several dips it is best to use the first and the last, dividing the time between them by the number of gaps, which is one fewer than the number of dips:
$P = (t_{last} − t_{first})/(n − 1)$
The orbit’s size, and why most planets never transit
The period tells us how big the orbit is, once we know the star’s mass. Newton’s form of Kepler’s third law, for a planet much lighter than its star on a circular orbit of radius $a$, is
$a^3 = GM_sP^2/4π^2$
with $G$ = 6.674 × 10⁻¹¹ N·m²/kg² and $M_s$ the star’s mass. In astronomers’ units (AU, years and solar masses) it becomes simply $a^3 = M_sP^2$. A lighter star pulls more weakly, so a planet with the same period must orbit closer. The star’s mass, like its radius, comes from its spectrum and brightness.
Planets with periods of a few days turn out to orbit at a few hundredths of an AU, far closer to their stars than Mercury is to the Sun. That matters for finding them. A transit happens only if Earth lies nearly in the plane of the planet’s orbit: seen from the planet, Earth must be within about one stellar radius of that plane. For an orbit tilted at random, the chance of that is about
$p ≈ R_s/a$
An Earth twin at 1 AU from a Sun-like star has a chance of only 0.47 %, about 1 in 215; a planet in a 3.5-day orbit, about 10 %. Most planets, even large ones, never transit for us. Tilt an orbit by a single degree and an Earth twin passes nearly four stellar radii above its star’s disk.
The wobble: Doppler shifts and minimum mass
A planet does not orbit a fixed star. The star and the planet pull on each other equally, so both orbit their common center of mass, the star in a small orbit and the planet in a large one, on opposite sides. The center of mass sits a distance $aM_p/(M_s + M_p)$ from the star’s center. For the Sun and Jupiter that is about 1.07 solar radii, just outside the Sun’s surface, and the Sun circles it every 11.9 years at about 12.5 m/s.
As the star moves toward us and away, the waves of its light are compressed and stretched: the Doppler shift. For speeds far below the speed of light $c$,
$v = cΔλ/λ$
where $λ$ is a spectral line’s wavelength at rest and $Δλ$ its shift. A shift to longer wavelengths (a redshift) means the star is moving away, by convention a positive radial velocity; a blueshift means it is approaching. The shifts are tiny: 25 m/s moves the 656.28 nm hydrogen line by 5.5 × 10⁻⁵ nm. Spectrographs reach such precision by measuring thousands of lines at once against a stable reference.
Over an orbit the star’s radial velocity swings back and forth: a sine curve for a circular orbit, whose half-height is the semi-amplitude $K$. A heavier planet pulls the star harder, so $K$ grows in proportion to the planet’s mass; a heavier star is harder to move; and a planet farther out moves more slowly, so $K$ falls as the period grows (as $P^{−1/3}$). The planet’s size does not enter at all.
Solving for the planet’s mass gives the RV method’s answer, for a circular orbit and a planet much lighter than its star:
$M_p\sin i = K(P/2πG)^{1/3}M_s^{2/3}$
The $\sin i$ is unavoidable. The spectrograph sees only the part of the star’s motion along our line of sight; an orbit seen at inclination $i$ (90° is edge-on) shows only $\sin i$ of it. A light planet seen edge-on and a heavier one seen more nearly face-on can give the same curve. So the RV method alone gives the minimum mass $M_p\sin i$. Remember to work in SI units throughout: a period left in days is the usual slip.
Noise, averaging, and density from both methods
Real light curves are noisy. A star’s brightness measured every half hour scatters by hundreds of ppm, often more than a small planet’s dip. The cure is to average. If a planet’s transits are folded on top of one another, lining up each at the period, the average of $N$ independent measurements scatters less than one does:
$σ_{avg} = σ/\sqrt N$
Random errors partly cancel, so the noise falls as the square root of the number averaged, not in proportion: four times the data halves the noise. A common slip is to divide by $N$. This is why small planets in long orbits are hard to find: an Earth twin transits once a year, so years of data give only a handful of transits.
The two methods measure different things, and together they are far more than the sum of their parts. A transit gives the planet’s radius, and shows that its orbit is nearly edge-on, so $\sin i$ is close to 1 and the RV gives the true mass. Mass and radius together give the planet’s mean density:
$ρ = 3M/(4πR^3)$
Density separates rock from gas. Earth averages 5,510 kg/m³ and Jupiter 1,240 kg/m³; water is 1,000. Many hot Jupiters come out at a few hundred kg/m³, puffed up by their stars’ heat.
Part 2 · Going further
Math you’ll need: trigonometry and inverse sines, Kepler’s equation (used, not derived), and least-squares fitting described in words.
Transit geometry, limb darkening and the star’s density
Signal sizes set the scale of the problem. An Earth twin dims a Sun-like star by 84 ppm and moves it by 9 cm/s; a Jupiter twin, by about 1 % and 12.5 m/s. The best spectrographs reach about 0.3 m/s, and a star’s own surface, boiling with convection, jitters by about as much: the Earth twin’s wobble is still out of reach.
A transit’s shape carries more than its depth. The impact parameter $b$ is the planet’s closest approach to the center of the disk, in stellar radii. For a circular orbit,
$b = (a/R_s)\cos i$
Edge-on, $b$ = 0 and the planet crosses the full diameter; near $b$ = 1 it grazes the limb. The transit has four contacts: the planet’s disk first touches the star (1), is fully inside (2), touches the far edge (3) and leaves (4). With $k = R_p/R_s$, the total and flat-bottomed durations of a circular transit are $T_{14} = (P/π)\arcsin\left[\sqrt{(1 + k)^2 − b^2}\,/\,((a/R_s)\sin i)\right]$ and $T_{23}$, the same with $1 − k$. A central transit has the simple form
$T_{14} = (P/π)\arcsin[(1 + k)/(a/R_s)]$
An off-center chord is shorter and its ingress and egress longer, so the ratio of $T_{23}$ to $T_{14}$ measures $b$. When $b > 1 − k$ the planet never fits wholly inside the disk: a grazing transit, V-shaped and shallower than $k^2$.
Real stars are not uniform disks. Near the edge, or limb, we look through the star’s atmosphere at a slant and see higher, cooler, dimmer gas. The quadratic law describes the brightness at a point where the line of sight meets the surface at angle $θ$, with $μ = \cos θ$: $I(μ)/I(1) = 1 − u_1(1 − μ) − u_2(1 − μ)^2$. Limb darkening rounds the bottom of the dip and makes it deeper than $k^2$ at the center of the disk, shallower near the limb; a fit that ignores it gets the radius wrong.
Because the shape gives $a/R_s$ directly, Kepler’s third law turns it into the star’s mean density, with no need for its mass or radius separately: from $a^3 = GM_sP^2/4π^2$ and $ρ_s = M_s/(\frac43πR_s^3)$,
$ρ_s = 3π(a/R_s)^3/(GP^2)$
This is a valuable check: a “planet” whose light curve implies a star far denser or less dense than the spectrum says is probably something else.
Searching: box searches, aliases and folding
A shallow transit can be invisible in any single night’s data. Surveys therefore search: for each of thousands of trial periods they fold the light curve and slide a box-shaped dip around the fold, recording how far the points inside the box sit below the rest, divided by that difference’s standard error. Plotted against the trial period, the best box’s signal-to-noise is a periodogram, and the true period stands out as its tallest peak.
Weaker peaks appear at simple fractions and multiples of the period, the aliases. Folded on twice the period, only every other transit lines up at one phase (the others sit half a cycle away); folded on half the period, half the box’s points fall where nothing happens. Folding on each candidate settles it: at the true period every transit lands at the same phase, and odd and even transits match.
Folding is averaging. If each transit has $n$ points in it and $N$ transits are folded, the depth is measured from $nN$ points, each with scatter $σ$, so its signal-to-noise is
SNR $= (δ/σ)\sqrt{nN}$
Surveys usually ask for an SNR of about 7 or more before calling a candidate, because a search over thousands of trial periods gives noise thousands of chances to fake a peak.
The full semi-amplitude and eccentric orbits
Without the approximations of Part 1, the star’s semi-amplitude for a planet of any mass on an orbit of any shape is
$K = (2πG/P)^{1/3}M_p\sin i/[(M_s + M_p)^{2/3}\sqrt{1 − e^2}]$
The $(M_s + M_p)^{2/3}$ is the total mass Kepler’s law really contains; the $\sin i$ projects the motion onto our line of sight; and the $\sqrt{1 − e^2}$ comes from eccentricity $e$, since an elongated orbit is swept faster near its closest point. Solving this for $M_p\sin i$ exactly takes a short iteration, because $M_p$ appears on both sides.
On an eccentric orbit the star’s velocity is no longer a sine wave: $v = K[\cos(ω + f) + e\cos ω] + γ$, where $f$ is the true anomaly (the angle from periastron, found from the time through Kepler’s equation $M = E − e\sin E$), $ω$ is the argument of periastron and $γ$ the system’s own velocity. By Kepler’s second law the motion is fastest at periastron, so the curve swings sharply there and drifts slowly for the rest of the orbit. Its lopsidedness gives $e$, and where the sharp swing falls gives $ω$. (This sim measures $ω$ for the planet’s orbit; many RV catalogs quote the star’s, 180° away.)
Fitting: χ², reduced χ² and uncertainties
Reading values off a curve by eye goes only so far. A fit adjusts a model’s parameters to make it agree with the data as well as possible, judging agreement by
$χ^2 = Σ[(v_i − \mathrm{model}_i)/σ_i]^2$
where each data point $v_i$ has error bar $σ_i$. Each term asks how many error bars the model misses a point by. A good model with honest error bars misses each point by about one error bar, so $χ^2$ comes out near the number of points less the number of fitted parameters, the degrees of freedom:
reduced $χ^2 = χ^2/(N − n_{par})$
A reduced $χ^2$ near 1 is what we hope for. Well above 1, the points scatter more than their bars claim: a missing noise source (a star’s own jitter, added in quadrature to each $σ_i$), or a wrong model (a second planet). Well below 1, the error bars are too large.
The fit’s uncertainties come from how fast $χ^2$ rises away from its minimum. Near the best value $K_0$ it is a parabola, $χ^2 = χ^2_{min} + ((K − K_0)/σ_K)^2$, so the values where it has risen by 1 mark one standard deviation either side, and a rise of 4 marks two. If the other parameters are refitted at each trial value, their correlations are folded into this one interval.
False positives, true masses and selection bias
Not every periodic dip is a planet. The commonest impostor is an eclipsing binary: two stars orbiting each other and taking turns in front. Alone, such a pair gives deep eclipses, often tens of percent; blended with a brighter star in the same pixel, its eclipses are diluted to planet-like depths. The tell-tales are dips of two alternating depths (each star eclipsed in turn), a secondary eclipse halfway between, a V shape, a density from the light curve that contradicts the star’s spectrum, and, when an RV is taken, a wobble of kilometers per second rather than meters. Grazing transits and spotted stars, whose brightness rises and falls smoothly as they rotate, can also confuse a search.
Once a planet is confirmed to transit, the two methods combine. The transit gives $b$ and $a/R_s$, and so the inclination, $\cos i = b/(a/R_s)$; the RV gives $M_p\sin i$, so the true mass is $M_p = (M_p\sin i)/\sin i$, and with $R_p$ the density. For a transiting planet $\sin i$ is within about a percent of 1, so the real gain is the radius beside the mass.
Finally, the planets found are not a fair sample. Both methods favor large planets close to their stars: their transits are deep and frequent and their orbits likely to be aligned, and their stars wobble hard. The first planets found around Sun-like stars were mostly hot Jupiters, yet surveys corrected for these biases find them around only about 1 % of such stars, while planets smaller than Neptune turn out to be common.
Reference
Common mistakes
Part 1
- Taking the radius ratio for the depth: the depth is its square, $δ = (R_p/R_s)^2$.
- Using the depth as a percent inside the square root: $R_p = R_s\sqrt δ$ needs it as a fraction.
- Dividing the time between the first and last dips by the number of dips instead of the number of gaps.
- Leaving the star’s mass out of Kepler’s third law, or the period in days inside SI formulas.
- Thinking every planet transits: only orbits seen almost edge-on do, about $R_s/a$ of them.
- Treating the RV mass as the true mass: without the inclination it is a minimum, $M_p\sin i$.
- Dividing the noise by $N$ when averaging: it falls as $\sqrt N$.
Part 2
- Using sin $i$ for the impact parameter: $b = (a/R_s)\cos i$, zero edge-on.
- Fitting a limb-darkened star with a uniform-disk model: the radius comes out wrong.
- Taking an alias for the period: check the fold at half and twice the period.
- Dividing $χ^2$ by $N$ instead of $N − n_{par}$, or reading $Δχ^2$ = 4 as one standard deviation.
- Calling any deep, periodic dip a planet: look for alternating depths, a secondary eclipse and a V shape.
The equation sheet
| Part 1 | |
|---|---|
| depth | $δ = (R_p/R_s)^2$; $R_p = R_s\sqrt δ$ |
| period | $P = (t_{last} − t_{first})/(n − 1)$ |
| Kepler | $a^3 = GM_sP^2/4π^2$ |
| transit chance | $p ≈ R_s/a$ |
| Doppler | $v = cΔλ/λ$ |
| minimum mass | $M_p\sin i = K(P/2πG)^{1/3}M_s^{2/3}$ |
| averaging | $σ_{avg} = σ/\sqrt N$ |
| density | $ρ = 3M/(4πR^3)$ |
| Part 2 | |
|---|---|
| impact parameter | $b = (a/R_s)\cos i$ |
| duration ($b$ = 0) | $T_{14} = (P/π)\arcsin[(1 + k)/(a/R_s)]$ |
| star’s density | $ρ_s = 3π(a/R_s)^3/(GP^2)$ |
| semi-amplitude | $K = (2πG/P)^{1/3}M_p\sin i/(M_s + M_p)^{2/3}$, divided by $\sqrt{1 − e^2}$ |
| folding | SNR $= (δ/σ)\sqrt{nN}$ |
| fit quality | reduced $χ^2 = χ^2/(N − n_{par})$ |
| transit chance | $p = (R_s/a)(1 + e\sin ω)/(1 − e^2)$ |
Symbols: $δ$ the transit depth; $R_p$, $R_s$ the planet’s and star’s radii, $k = R_p/R_s$; $M_p$, $M_s$ their masses; $P$ the period; $a$ the orbit’s semi-major axis; $i$ the inclination (90° edge-on); $b$ the impact parameter; $T_{14}$ the transit’s duration; $e$ the eccentricity and $ω$ the argument of periastron; $K$ the star’s RV semi-amplitude; $v$ a radial velocity, positive moving away; $λ$, $Δλ$ a line’s wavelength and its shift; $c$ the speed of light; $G$ the gravitational constant; $σ$ one point’s scatter; $n$ the points in one transit (or the number of dips); $N$ the transits folded (or the data points in a fit); $n_{par}$ the fitted parameters; $ρ$ a density; $p$ a probability.
Sources and further reading
- OpenStax, Astronomy 2e, chapter 21, §21.4 Planets beyond the Solar System: Search and Discovery (openstax.org): an open introduction to both methods.
- J. N. Winn, “Transits and Occultations,” in Exoplanets, ed. S. Seager (University of Arizona Press, 2010): the geometry, durations and impact parameter used here.
- S. Seager and G. Mallén-Ornelas, The Astrophysical Journal 585, 1038 (2003): the star’s density from a light curve.
- K. Mandel and E. Agol, The Astrophysical Journal 580, L171 (2002): transits with limb darkening; G. Kovács, S. Zucker and T. Mazeh, Astronomy & Astrophysics 391, 369 (2002): the box-search periodogram.
- C. Lovis and D. Fischer, “Radial Velocity Techniques for Exoplanets,” in Exoplanets (2010); M. Perryman, The Exoplanet Handbook, 2nd ed. (Cambridge, 2018).
- This sim’s Field Manual (H) for the model and its limits; Gravitation & Satellite Orbits and the Three-Body Lab for Kepler’s laws and the center of mass.
Exoplanet Detection Lab v1.1 · © 2026 Dr. David Meier · CC BY-NC-SA 4.0 · glidepathscience.org