Kepler's Third Law: Orbital Period, Total Mass and the Correct Semimajor Axis

Which form of Kepler's third law should you use?

For an isolated Newtonian two-body system in a bound elliptical orbit, P² = 4π²a³ / [G(M₁ + M₂)]. Here P is the orbital period and a is the semimajor axis of the relative orbit: the orbit of one body's position measured from the other. Both masses belong in the denominator.

The familiar P² ≈ a³ shortcut needs carefully chosen units and a total mass close to one solar mass. It is not a universal formula for arbitrary units, every binary star or every satellite. The worked examples below use an explicit gravitational parameter, SI conversions and a defined Julian year so the rounding can be reproduced.

The most consequential mistake usually occurs before pressing the square-root key: selecting an instantaneous separation or one star's orbit about the barycentre in place of the relative semimajor axis.

Semimajor axis is not a snapshot separation

An ellipse has a major axis running across its longest dimension. The semimajor axis is half that length. For the relative orbit, its closest and farthest separations satisfy rₚ = a(1 − e) and rₐ = a(1 + e), giving a = (rₚ + rₐ)/2. Only for a circle does the separation stay equal to a throughout the orbit.

For example, closest and farthest separations of 1 and 3 au give a = 2 au and eccentricity e = 0.5. Substituting the 3 au farthest separation into the period formula would produce a different, longer period. A photograph taken at that instant cannot convert the farthest distance into the semimajor axis by changing its label.

“Average distance” can also mislead: a is the arithmetic mean of those two extremes, not generally the separation averaged over equal time intervals. The body spends more time near its distant part of an elliptical orbit. See perihelion, aphelion and orbital speed for the geometry behind that distinction.

Why the two masses add

Let r be the separation vector from body 1 to body 2. Each body accelerates under the other's gravity. Subtracting their two acceleration equations gives r̈ = −G(M₁ + M₂)r/|r|³. This is the relative equation of motion, with gravitational parameter μ = G(M₁ + M₂).

For a circle, the relative speed is 2πa/P. Setting its centripetal acceleration equal to μ/a² gives 4π²a/P² = μ/a² and hence the period law. For an ellipse, the same result follows from the area πab, constant areal rate h/2, b = a√(1 − e²), and h² = μa(1 − e²). Substituting these into P = 2πab/h cancels the eccentricity factors.

This cancellation is the reason an ellipse and a circle can have the same period at the same a and μ. It does not say that their speeds or separations match at every point. The derivation assumes two-body gravity; significant extra forces require a fuller model.

Declare units before calculating

The IAU nominal solar parameter is an exact conversion constant. It is not a claim that the physical Sun has an eternally exact mass. Using μ directly also avoids splitting a measured gravitational parameter into a mass and an independently estimated G when that extra step is unnecessary.

With a in au, m = μ/μ☉ᴺ, and P in the Julian years defined above, the formula is P = k√(a³/m), where k = 1.000018886746. The slight departure from one matters when explaining the units, even though it is tiny for a classroom estimate.

At a = 1 au and m = 1, this model gives 365.256898 days. That is not exactly 365.25 days, nor should it be called a precise prediction of the Earth's year: the example uses a nominal solar parameter, specified ideal inputs and an isolated two-body model.

Worked period table for three total mass parameters

Each row keeps the relative semimajor axis fixed and changes the total gravitational parameter. For a = 4 au and m = 2, a³/m = 64/2 = 32. Taking its square root and multiplying by k gives 5.656961 Julian years.

Keeping m = 1 while doubling a from 1 to 2 au multiplies P by 2√2, approximately 2.828427. Keeping a fixed while doubling m divides P by √2. These ratios are more informative than memorising a long table, because the common unit coefficient cancels.

How orbit size and mass change the period

The graph has linear axes. The upward curvature comes from P being proportional to a raised to the power 3/2. The blue curve is highest because the lowest total mass parameter gives the slowest orbit at the same size. At 4 au its period is twice the red curve's period, since their parameters differ by a factor of four.

The curves deliberately use synthetic systems. Reading a point from the chart does not identify a real planet, specify its eccentricity or tell you its current position. The accompanying table gives numbers where the width of a plotted line would otherwise limit precision.

Eccentricity changes the shape, not this fixed-a period

At e = 0.8 the separation ranges from 0.4 to 3.6 au, yet the period matches the circular 2 au orbit at the same total parameter. The larger farthest separation is balanced by a much smaller closest separation while a stays fixed.

If instead you hold the closest separation fixed and increase eccentricity, then a = rₚ/(1 − e) increases, so the period does change. “Independent of eccentricity” is meaningful only after specifying which other quantities remain fixed. The four rows above make that condition explicit.

At e = 1 or greater, a bound elliptical orbital period is no longer supplied by this formula in the same way. An unbound pass does not have a recurring revolution time; inserting an unsuitable or negative axis into a square root will not create one.

A binary-star example: use a₁ + a₂

Consider a fictional binary whose component gravitational parameters are 1.5 and 0.5 nominal solar units, with relative a = 4 au. Their total is 2. The centre-of-mass condition gives a₁ = aM₂/(M₁ + M₂) = 1 au and a₂ = aM₁/(M₁ + M₂) = 3 au.

Both stars complete their barycentric orbits in the same 5.656961 years. The more massive component travels on the smaller ellipse. The relative axis is the sum of their barycentric axes, not either one separately.

Using a₁ = 1 au with the total parameter underestimates this period by a factor of eight. Using the correct 4 au but ignoring the 0.5-unit companion also fails, in the opposite direction. A planet's mass can sometimes be neglected because its contribution is small; that approximation must be checked rather than carried automatically into a stellar binary.

Use ratios when absolute constants cancel

For two systems obeying the same model, P₂/P₁ = √[(a₂/a₁)³(μ₁/μ₂)]. This avoids unnecessary unit conversions when all like quantities already share a unit. It also makes a useful independent check on a calculator result.

The row with twice the size and eight times the parameter returns the same period. That provides a quick dimensional sanity check: the cubic increase in size has exactly matched the increase in μ. A result that doubles the period in that case probably used a linear distance scaling instead.

Infer total mass from a measured orbit

Rearranging gives μ = 4π²a³/P², or m = k²a³/P² in this article's units. A hypothetical orbit with a = 4 au and P = 5 Julian years therefore implies m = 2.560097. That is the total parameter; it does not independently separate the two component masses.

To find the mass ratio as well, you need additional information, such as both barycentric axes or appropriate velocity measurements and an orbital model. In our earlier binary a₁/a₂ = 1/3 implies M₂/M₁ = 1/3. Combining that ratio with the total partitions it into 0.5 and 1.5 units.

A resolved binary's angular orbit also needs a distance scale and a fit that accounts for orientation. In the small-angle relation, a correctly fitted angular relative axis in arcseconds multiplied by distance in parsecs gives its physical axis in au. A single projected separation on an image is not that fitted axis. The parallax and parsec guide explains the distance conversion separately.

Why a modest distance error can dominate the mass estimate

Taking logarithmic differentials gives dP/P = (3/2)da/a − (1/2)dμ/μ. For the inverse problem, dμ/μ = 3da/a − 2dP/P. These are first-order relations for small changes, not exact equations for large errors.

If independent one-standard-deviation fractional uncertainties are 2 percent in a and 1 percent in P, the first-order uncertainty in μ is √[(3 × 0.02)² + (2 × 0.01)²] = 0.063246, about 6.32 percent. The cubic distance dependence dominates this example.

Real orbit fits can correlate a and P. Then the variance includes the covariance term −12 Cov(a,P)/(aP), and independent quadrature may misrepresent the uncertainty. A short observed arc can also leave a wide range of permissible orbits. Extra displayed decimals do not supply the missing observational information.

A calculation workflow that keeps the physical meaning intact

1. State whether the orbit is a bound two-body approximation.
2. Identify the relative semimajor axis and record its unit.
3. Include both masses or state why one is negligible.
4. Use μ directly when available, and declare any nominal conversion constant.
5. Convert the period to a named time unit, not an undefined “year.”
6. Check the answer with size and mass ratios.
7. Round to match the precision and uncertainty of the inputs.

Mean motion n = 2π/P is an average angular phase rate. It is not generally the instantaneous rate at which the true position angle changes around an eccentric orbit. Period alone also cannot supply where the body is now: that requires an orbital phase reference and the other orbit elements.

For practical sky planning, local sunrise and sunset provide a different kind of timing information. Telescope resolution and observing conditions determine whether a binary can be separated in an image; the resolution guide covers that question. Neither observing-time tables nor an apparent double-star image replaces a fitted dynamical orbit.

Frequently asked questions