Orbital perturbations

Draft

The two-body model has an exact closed-form solution and is wrong. Everything useful in operational astrodynamics comes from understanding how it is wrong, by how much, and over what timescale.

Magnitudes

Accelerations for a 400 km circular Earth orbit, where the central term is 8.68 m/s²:

Perturbation	Acceleration	Relative	Character
$J_2$ oblateness	1.2 × 10⁻² m/s²	1.4 × 10⁻³	Secular in $\Omega$ and $\omega$
Higher harmonics	10⁻⁵ and below	10⁻⁶	Mostly periodic; resonant for some orbits
Atmospheric drag	10⁻⁷ to 10⁻⁵ m/s²	Varies	Dissipative, secular in $a$
Lunar third body	1.2 × 10⁻⁶ m/s²	1.3 × 10⁻⁷	Long-period and secular
Solar third body	5.4 × 10⁻⁷ m/s²	6.2 × 10⁻⁸	Long-period and secular
Solar radiation pressure	10⁻⁸ to 10⁻⁷ m/s²	~10⁻⁸	Periodic, secular with eclipses
Relativistic correction	~10⁻⁸ m/s²	~10⁻⁹	Secular in $\omega$

The column that matters is the last one. A perturbation two orders of magnitude smaller than another can dominate the error budget if it acts secularly while the larger one averages out over an orbit.

Non-spherical gravity

Earth’s gravitational potential expands in spherical harmonics:

U = \frac{\mu}{r}\left[1 - \sum_{n=2}^{\infty}J_n\left(\frac{R_\oplus}{r}\right)^nP_n(\sin\phi) + \text{tesseral and sectorial terms}\right]

$J_2 = 1.0826 \times 10^{-3}$ is the equatorial bulge, and it is a thousand times larger than every other harmonic combined. To first order it produces three secular rates and no others.

J2 nodal regression

\dot{\Omega} = -\frac{3}{2}\,n\,J_2\left(\frac{R_\oplus}{p}\right)^2\cos i

J2 apsidal rotation

\dot{\omega} = \frac{3}{4}\,n\,J_2\left(\frac{R_\oplus}{p}\right)^2\left(5\cos^2 i - 1\right)

Neither $a$ nor $e$ nor $i$ has a secular $J_2$ term. The bulge rotates the orbit but does not change its size or shape.

Orbit	$\dot{\Omega}$	Consequence
ISS, 400 km, 51.6°	−5.00°/day	Plane precesses fully in 72 days; drives the beta angle cycle
Sun-synchronous, 800 km, 98.6°	+0.986°/day	Matches Earth’s motion about the Sun, by construction
GEO, 0°	−0.013°/day	Small, but folded into east-west station keeping

These two expressions are not corrections to be minimised. They are design tools: sun-synchronous orbits exist because of the first, and Molniya orbits at the 63.435° critical inclination exist because the bracket in the second vanishes there. See Orbit types and regimes.

Tesseral harmonics, which vary with longitude, matter mainly for orbits in resonance with Earth’s rotation. $J_{22}$ is what drives geostationary satellites toward the stable longitudes discussed in Station keeping.

Atmospheric drag

\mathbf{a}_{\text{drag}} = -\frac{1}{2}\rho\,\frac{C_DA}{m}\,v_{\text{rel}}^2\,\hat{\mathbf{v}}_{\text{rel}}

Drag is the only perturbation in the list that is dissipative. Everything else exchanges energy between orbital elements; drag removes it permanently, so its effect on semi-major axis accumulates monotonically and ends in re-entry.

It also produces a genuinely counterintuitive result. Drag decelerates the spacecraft, but the resulting lower orbit is faster. The satellite loses energy and gains speed. This is the satellite drag paradox, and it follows directly from vis-viva: a lower $a$ means a higher $v$ at a given $r$ .

The hard part is not the equation, it is $\rho$ .

Solar cycle: Density at 400 km varies by more than an order of magnitude between solar minimum and maximum.
Geomagnetic storms: Can double or triple density within hours, with no useful advance warning.
Diurnal bulge: The heated dayside atmosphere is denser than the nightside at the same altitude.
Models: NRLMSISE-00, JB2008 and DTM are the standards. All are driven by solar and geomagnetic indices, which must themselves be forecast.

Because density prediction is really solar activity prediction, re-entry forecasts carry wide error bars until the final few orbits. The drag coefficient is equally uncertain: in free molecular flow $C_D$ depends on surface accommodation and attitude, and values between 2.0 and 2.6 are all defensible. In practice $C_D$ and $A$ are absorbed into a single fitted parameter estimated from tracking data rather than computed from first principles.

Third bodies

\mathbf{a}_3 = \mu_3\left(\frac{\mathbf{r}_3 - \mathbf{r}}{|\mathbf{r}_3 - \mathbf{r}|^3} - \frac{\mathbf{r}_3}{|\mathbf{r}_3|^3}\right)

The second term is essential and often omitted by mistake. The perturbation is the difference between the third body’s pull on the spacecraft and on the central body, because the frame is centred on the latter. Dropping it produces an acceleration several orders of magnitude too large.

Third-body effects scale as $r^3$ , so they are negligible in LEO and dominant in GEO and beyond. Lunisolar gravity is what drives the 0.85° per year inclination growth that consumes most of a geostationary satellite’s propellant.

Solar radiation pressure

\mathbf{a}_{\text{SRP}} = -P_{\odot}\,C_R\,\frac{A}{m}\,\hat{\mathbf{s}}

with $P_{\odot} = 4.56 \times 10^{-6}$ N/m² at 1 AU. The reflectivity coefficient $C_R$ runs from 1 for a perfect absorber to 2 for a perfect specular reflector.

SRP would average out over an orbit were it not for eclipses. Because the force switches off in Earth’s shadow, the average over a revolution is non-zero, and the result is a secular growth in eccentricity. For high area-to-mass spacecraft, large solar arrays or deployed sunshades, this is a first-order effect that must be actively controlled.

The same physics is the basis of solar sailing, where a deliberately enormous area-to-mass ratio turns a nuisance into propulsion. IKAROS and LightSail 2 both demonstrated measurable orbit change from radiation pressure alone.

Special and general perturbations

Two philosophies, with a clean division of labour.

	Special perturbations	General perturbations
Method	Numerically integrate the full force model	Analytically average the equations, then propagate mean elements
Accuracy	As high as the force model and integrator allow	Limited by the truncation of the analytical theory
Cost	Expensive; scales with time span	Very cheap; nearly independent of time span
Output	Osculating elements, the instantaneous true orbit	Mean elements, with short-period terms removed
Used for	Precise orbit determination, manoeuvre planning, conjunction assessment	Catalogue maintenance, SGP4, constellation design studies

Cowell’s method integrates the total acceleration directly and is what operational orbit determination uses. Encke’s method integrates only the deviation from a reference conic and was important when computation was scarce.

The general-perturbation approach is why the public satellite catalogue is distributed as two-line elements: mean elements plus a matched analytical theory compress an orbit into two lines of text and propagate in microseconds. The price is that TLE elements are meaningless outside SGP4, and that kilometre-level accuracy is the ceiling. See Orbital elements.

References

Vallado, D. A. Fundamentals of Astrodynamics and Applications, 5th ed., Microcosm Press, 2022, chapters 8 and 9.
Montenbruck, O. and Gill, E. Satellite Orbits: Models, Methods and Applications, Springer, 2000.
Picone, J. M. et al. “NRLMSISE-00 empirical model of the atmosphere”, Journal of Geophysical Research, 107(A12):1468, 2002.
Bowman, B. R. et al. “A New Empirical Thermospheric Density Model JB2008”, AIAA 2008-6438.
Brouwer, D. “Solution of the problem of artificial satellite theory without drag”, The Astronomical Journal, 64:378, 1959.