The precession that switches the cycle off

An inclination can turn into an eccentricity, and the previous rung of this anchor established the mechanism: a distant companion, averaged over both orbits, conserves one combination of the inner orbit’s eccentricity and inclination and lets the two exchange. Above a critical inclination the exchange becomes an oscillation that can drive a nearly circular orbit to an eccentricity of 0.99 and back.

That is the mechanism in isolation. Nothing in astronomy is in isolation, and the thing that most reliably interferes with it is the inner orbit’s own relativity.

Why the mechanism needs the pericentre to stay put #

The cycle is not a resonance between periods. It is a secular effect: both orbits are averaged over, so the individual positions of the bodies do not matter, and what is left is a slow exchange between the shapes of two rings of mass.

What makes the exchange coherent is the argument of pericentre. The outer body’s averaged quadrupole field is not spherically symmetric about the inner orbit’s plane, so the torque it exerts on the inner ring depends on where that ring’s long axis points. If the long axis stays put, the torque acts in the same sense for a long time and the eccentricity climbs. If the long axis rotates, the torque reverses as it goes round and averages away.

So the mechanism is a competition between two rotations of the same pericentre: the one the outer body drives, and every other one.

The competitor that cannot be avoided #

An orbit precesses for many reasons. A planet’s oblateness precesses the orbits of its moons; the other planets precess each other’s, through the secular exchange that gives no planet an eccentricity of its own; and a tidal bulge on either body precesses the pair. Those depend on the system, and in principle a system can be found without them.

Relativity is not optional. Any two bodies in orbit have a pericentre that advances, at a rate

$$\dot\omega_{\rm GR} = \frac{3 G (M_1+M_2)\, n}{c^2 a (1 - e^2)},$$

and the rate has two properties that matter here. It is always in the same direction, prograde. And it grows as the pericentre falls — the factor of one over one minus the square of the eccentricity is exactly the term that blows up as the cycle drives the orbit inward.

The threshold, and why it is sharp #

Putting the relativistic precession into the averaged equations adds one term to the rate of change of the argument of pericentre, and the hero figure is the result of integrating the modified system for a range of values of that term.

Two features of the result are worth separating. The first is that the transition is at a ratio of order one — which is not a surprise, since the only dimensionless quantity in the problem is the ratio of the two precession rates. The second is that it is sharp: the greatest eccentricity falls from its full value to a third of it over less than a decade in the ratio.

The sharpness comes from the shape of the phase portrait. Adding a drift shifts the fixed points and shrinks the islands; when the drift exceeds a threshold the islands disappear entirely, and there is no intermediate configuration in which a small island supports a small oscillation. What is on the other side is not a weaker cycle but no cycle.

The floor it puts on delivery #

The reason anyone cares is that the mechanism is the leading candidate for making close-in giant planets and close binaries, by a route that does not require anything to migrate through a disc.

The route is: a distant companion drives the inner orbit to high eccentricity; at the pericentre of that eccentric orbit, tides raised on the inner body dissipate energy; the orbit shrinks and circularises at roughly twice its pericentre distance. It is called high-eccentricity migration, and unlike disc migration it can happen long after the gas has gone and it naturally produces the misaligned orbits that are observed.

The quench sets a floor on that pericentre. As the cycle drives the eccentricity up, the relativistic precession rate rises as one over one minus the square of the eccentricity, so a mechanism that was subdominant at the start of the cycle can become dominant partway through and stop the cycle before it reaches the eccentricity that would have delivered the planet.

The result is a maximum eccentricity that depends on the inner orbit’s semi-major axis, and hence a minimum pericentre. Planets cannot be delivered arbitrarily close by this route: below a certain final orbital period, relativity would have stopped the cycle that was supposed to produce them.

What sets the ratio in practice #

Written out, the ratio of the two rates is

$$\epsilon_{\rm GR} \;\sim\; \frac{3 G M_1^2\, a_2^3 (1-e_2^2)^{3/2}}{a_1^4\, c^2\, M_2},$$

with the subscripts distinguishing the inner pair from the outer companion. The dependences are steep in every variable. A closer inner orbit quenches harder, as the inverse fourth power. A more distant or lighter companion quenches harder, because it drives the cycle more weakly.

The practical consequence is a region of parameter space rather than a line. Triples with a wide, light outer companion and a compact inner binary are quenched; triples with a close, heavy companion and a wide inner binary are not. Both exist.

What was actually measured #

None of this has been watched. A Kozai cycle in a stellar triple takes ten thousand to a million years, and the quench is a statement about what does not happen.

Three lines of evidence bear on it, and all are statistical.

The first is the period distribution of close binaries. A substantial fraction of short-period binaries have a distant third companion — a much higher fraction than for wide binaries — which is what the mechanism predicts, since the tight pairs are supposed to have been made by it. Establishing that a pair has a third member at all often requires resolving two components a Fourier transform can barely tell apart.

The second is the obliquity distribution of hot Jupiters. Planets delivered by high-eccentricity migration should arrive on orbits misaligned with their star’s spin, because the mechanism works by tilting the orbit; planets delivered through a disc should arrive aligned. Both populations are observed, in a proportion that depends on stellar type in a way tidal realignment explains. The third is the pile-up in orbital period. Hot Jupiters cluster near three days, and twice the tidal circularisation radius for a Jupiter around a Sun-like star is about that — a pile-up visible in the occurrence-rate census rather than in any individual system. Whether the inner edge of that pile-up shows the floor this essay predicts is a question about a few dozen objects, and the answer is not yet decisive.

Three rates, and which one is largest where #

It helps to put numbers on the competition for two systems at opposite ends of the range.

Take a Jupiter at five astronomical units from a solar-mass star, with a stellar companion of half a solar mass at five hundred astronomical units. The cycle’s period is of order a million years. The relativistic precession of the inner orbit is about a hundredth of an arcsecond a century at zero eccentricity — utterly negligible — but at an eccentricity of 0.99 the factor of one over one minus the square of the eccentricity is fifty, and the pericentre has fallen by a factor of a hundred, so the rate has risen by five thousand. The cycle is unquenched at the start and quenched near the top, which is exactly the case the floor argument is about.

Now take a pair of white dwarfs at a hundredth of an astronomical unit with the same outer companion. The relativistic precession is a degree a year at zero eccentricity, and the cycle’s own rate, which falls as the cube of the ratio of the two semi-major axes, is nothing at all. This system never cycles.

The lesson is that the same triple geometry gives completely different answers depending on the inner orbit’s size, and that the dependence is a fourth power rather than a linear one. There is no gradual regime.

The other things that precess #

Relativity is the quencher this essay is named for and it is not the only one. Anything that turns the argument of pericentre faster than the cycle does will do the same job, and in most real systems something else gets there first.

Tidal bulges. Each body raises a bulge on the other, and a bulge is a mass distribution that is not a point, so the orbit precesses. The rate falls as the eighth power of the separation — far steeper than the relativistic term’s fourth — so at very small pericentre distances the tidal term dominates everything. For a Jupiter approaching a solar-type star inside about ten stellar radii, tidal precession quenches the cycle before relativity does.

Rotational flattening. A rapidly rotating star is oblate, and an oblate primary precesses an orbit around it by the same mechanism that regresses a satellite’s node about the Earth. Young stars rotate fast, so the term is largest exactly when the systems in question are being assembled, and it also falls steeply with separation.

A third orbit. In a system with more than one planet, each planet precesses the others, and those secular rates are often faster than a distant companion’s cycle. That is the standard reason a compact multi-planet system is immune to a stellar companion that would otherwise wreck it: the planets protect each other by precessing each other, and removing one of them can destabilise the rest.

The competition is therefore between four rates with four different dependences on separation — the cycle itself scaling with the cube of the ratio of the two semi-major axes, relativity as the inverse fourth power of pericentre, tides as the inverse eighth, and mutual planetary precession as whatever the system’s own architecture gives. Which one wins is a different answer at every separation, and the boundaries between them are sharp for the reason the whole essay is about: a ratio of rates crossing one is not a gradual transition.

How high the eccentricity would have gone #

The quench is a statement about a cycle that would otherwise have run, and quantifying it needs the unquenched amplitude — which at quadrupole order has a closed form worth having.

For an inner orbit starting nearly circular at inclination $i_0$ to the outer one, the cycle exchanges inclination for eccentricity while conserving $\sqrt{1-e^2}\cos i$, and the maximum reached is

$$e_{\max} = \sqrt{1 - \tfrac{5}{3}\cos^2 i_0}.$$

Two things follow immediately. The cycle does nothing at all unless $\cos^2 i_0 < 3/5$, which is the critical inclination band from 39.2° to 140.8°; and at exactly 90° the formula gives $e_{\max} = 1$, meaning the inner orbit is driven to a collision.

The consequence is the whole reason the mechanism matters. A pair starting at a separation of five astronomical units and 85° of mutual inclination reaches an eccentricity of 0.998, and its pericentre falls to a hundredth of an astronomical unit — from a wide orbit to a grazing one, with no dissipation and no change in the semi-major axis, purely by moving angular momentum between two orbits.

That is a factor of five hundred in pericentre distance from a geometric condition on an angle, and it is why the quench is decisive rather than a correction. The relativistic precession at the top of that excursion is thousands of times its value at the start, because the rate goes as the inverse fourth power of pericentre — so the mechanism generates, at its own maximum, exactly the term that switches it off.

It also explains why the mechanism is so often invoked and so rarely demonstrated in a specific system. The excursion happens once, over a million years, and leaves behind an orbit that carries no label saying how it got there — so the evidence is always a population’s shape rather than an individual’s history.

The same reasoning applies to the systems where nothing happened. A wide, coplanar pair with a distant companion is not evidence that the mechanism is weak; it is evidence that the inclination was outside the band, which is a statement about how the system formed rather than about how it evolved.

The wider lesson #

The quench is a specific instance of a general rule about secular mechanisms, and the rule is worth stating on its own.

A secular effect works by accumulating a small perturbation coherently over many orbits. Anything that rotates the geometry of the accumulation faster than the effect itself destroys the coherence, and the competition is decided by a ratio of rates rather than by a ratio of forces. So a mechanism can be switched off by something far weaker than itself, provided the weaker thing acts on the right angle.

That is why precession budgets are worth constructing at all. A term that is negligible in the force is not negligible in the argument of pericentre, and the second is what secular theory is about.

There is a second reading of the same rule that is worth keeping, because it runs the other way. If a weak competitor can switch off a strong mechanism, then measuring whether a mechanism operated is a measurement of the competitor. A triple observed to have a tight, misaligned inner pair is a triple in which the relativistic term was small enough to lose, which bounds the inner orbit’s size at the time the cycle acted — a bound on a configuration that no longer exists, obtained from one that does.

That kind of inference is characteristic of this whole anchor. Secular mechanisms leave no trace of their operation except the final state, so every argument about them is a reconstruction: this arrangement is reachable and that one is not, and the observed population is the intersection of what is reachable with what survives. The quench is valuable precisely because it is a sharp boundary rather than a gradual falling-off, and a sharp boundary in a reconstruction is worth several soft ones.

One practical consequence deserves stating because it is the reason the quench appears in the literature far more often than its size would suggest. Numerical work on triples is usually done with the secular equations rather than by integrating every orbit, because the cycle takes millions of inner orbital periods and a direct integration is unaffordable. The relativistic term has to be added to those secular equations by hand — it is not there unless somebody puts it there — and for two decades a good deal of published work omitted it. The results were not wrong so much as answers to a different question: they described what a triple would do in a universe governed by Newtonian gravity, and the systems most affected were exactly the compact ones the calculations were aimed at. The correction is one line of code and it changes the predicted merger rate of compact binaries formed this way by an order of magnitude, which is a large return on a term nobody can measure directly.

One more starting argument of periapsis shows that the quenching does not depend on the phase.

Where the ladder goes #

The next rung is the octupole term, which is what happens when the outer orbit is itself eccentric. At quadrupole order the inner orbit’s angular momentum along the outer orbit’s axis is conserved and the inclination cannot pass through ninety degrees; at octupole order it is not, and the inner orbit can flip from prograde to retrograde. That extends the reach of the mechanism enormously and is much harder to quench.

The other direction is the competition with gravitational radiation in compact triples, where the cycle drives the pericentre down and the radiation shrinks the orbit — a race whose outcome is the rate at which black hole binaries in dense environments merge.