Asteroids in the asteroid belt between Mars and Jupiter jostle for position, and occasionally collide with one another. When a large asteroid shatters as a result of such a collision, it yields a shower of fragments which comprises an asteroid 'family'.

Such family break-ups may lead to an increase in the number of asteroids that end up in the inner Solar System - and to times when the Earth is subject to potentially catastrophic asteroid showers lasting millions of years. This is the conclusion of a study by Vincenzo Zappalà of the Astronomical Observatory of Torino, Italy and colleagues, reported in the planetary science journal Icarus.

It is estimated that the impact on Earth of an asteroid 1.5 kilometres in diameter would produce a 200,000-megatonne explosion and lead - directly or indirectly - to the deaths of 1.5 billion people and the collapse of civilization. An impact of this magnitude occurs perhaps once every half-million years. Another estimate suggests that there are, at present, 1,500 so-called 'near-Earth' asteroids (NEAs) more than a kilometre in diameter, any one of which has the potential to hit the Earth. The largest known has a diameter of 8 km.

Calculation of impact hazard rests on the assumption that asteroids hit the Earth at random intervals, and that the population of NEAs is more or less constant. Zappalà and colleagues challenge those assumptions, showing that "the number of asteroids escaping the belt after some family forming events rivals or exceeds that of the entire present NEA population." Flooding the population of NEAs at certain times would lead to an increased potential for terrestrial impact.

When asteroids shatter, the family members adopt new orbits within the asteroid belt. Occasionally, though, they become injected into 'resonant' orbits, which are particularly susceptible to the gravitational push and pull of the Sun and planets. Asteroids in such orbits follow orbits which rapidly become more and more elliptical, so that they are carried out of the Asteroid Belt and into the inner Solar System. Most will be swallowed by the Sun, but a fraction will collide with the inner planets.

About twenty distinct asteroid families can be seen today, each one the product of collisions some time in the past. Zappalà and colleagues looked at eight families sufficiently close to resonant orbits that they could have been significant NEA sources at the time of the break-up events, estimated to have occurred hundreds of millions to billions of years ago.

Some of these family break-ups produced large numbers of potential impactors. For example, the break-up of the Eunomia family produced, at a conservative estimate, 9,700 1-km-diameter objects, 12 of which could have struck the Earth over a 15-million-year period. A slightly less conservative estimate (based on extrapolating the numbers of small, and thus not directly observable fragments), raises the number of 1-km-diameter objects to 110,000, of which 139 could have hit the Earth over the same period. This works out at an impact rate of, on average, of one 1-km impactor every 100,000 years or so.

This is roughly the same as the estimated frequency of impact for bodies in the 1-km size class, assuming random strikes. However, this is just one example, from one family of asteroids: nothing says that families should break up over evenly spaced intervals. From this, the researchers conclude that impacts over time ought to occur in clumps, rather than be spread out in time. How could this be tested?

The cratering record of the Earth samples predominantly the past 200 million years: craters on Earth are rare, and are quickly erased by our planet's active surface. On the plus side, the craters are well-dated - but there is "no obvious signature of a massive increase in impactors" lasting several million years over the past 100 million years or so. However, the coincidence of two known impact craters dating from around 36 million years ago is "intriguing."

Craters on the Moon are less well dated, but there are a lot more of them, and the cratering record stretches back almost to the beginning of the Solar System itself. The problem here is the reverse of that found on the Earth. Instead of too few craters to preserve a signal, there are too many. If the cratering rate increased 100-fold for a few million years, this would contribute only 10 per cent of the craters accumulated on the lunar surface - an excess, say the researchers, "below the accuracy of crater counting studies". The results of these showers would simply be lost amid the profuse cratering of the lunar surface.

Family break-up could account for a perplexing mystery: the so-called 'lunar cataclysm' around 4 billion years ago, when large areas of the lunar surface were pounded by large impactors over a short period. The mystery is how the impact rate could have increased from the normal, background rate, to produce the observed effects. The break-up of a large family close to a suitably resonant orbit could have created an effective asteroid shower. Some of these impactors would have been more than 100 km in diameter, so some of the members of that broken family would have been large indeed.