Finding Planets We Cannot See

AThe catalogue of planets known to orbit other stars now runs to several thousand entries, and it is routinely presented, in textbooks and in press releases alike, as a census of what the Galaxy contains. It is nothing of the kind. A catalogue records what a particular set of instruments happened to be capable of noticing during the years in which they were pointed at the sky, and the two indirect techniques that produced almost all of those entries notice some kinds of planet far more readily than others. The distinction is easy to state and surprisingly hard to keep in mind. What has changed since the first confirmation of a companion to a Sun-like star in 1995 is not the sensitivity of the instruments, which improves steadily and predictably, but the nature of the argument: the hardest question facing the field is no longer whether a faint periodic signal can be measured at all, but whether the star itself, rather than any unseen body, is producing it.

BEvery star of the kind most often surveyed is covered, to a greater or lesser extent, in regions of intense magnetism that are cooler and darker than their surroundings, and as the star turns, those regions sweep across the visible face and distort the light that eventually reaches a spectrograph. The distortion imitates, with unhelpful precision, the rhythmic shift produced by a companion tugging its parent back and forth. A blemish covering barely one per cent of the visible disc can generate an apparent velocity of three metres per second, comfortably larger than the signal expected from a rocky world in a temperate orbit. Marisol Quintero, of the Cerro Blanco Observatory, has spent a decade developing a way of telling the two apart. Her filter ignores the overall shift and examines instead the changing asymmetry of individual spectral lines, a quantity that a genuine orbit leaves untouched. Applied to 84 published candidate signals, the procedure dissolved 19 of them.

CAnton Kreidler, an instrument scientist at the Freiburg Institute for Stellar Physics, is sceptical of what he calls the arithmetic of subtraction. He does not dispute that magnetic activity forges convincing imitations; he contends, rather, that any filter aggressive enough to remove them will discard a comparable number of real detections at the same time, and that the field has no means of counting what it has thrown away. In work published in 2019 he inserted artificial signals of known strength into archived observations and then passed the contaminated data through Quintero's filter, which recovered 62 of the 100 buried planets. The result has been widely quoted. Kreidler concedes, however, that his own trial drew on a single class of unusually placid star and on baselines of no more than three years, whereas the signals now in dispute are precisely those that emerge only after a decade of patient monitoring.

DThe second great technique, which watches for the slight dimming that occurs when a planet crosses in front of its star, suffers from a bias of an entirely different character, and one that no improvement in detector design can remove. A transit is visible only when the plane of the orbit happens to lie almost edge-on to the line of sight, and the geometrical probability of that alignment falls away as the orbit widens: for a world circling at the distance of Mercury the chance is close to one in ten, while for a world at the distance of the Earth it is nearer one in two hundred. The result has been a catalogue crowded with scorched planets on short orbits. Nobody has ever claimed that such worlds are typical. Yet the shape of the catalogue has quietly shaped expectations of what an ordinary planetary system contains.

ECorrecting for these effects is standard practice, and the corrections are large. Astronomers estimate, for each survey, what proportion of planets of a given size and period the pipeline would have retrieved had they been present, and then divide the raw counts by that proportion; the exercise turns a few hundred detections around Sun-like stars into the claim that small worlds are common enough for roughly one to accompany every such star. The procedure has an awkward feature that its practitioners acknowledge readily enough. The recovery fraction is itself measured by inserting synthetic planets into the very pipeline whose blind spots are under investigation, so any systematic weakness shared by the insertion and the search will remain invisible to both. Between 2011 and 2016 three independent groups, working from a single archive, published occurrence rates for temperate rocky planets that differed by a factor of four, and no error was ever demonstrated in any of them.

FWhere does this leave the sceptic and the enthusiast? This article inclines towards Quintero, though with a restriction that she herself insists upon: her filter has been shown to work only on stars that rotate more slowly than about twenty-five days, since faster rotators smear the asymmetries she depends upon beyond recognition. For the remainder of the surveyed population Kreidler's warning stands, and the honest position is that some fraction of the disputed detections are planets and some are blemishes, with nothing at present fixing the ratio between them. That admission is less damaging than it sounds. The one development capable of settling the matter is unglamorous and slow, namely the accumulation of observations spanning several rotation cycles of the star and several orbits of the candidate, because magnetic patterns wander while planetary orbits do not.