Writing in Physical Review Letters, Paoletti and Lathrop1 describe the breakdown of smooth flow into turbulence in a laboratory investigation. What makes this so noteworthy is that the flow in question is similar to what is generically expected to be present in astrophysical disks2. Moreover, the authors reach precisely the opposite conclusion to what had been considered the state-of-the-art laboratory investigation3, which found no such breakdown into turbulence and implied that, in the absence of an external agent such as a magnetic field, the rotation pattern of astrophysical disks should be stable.

Understanding how a fluid with internal motion makes a transition from smooth (laminar) flow to turbulence remains a stubborn, long-standing theoretical challenge to the hydrodynamics community4. A particularly relevant astrophysical application of this problem is the study of star formation, because stars similar to the Sun are thought to pass the earliest stages of their lives forming at the centres of gaseous 'protostellar' disks5. The question is whether the gas, which orbits in nearly circular motion (much like the planets orbiting the Sun), is a turbulent or laminar fluid. Depending on the answer to this question, the astrophysical consequences may be very different. The efficiency with which a forming star is able to accrete gas from a surrounding disk is much greater if the gas is turbulent, for example. Furthermore, planet formation in the disk may be aided by the gathering of dust grains into transient whirlpool structures6.

The formal problem of the stability of rotating flow was first addressed by Lord Rayleigh in the late nineteenth century7. Rayleigh found that if the rotational velocity of a fluid decreases more rapidly with radius than the reciprocal of the distance from the axis of rotation, such a system is unstable to infinitesimal perturbations. Astrophysical disks, by this criterion, should be stable. But Rayleigh's analysis was restricted to vanishingly small disturbances, and the geometrical shape of the perturbations was in the form of rings with cylindrical symmetry. It is still not known what types of flow that are formally stable by this Rayleigh criterion might still be unstable to more general forms of disturbance; it is known, however, that some types of Rayleigh-stable flow certainly can be destabilized4,8. The issue of interest is whether the rotation of an astrophysical gas disk about a central mass falls into this unstable category.

This problem can be investigated in the laboratory by studying what is known as Couette flow. In a Couette apparatus, water is confined to flow in the space between two coaxial cylinders. There should be no motion along the central axis, only rotational flow about the axis. The cylinders rotate independently of one another, so that small frictional viscous forces near the cylindrical walls will set up a hydrodynamical flow in which the rotational velocity depends on the distance from the rotation axis. By choosing the rotational velocities of the rotating cylinders appropriately, a small section of an astrophysical disk can be mimicked in the laboratory. In such a disk, the flow velocity is inversely proportional to the square root of the distance from the centre, a pattern known as Keplerian flow. The question to be answered is whether Keplerian flow, formally stable by the Rayleigh criterion, actually breaks down into turbulence.

It is this question that Paoletti and Lathrop1 have sought to address. When a Couette flow becomes turbulent, one of the consequences is a greatly enhanced outward flux of angular momentum, which is imparted to the outer cylinder in the form of a torque. In their experiment, the authors measure this torque directly. An earlier investigation9 had claimed to detect this torque, but the new experiment1 was conducted under conditions in which (undesirable) viscous effects were more effectively minimized.

Close on the heels of Paoletti and Lathrop's claim, however, comes a report by Schartman et al.10 on a related experiment. These investigators found no transition to turbulence for Keplerian flow with the same controlled level of viscosity. This null result was first reported3 in 2006, and the most recent paper maintains its original conclusion that there is no evidence of a turbulent breakdown of Keplerian-like laminar flow for very small values of the viscosity.

Although the two experiments1,10 are in principle investigating the same type of rotational flow profile, the measured quantities and the Couette apparatuses themselves are different. Paoletti and Lathrop1 work with a container in which the end caps at the top and bottom of the cylinders rotate as solid disks. The fluid immediately adjacent to an end cap also rotates uniformly and therefore must make a sharp transition, within a narrow region, to join onto the shearing rotation in the bulk of the flow. The transition causes additional stresses to be present throughout the fluid, which in turn drive vertical circulation patterns (technically known as Ekman flow) that would not be present in astrophysical disks. To avoid this unwanted Ekman flow, the authors1 work with long cylinders, so that near the midplane of the apparatus these effects are minimized.

Schartman et al.10 use instead a Couette system with split end caps comprised of two sliding annuli, so that four velocities can be adjusted: those of the two confining cylinders, and the rotation rates of the annuli. The Ekman circulation may thus be directly controlled. The experiment10 is also equipped to measure the internal velocity of the fluid directly. This ability serves to verify that the correct rotation profile has been achieved, as well as to detect the characteristic fluctuations in velocity that accompany a breakdown into turbulence. Paoletti and Lathrop's experiment1, by contrast, is designed to measure only the enhanced torque, not the velocity field itself.

The two approaches have led their respective investigative teams to exactly opposite conclusions. Paoletti and Lathrop argue that their results show that astrophysical disks would be unstable to large-amplitude disturbances (as opposed to the infinitesimal perturbations assumed in Rayleigh's mathematical analysis) and become turbulent. By contrast, Schartman et al. maintain that there are no significant dynamical instabilities in which the Keplerian rotation field of the Couette flow — or, presumably, of an astrophysical disk — is the driving source of energy for the onset of turbulence.

By way of support, Paoletti and Lathrop can point to a recent fluid experiment by van Gils et al.11 that finds the same quantitative relationship between the rotation parameters of the cylinders and the ensuing turbulent torque. It must be noted, however, that in the study by van Gils and colleagues, the unstable profiles are not near the Keplerian regime. On the other hand, the null result of Schartman et al. is itself supported by direct numerical simulations12,13 showing stability of the flows in question. Here, the caveat is that the simulations do not yet have viscous effects controlled at the same level that the laboratory experiments can now achieve.

Because of its central importance to astrophysics, the possibility that disks may be turbulent for purely hydrodynamical reasons will probably excite another round of intense investigative activity, both in the laboratory and on the computer. For the time being, however, we must wait a little longer for a laboratory consensus on whether Keplerian disks are, after all, intrinsically unstable, or whether, as is currently suspected by most accretion-disk theorists, magnetic effects have an essential role in the destabilization process.