How can tiny particles such as molecules be weighed? A cantilever, if small and flexible enough, will bend under the weight of a molecule adsorbed onto its surface. A rather more sensitive mass measurement is given by the shift in the cantilever's resonance frequency as mass is adsorbed. This tiny resonance signal can be read out electronically by integrating the cantilever onto a silicon chip.

Previous demonstrations of chip-based mass sensors include the detection of a single virus weighing as little as 9.5 femtograms (1 femtogram is 10−15 grams) in air (A. Gupta et al. Appl. Phys. Lett. 84, 1976–1978; 2004), and weighing a cluster of about 30 xenon atoms — equivalent to a mere 7×10−21 grams — in a vacuum (Y. T. Yang et al. Nano Lett. 6, 583–586; 2006). But where the particles to be weighed are suspended in solution (as is the case, for example, in many biological settings), the technique hits a snag: the viscosity of the fluid damps the resonator, and significantly decreases its sensitivity.

Elsewhere in this issue, Scott Manalis and colleagues present what might be described as a radical solution to the problem: putting the fluid inside the resonator (T. P. Burg et al. Nature 446, 1066–1069; 2007). They have designed a vacuum-sealed silicon microcantilever with hollow channels (pictured), connected to pressure-controlled inlets and outlets for fluid delivery.

The device works in two modes. In the first, which is particularly suited to selective detection of biomolecules, a solution is continuously run through the channels and particles can adsorb on the channels' inner surface — which must be specially prepared for the purpose. The authors demonstrate how the mass change can be followed by monitoring shifts in the resonance frequency of the cantilever in real time as proteins in solution become bound to appropriate receptor molecules that have been grafted onto the tube surface.

In a second mode, particles are detected in transit through the channels; this is useful for weighing particles in dilute solutions. An experiment in this mode determined the distribution in masses, with a resolution of one femtogram, for two types of live bacterial cell, of average masses 110 and 150 femtograms.

The work is an example of the steady progress that is being made in designing practical, inexpensive and portable lab-on-a-chip diagnostic devices. Although further advances are required to demonstrate a 'killer application' for Burg and colleagues' fluidic sensor — a medically relevant specific detection of viruses in blood samples, for example — it is already an elegant method for weighing tiny particles in solution.