Paul Lauterbur died on 27 March 2007, at the age of 77. He will be remembered as one of the inventors of magnetic resonance imaging (MRI), a technique that revolutionized medical diagnosis. Lauterbur's now classic paper (Nature 242, 190–191; 1973) was rejected initially, something that he felt was a mistake, “not because I foresaw all of the medical applications that would follow, but because of the physical uniqueness of the concept”, as he said in his 2003 Nobel Lecture.

Lauterbur's concept of 'image formation by induced local interaction' — which he proposed be known as 'zeugmatography', although the name didn't stick — uses the combination of two magnetic fields to find the location of spins: one field induces an interaction (that is, radio waves drive NMR transitions), whereas the other restricts this interaction to certain spatial regions. In this way, the image resolution is not limited by the wavelength of the radiation involved, as it is in conventional optics, but by the ability to disperse the interaction spatially using the second field.

This is key for MRI of humans. The radiofrequency fields used in a typical clinical MRI scanner have wavelengths of the order of a metre — wavelengths that ordinarily are “not sufficient even for imaging elephants”, as Richard Ernst remarks in his textbook. But in his first demonstration of his zeugmatographic approach, Lauterbur produced resolved images of two water-filled capillaries, each with an inner diameter of only 1 mm (pictured, left). Over the years, the resolution of MRI experiments has been pushed further, down to a couple of micrometres. But this is still a long way from direct imaging of molecular structure — one of the visions of Peter Mansfield, who shared the 2003 Nobel Prize in Physiology or Medicine with Lauterbur.

The nanometre scale does, however, come within reach when MRI is combined with scanning probe microscopy, as John Mamin and colleagues now report (Nature Nanotech. 2, 301–306; 2007). In magnetic resonance force microscopy (MRFM), a sharp tip that creates a localized magnetic field gradient is scanned across an object. In the experiment conducted by Mamin et al. the sample sits on a cantilever that has vibration characteristics that change whenever the sample is brought close to the tip. MRFM has been demonstrated before, but these authors have improved its resolving power considerably, not least by fabricating very sharp and highly magnetized tips, which guarantees both high spatial resolution and a strong signal. This and other improvements have paved the way to a record resolution of 90 nm, or an imaging volume of less than an attolitre (the objects in the coloured images above are each less than 300 nm). Mamin et al. estimate that only a few thousand nuclear spins contributed to the signals they detected.