A fine marriage between two approaches to X-ray microscopy — computed tomography and ptychographic imaging — delivers high-resolution, three-dimensional images of samples without the need for lenses. See Letter p. 436
The 1895 radiograph showing the bones and wedding ring of Wilhelm Röntgen's wife's hand famously demonstrated the penetrating power of X-rays. Along with the very short wavelength of X-rays, this capacity to 'see into' opaque objects has spurred the development of X-ray microscopes to directly resolve features that are too small to be accessible by optical microscopes in samples that are too thick for use in electron microscopes. Practically all X-ray microscopes at synchrotron facilities require exquisite, nanofabricated, diffractive X-ray optics as lenses, and form images through the absorption of X-rays by the constituent materials of the sample. On page 436 of this issue, Dierolf et al.1 neatly sidestep both of these requirements: they obtain quantitative, three-dimensional images of a thick bone sample without the use of optics, while delivering a less damaging dose of X-rays to the sample.
X-ray microscopy has found wide application in the imaging of cells, coal and fossils, interplanetary dust, magnetic materials, polymers, catalysts and many other biological and technologically relevant materials2. The key technology for high-resolution X-ray microscopy is the zone plate. This device bends X-rays by diffraction — rather than refraction — to make a lens. Instead of the careful polishing of glass, a good zone plate requires line structures to be fabricated at precisions of tens of nanometres. In fact, the highest resolution of the lens is given directly by the smallest 'linewidth', or feature size, that can be fabricated. A good optical microscope can resolve features comparable in size to the wavelength of light used to illuminate the sample. However, by this measure, the resolution obtained by zone plates is more than ten times worse, so there is plenty of room for improvement.
A new disruptive technology is set to change X-ray microscopes. Referred to as ptychography3,4, it is a method that builds up an image by means of high-speed detectors and substantial data acquisition and computing power. The images are computed from transmitted X-ray micro-diffraction patterns collected at many positions of an X-ray probe beam as it is scanned across the sample. The real breakthrough in this technique is that the spatial resolution of the computed images is no longer limited by the quality or resolving power of a lens: it is dependent only on the X-ray wavelength and the highest scattering angles recorded in the micro-diffraction patterns. (Dierolf et al. forsake the use of a lens altogether, and form the probe beam with a pinhole.) It is now the sample that bends the beam by scattering, with the smallest sample linewidth giving rise to the largest diffracted angles to be collected by the far-field detector.
This information must be decoded by numerical calculations that act to reverse the propagation of the X-ray wave field that occurred from the object to the far-field detector. Such a computational task is akin to playing a movie backwards to reconstruct a cup from its broken fragments, and leads to an image of the wave field directly at the sample. To achieve this, a problem similar to the 'phase problem' in crystallography must be solved — the detector records the intensity of the light that strikes it, and not the phase (where a wave's peaks and troughs lie). Ptychography determines the phases of the diffracted X-rays by using the huge redundancy of information encoded in the four-dimensional data set; a two-dimensional pattern is recorded for all points of the two transverse dimensions of the probe beam. Because the form of the probe-beam wave field is known, the sample image can be extracted directly, just as in a holographic image reconstruction, which was first demonstrated5 for X-rays in 1996. Dierolf et al. use an iterative algorithm that robustly determines the image.
In contrast to Röntgen's radiograph (Fig. 1), this mode of image formation does not rely solely on X-ray absorption, but directly maps the deviations of the beam wave field as it passes through the sample. In fact, these are represented by the phase of the wave field (not to be confused with the diffraction phases). In the X-ray regime, in which the refractive indices of materials are less than unity, a dense region in the object would cause the phase of the wave field to advance relative to its surroundings. Unlike 'phase contrast' images6, in which such local-beam deviations are turned into intensity changes that can be seen, directly mapping the phase is quantitative and does not sacrifice resolution (an easy way to 'see' a transparent cell in an optical microscope is to defocus, at the expense of resolution). By mapping the phase instead of the absorption, shorter and more penetrating X-ray wavelengths can be chosen, which make the sample more transparent — meaning that there is less energy deposition in the sample. This reduces the radiation damage that causes changes to the very structure under investigation, wiping out details7.
When the resolved features in an image are smaller than the sample thickness, a complete and quantitative understanding of the image requires that it be in the context of three dimensions. In Dierolf and colleagues' study1 of density variations in bone (Fig. 1), three-dimensional (3D) images were assembled by tomography: 2D projected images are acquired from a sample as it is rotated into different views, which are then back-projected into a 3D volume. The resulting images provide precise measurements of mass density that may offer insights into bone growth and diseases such as osteoporosis. These density variations are invisible in the usual radiographic image, and the accuracy of the densities obtained from the images provides a new tool for nanoscience.
Further improvements are certainly expected. Dierolf et al. performed computer-aided tomography in a conventional way by back-projecting images into a volume, similar to a computed tomography scan. Each ptychographic reconstruction was carried out independently without using the fact that they were all views through a common object. It should, however, be possible to exploit this commonality when recovering the images to correct for depth-of-focus artefacts that occur in 2D images, and allow further significant reductions in X-ray dose according to the dose fractionation theorem8. Ptychography requires a coherent beam (such as a laser), and exposure times should drop dramatically with the introduction of new high-brightness X-ray sources such as Petra III in Hamburg, Germany, NSLS-II in New York, and high-repetition sources based on linear particle accelerators. These improvements may soon yield 3D imaging of thick, radiation-tolerant objects approaching one-nanometre resolution.