"I know what the atom looks like!" Ernest Rutherford's excited announcement at a Sunday evening dinner party almost a century ago stemmed from his remarkable ability to distil the results of a series of pain_staking scattering experiments into an elegant and appealing model of the atom. Despite the revolution in our understanding of the atom brought about by quantum mechanics, Rutherford's iconic model persists. It is the Rutherford–Bohr picture of the atom that non-scientists, and quite a few scientists too, tend to hold in their head, rather than the probability density distributions for s, p, d and other orbitals. In the minds of most, an atom is a solar system writ small.

To see small things, we, of course, use a microscope. Indeed, the word microscope, which has Greek origins, means "to see small". But the traditional optical microscopes many of us have used to bring small structures into focus have a basic problem when it comes to resolving something as small as an atom: the wavelength of a photon of visible light is huge on the atomic scale. Visible light spans from about 400–750 nm and there is a fundamental limit – the diffraction limit – that dictates just how small an object we can resolve using photons of these wavelengths. It turns out that about the best we can do with traditional optical microscopy is to resolve objects about 200 nm across – almost three orders of magnitude larger than the diameter of an atom.

In order to get anywhere close to imaging on atomic length scales, a radically different approach is required, with the most logical step being to reduce the wavelength of the light. We need X-rays to reach nanometre and sub-nanometre resolution, and there have been dramatic and impressive breakthroughs in the resolving power of X-ray microscopes in recent years. For example, last year researchers managed to break the 10 nm resolution barrier for the first time.

But photons are not the only particle available to us, of course. The electron has an even smaller wavelength, and there have been amazing developments in pushing back the resolution frontier for transmission electron microscopes (TEMs). A little more than a year ago, sub-50 pm resolution was reported by a pioneering group at the Lawrence Berkeley National Laboratory in California using a state-of-the-art TEM equipped with aberration-correcting electron optics. What is not widely realized, however, is that the image formed in a TEM does not directly show the positions of atoms. Instead, the data recorded at the TEM imaging plane are patterns due to the interference of electron waves. This "exit-plane wavefunction" can only be converted into an image of the atomic structure via sophisticated computer algorithms and modelling. Whether one could describe the resulting picture of atoms as an image in the conventional sense is a moot point (see box).

But what if we could design a microscope that did not need lenses or other optics to work? Enter, stage left, the scanning probe microscope (SPM). Probe microscopes, which can achieve not only atomic-, but also sub-atomic resolution, come in a variety of different forms. However, all use a concept very different to that underpinning the microscopies discussed so far: an SPM image is generated by measuring the interaction of a sharp tip with a surface – no lenses, no mirrors, no optical elements. We now have a family of scanning-probe techniques that exploits different local interactions between a tip and a surface. This includes the grandaddy of scanning probe techniques – scanning tunnelling microscopy (STM), which is based on the measurement of tip–sample current. Atomic force microscopy (AFM), invented shortly after the introduction of STM, has sired a variety of related techniques that can map a range of forces, including electrostatic, magnetic and frictional. However, it is AFM's capability to image atoms by probing short-range chemical forces that means it is now poised to supersede STM as the technique of choice for atomic-resolution probe microscopy.

Resolving individual atoms and molecules using advanced microscopy is now a regular, some might even say mundane, occurrence in many research laboratories. But Rutherford's statement still warrants careful consideration: just what does an atom or, for that matter, a molecule actually look like? How much detail is it possible to resolve? And to what extent do the particular sample, the type of microscopy and, increasingly, image processing and digital "artistry" define not only the form of an atom in a given image but, more broadly, the public perception of the atomic and molecular world?

Tunnelling ahead

The very first images of atoms were observed one day in the autumn of 1955, when Kanwar Bahadur and his PhD supervisor, Penn State University professor of physics Erwin Müller, strained their eyes in the gloom of a darkened lab and looked at the fluorescent screen of a field ion microscope (FIM). This was the culmination of more than two decades of work for Müller, which had begun with the invention of the field emission microscope (FEM) in the 1930s and the subsequent development of the FIM during his time at the Kaiser Wilhelm Institute in Berlin.

Both types of microscope use an atomically sharp metal tip as the sample. Müller biased his tips to a high, few-kilovolt potential to produce large electric fields, of the order of 1010 V m–1, right at the tip. In the first version of the microscope – the FEM – fields of this magnitude enabled electrons to quantum-mechanically tunnel from the tip into the vacuum, through the potential barrier that usually confines electrons to the metal, in a process known as "field emission". Subsequent acceleration of the field-emitted electrons along the electric field lines onto a fluorescent screen produced a magnified image of the end of the tip. However, it was only when Müller switched to using gas ions, rather than electrons, that atomic resolution was achieved.

Despite the excitement that the FIM images generated, and notwithstanding a great deal of expectation in the FIM/FEM community, Müller, who died in 1977, did not receive a Nobel prize. A Nobel prize was instead awarded in 1986 for work that exploited quantum-mechanical tunnelling in a rather different fashion to produce images of atoms: the invention of the STM by Gerd Binnig and Heinrich Rohrer of IBM Zurich. (They shared the prize with Ernst Ruska for his development of the first electron microscope.)

Scanning revolution

With the invention of the STM in 1981 – more than 25 years after the FIM first resolved atoms – it became possible to see individual atoms and molecules anywhere on an entire solid surface, not just those on a pointed tip. This time the tip is used as a scanning probe that is rastered across a sample surface to map out its 3D topography (figure 1). Figure 2a shows one of the highest resolution STM images to date of a surface that the vast majority of scanning probe microscopists use to check that their STM is performing correctly – the (7 × 7) reconstruction of the silicon(111) surface. The (7 × 7) notation comes from the fact that when a silicon crystal is cut through the (111) plane and then annealed, the atoms on the new surfaces rearrange themselves, to lower their energy, in a pattern made up of tessellating diamonds – the "unit cells" of the surface reconstruction (figure 2b). The two vectors that describe these unit cells are both seven times larger than those that describe the spacing of atoms on the uncut (111) plane – hence (7 × 7). Binnig and Rohrer focused their efforts on imaging this partic_ular surface, which is the prototype surface for scanning-probe studies under ultrahigh-vacuum conditions.

Although FEM, FIM and STM all use a sharp tip from which electrons are emitted via quantum-mechanical tunnelling, the STM differs substantially from its field-emission predecessors in that electrons tunnel not from the tip into the vacuum, but through a tiny vacuum gap between the tip and a sample. The electrons can either travel from the tip to the sample or, with a change in polarity of the voltage, from the sample to the tip. The probability for electrons to tunnel increases exponentially as the gap between the tip and the sample decreases.

The SPM tip can be positioned with sub-angstrom precision above the surface using piezoelectric actuators. These devices are based on piezoelectric crystals that produce a voltage when mechanically stressed – an effect many of us are familiar with as it is exploited to generate the spark in cigarette and gas lighters. Conversely, a piezoelectric crystal will deform when a voltage is applied across it. It is this latter phenomenon that is exploited in scanning probe microscopes. With low-noise voltage sources and high-quality piezoelectric actuators, control of the tip position down to the picometre level is possible.

Seeing atoms?

But what do the intensity maxima in an STM image actually represent? Each peak originates from the tunnel current flowing between the tip and sample, the magnitude of which is determined by the overlap of the electronic wavefunctions of the tip and sample. An STM image is, in essence, a map of the local density of electron states within an energy window defined by the bias voltage applied to the tip or sample. The overlap of tip and sample wavefunctions results in a convolution of tip and surface structure and deconvolving one from the other is generally a far from trivial task.

For the silicon(111)-(7 × 7) surface there is a fortuitous match between the positions of the surface atoms and the peaks in the STM image, largely because the dangling-bond orbitals of the silicon atoms are oriented so that they point directly out of (i.e. normal to) the surface. Nevertheless, a different voltage can produce a distinct change in the contrast of the image (figure 2c) because the energy window available for electron tunnelling is modified. With an STM we therefore do not see atoms as such, i.e. we do not map the nuclear positions, rather, we map out the variation in electron density.

It is in the field of molecular imaging where the most striking high-resolution images of electron-density variations are produced. Buckminsterfullerene, the football-shaped C60 molecule, has been particularly intriguing in this context, with a variety of fascinating STM studies revealing its internal electronic structure. Intramolecular contrast in STM images arises from the spatial distribution of the molecular orbital electron density. (Uniquely, STM is capable of mapping, with sub-nanometre resolution, both the orbitals occupied with electrons and those without.) A particularly impressive example of this is shown in figure 3. Taken from the work of Guillaume Schull and Richard Berndt at Christian Albrechts University in Kiel, Germany, the data show how the intramolecular contrast observed in the STM images varies with the orientation of the molecule on the surface.

Feel the force

We are not constrained to just using tunnel current to probe the tip–sample interaction. Binnig, Quate and Gerber, inspired by a talk by John Pethica at the first international workshop on scanning tunnelling microscopy in 1985, extended STM to the detection of atomic forces, reporting the development of the atomic force microscope in 1986. The most significant recent advances in high-resolution imaging have come from a particular breed of AFM that uses a mode known as "non-contact" (figure 4).

Non-contact AFM (NC-AFM) is an excellent example of the application of first year undergraduate physics – in this case, in the area of vibration and waves – to state-of-the-art science. In NC-AFM, the sharp tip that is common to all forms of scanning probe microscopy is found at the end of a cantilever, which is driven at its resonant frequency. While the control variable for STM is the tunnel current, in NC-AFM the key experimental observable is the frequency shift – the difference between the "free space" resonant frequency for the cantilever when it is far from the sample and the resonant frequency when the tip interacts with the surface. Just as the resonant frequency associated with a mass on a spring shifts when the mass is changed, the frequency shift observed in NC-AFM arises from the force with which the tip interacts with the surface. It is now possible to measure directly, and with sub-atomic precision, the force due to a single chemical bond between two atoms.

NC-AFM research is currently divided into two broad camps. Some research groups use rather "floppy" silicon cantilevers with relatively small spring constants (of the order of tens of newtons per metre) whereas others are increasingly focusing on the application of quartz tuning-fork sensors, just like those used in quartz clocks and watches, for high-resolution imaging.

In the so-called qPlus variant of NC-AFM, pioneered by Franz Giessibl, now at the University of Regensburg, the tip is glued to one tine of a quartz tuning fork to produce a very stiff (of the order of 1800 N m–1) cantilever. As the cantilevers are so stiff, much smaller oscillation amplitudes can be used, down to tens of picometres, which makes the technique more sensitive to the short-range forces of chemical bonding. Silicon cantilevers can also produce excellent results: Oscar Custance and Seizo Morita of Japan's National Institute for Materials Sciences and Osaka University, respectively, have carried out a series of groundbreaking NC-AFM experiments that have involved not only chemical identification at the atomic level, but also the manipulation of individual atoms on semiconductor surfaces.

Last year, in a groundbreaking paper in Science that received wide media attention, Leo Gross and co-workers used the qPlus NC-AFM technique to resolve the structure of pentacene, achieving the highest resolution of a molecule to date (figure 4a). The stunning image reveals the "architecture" of the pentacene molecule, clearly resolving the internal bonds and far outstripping the resolution of the STM topograph also shown in the figure. Two key advances were required in order to achieve such unprecedented resolution. First, rather than just using a plain tip, a single carbon-monoxide molecule was attached to its end, which significantly enhanced the resolution. Second, it was necessary to place the tip very close to the molecule, into the repulsive Pauli-exclusion component of the interaction potential, in order to achieve high-quality intramolecular resolution. Meanwhile, researchers at the Forschungszentrum Jülich and the University of Osnabrück, led by Ruslan Temirov, have shown that high-resolution imaging in the Pauli-repulsion regime is also possible using STM.

Orbital imaging

As highlighted earlier, one must always remember that any scanning-probe image involves, at some level, a convolution of tip and sample structure. In other words, the sharpest thing, ideally the tip, does the probing and the bluntest thing, ideally the surface, is what gets imaged. On the smallest length scales, however, atomic orbitals at the sample surface may well have a smaller spatial extent than those at the tip. This means that an SPM image can incorporate sub-atomic information on the electronic structure, i.e. orbital symmetry and spatial extent, of the probe. Giessibl and co-workers have presented evidence for this type of orbital imaging over the past decade, and while the interpretation of the images has been somewhat controversial, reproducible experiments and high-level theoretical calculations provide support for their claim that orbital information can be extracted from NC-AFM data. (I use the term "orbital imaging" very loosely here – orbitals are, of course, not experimental observables. It is the force distribution associated with an orbital that is resolved in a NC-AFM image.)

With certain tips, Giessibl and his colleagues have acquired images of the silicon(111)-(7 × 7) surface where each atomic feature has internal fine structure (figure 4d). They interpreted the sub-atomic structure as arising from two dangling-bond orbitals on the atom at the end of the tip, either of which can interact with the orbitals of the atom on the surface. Giessibl's group has since imaged features interpreted as arising from 4f, 3d and 3p atomic orbitals of the probe when scanning the silicon(111)-(7 × 7) surface with samarium, cobalt and silicon tips, respectively. They have also acquired sub-atomic-resolution images, observing features with a separation of 77 pm, stemming from the spatial variation of the charge density of a tungsten tip used to image a graphite lattice.

That we can now not only measure interactions as_sociated with a single chemical bond, but also see the force distributions arising from individual atomic and molecular orbitals is a staggering achievement. But where could SPM go from here? Arguably the most exciting, and certainly the most challenging, advance in SPM in the coming years will involve the combination of the spatial-resolution capabilities described here with improvements in temporal resolution. At the moment, atomic-resolution SPM is generally a pain_fully slow process. Nonetheless, there have been encouraging recent developments of high-speed techniques, including, while I was writing this article, a highly impressive report by Sebastian Loth and researchers at the IBM Almaden Research Center in California of the measurement of spin relaxation times using an STM pump-probe technique with a temporal resolution in the nanosecond regime.

But even without the added excitement of high-speed imaging, the visual appeal of SPM images provides a thrilling insight into the complexity of the quantum world that will continue to fascinate scientists and non-scientists alike. Rutherford's solar-system-in-miniature model may yet be superseded in the public imagination by iconic images of atomic orbitals.

Style over substance?

To what extent do scanning-probe images, with features resolved down to sub-atomic and sub-molecular length scales, represent a "true" or accurate picture of reality? Do atoms and molecules really look like they do in scanning-probe images? A workshop in June this year entitled "Ethics and aesthetics in the age of advanced visual engineering", organized by the University of Nottingham's Institute for Science and Society, brought together scientists, sociologists, psychologists and artists to explore questions along these lines. Some at the forefront of critical debate in this area argue that we never really "see" atoms with scanning probe microscopy because it uses an entirely different, i.e. lens-less and photon-less, approach to imaging compared with conventional microscopy (and our eyes!). As such, it is claimed that the technique cannot provide an accurate representation of atoms and molecules. Critics also point to the use of false colours, exaggerated aspect ratios and artificial shading as being particularly troublesome in distorting the public's perception of the atomic world.

I must admit that I find arguments of this type disconcerting and unconvincing. After all, we do not use traditional optical microscopy to image internal organs or tissues in the body, but few of us would claim that the pictures produced in an ultrasound or CT scan are not an accurate representation of what is inside us. Yes, there are perceptual difficulties associated with the convolution of tip and sample states in any scanning probe microscopy (SPM) image, but it is difficult to argue that, for example, the image of pentacene shown in figure 4c is a misrepresentation of the molecule. I am confident that this particular image will soon feature regularly in undergraduate textbooks – the agreement between the non-contact atomic force microscope image and the classical ball-and-stick model of pentacene is remarkable.

The use of image processing to improve the aesthetic appeal of scientific data is, of course, not limited to SPM images. Right at the other end of the scale, the stunning images from the Hubble Space Telescope are enhanced via false-colour palettes and contrast adjustment. Claims that this somehow misinforms the public are misplaced in my opinion. Although it is essential that images and data with minimal, if any, processing are used throughout the peer-review process, these raw data simply may not have the visceral aesthetic appeal to attract a non-scientist. Just because we do not, or cannot, see precisely the same view down an optical microscope, or though a telescope, does not invalidate the scientific content of the image.

More about: Scanning probe microscopy

F J Giessibl 2003 Advances in atomic force microscopy Rev. Mod. Phys. 75 949
F J Giessibl et al. 2000 Sub-atomic features on the Si(111)-(7 × 7) surface observed by atomic force microscopy Science 289 422
L Gross et al. 2009 The chemical structure of a molecule resolved by atomic force microscopy Science 325 1110
M Jacoby 2005 Atomic imaging turns fifty Chem. Eng. News 83 13
S Loth et al. 2010 Measurement of fast electron spin relaxation times with atomic resolution Science 329 1628
C Weiss et al. 2010 Imaging Pauli repulsion in scanning tunneling microscopy Phys. Rev. Lett. 105 086103

•  This article first appeared in the November issue of Physics World.