Metal Clusters and Magic Numbers

Investigations of tiny lumps of metal can help bridge the gap in physicists understanding of the differences between isolated atoms and bulk solids
by Matthias Brack

H
SLIM FILMS

eat a small piece of metal until it starts to evaporate. Blow the vapor through a slim nozzle into a vacuum chamber. What happens? The spray of particles will condense into tiny clusters containing anywhere from a few to several thousand atoms of the metal. These metal clusters, much smaller than drops

of water or grains of sand, constitute a fascinating new arrangement of matter. Although most cluster experiments involve rather tiny objects, in principle, metal clusters can be arbitrarily large. These aggregates thus fall between isolated atoms or molecules and bulk solids or liquids. In this sense, they link the microcosm with the macrocosm. To

a theoretical physicist, clusters present a host of intriguing questions: Why are some clusters more stable than others? How many atoms are needed before the properties of a cluster begin to resemble those of bulk matter? And how does the structure of a cluster change as more and more atoms join together?

But interest in metal clusters is not entirely restricted to theoretical physicists. For example, industrial chemists know from practice that metal clusters might make exceptionally good catalysts. Yet metal clusters may be difcult to harness more effectively until the principles underlying their formation are well understooda goal that has come increasingly close to being realized over the past decade. The Magic of Stability

fundamental characteristic of metal clusters that scientists must explain is why certain sizes occur preferentially. Chemists grappled with a similar problem more than 100 years ago, when they formulated the periodic table of elements. They found that certain elements had heightened stability because those atoms possessed special numbers of electrons. They called the gaseous elements with 2, 10, 18, 36 and 54 electrons noble gases because, being chemically inert, these atoms did not interact with the hoi polloi sharing space on the periodic table. During this century, physicists discovered that atomic nuclei containing 2, 8, 20, 28, 50, 82 and 126 protons or neutrons are exceptionally stable. Hence, they called these numbers magic. Lead, for example, with its 82 protons, is magically stable. (One form of this element, with 126 neutrons, is said to be doubly magic.) Metal clusters, it turns out, can be magic as well. In 1984 a group of investigators at the University of California at Berkeley examined clusters formed from hot sodium metal. They found that clusters containing 8, 20, 40 or 58 atoms were much more

abundant than other sizes. Clusters with these numbers of atoms predominated because, for some reason, they proved exceptionally stable. Physicists now realize that unstable clusters produced at sufciently high temperatures give off extra atoms and quickly turn into smaller, more stable clusters. Experimenters have found that the magic number sequence for stable clusters of hot sodium continues with 92, 138, 198, 264, 344, 442, 554 and higher numbers. Sodium clusters (and those of several other metals) with these numbers of atoms remain stable even when they are so hot they cannot solidify and remain only as droplets of liquid. The tendency for clusters to form in exactly these sizes arises from the rules of quantum mechanics, which dictate that bound electrons can have only certain energies. In isolated atoms, electrons in excess of the numbers found in noble gas atoms are very loosely held and tend to stray far from the nuclei. These electrons are known as valence electrons and are responsible for various chemical properties of the different elements. In a cluster of metal atoms (or a larger piece of metal), these valence electrons do not remain attached to the particular atoms in which they originally resided. Instead they ow rather freely between the atoms and are said to be delocalized. But they stay in or around the cluster because they are attracted to the now positive charges of the atoms they left behind. (With the valence electrons removed, the positive protons in the nuclei are no longer completely balanced by an equal number of negative electrons, and the formerly neutral atom becomes a positively charged ion.)

PARTICLE BEAM experiments send tiny amounts of metal coursing from a hot source through a long vacuum chamber. Depending on the experimental conditions imposed, the metal clusters created in this way may be small molecules with rigid bonds holding the atoms together, large groupings of atoms packed into regular polyhedral shapes or jellylike droplets without solid internal structure. Particularly stable examples of all three varieties arise in abundance, but not during a single experiment.

To understand why select numbers of atoms yield stability in a cluster, physicists would, ideally, want to determine the detailed conguration of all internal ions and valence electronssomething that is extremely difcult to ascertain. It turns out, however, that they can obtain answers for a metal cluster by modeling it as a smooth jelly of positive charge to which the valence electrons are attracted (a simplication known as the jellium model). According to the rules of quantum mechanics, the energies of these electrons must be quantizedthat is, the energy levels are of set amounts and never occur in intermediate values. And in a metal cluster, just as in an atom, the available energy levels for electrons are not equally spaced. They are grouped into bundles of close-lying levels, separated by larger spacings. For historical reasons, such bunches of energy levels in atoms are called electronic shells, although the electrons are not actually conned to shell-shaped regions. Quantum-mechanical rules limit the number of electrons that can reside in each shell, and if the electrons ll one or more shells in an atom completely, the atom does not react with others and thus proves exceptionally stable. So the existence of magic numbers for metal clusters makes some sense: they correspond to the number of valence electrons that completely ll one or more shells in a cluster and make it especially sturdy. (The same mechanism works for lled proton and neutron energy shells in atomic nuclei and explains their magic numbers.) The metal clusters that show the greatest stability are nearly spherical. If the available electrons cannot ll the highest energy shell to capacity, however, the cluster can become attened or elongated, and it begins to resemble a pancake or a football. Or it may take up a more complicated pear, lemon or diamondlike shape or a conguration with no particular symmetry. Such shapes reduce the total energy of the cluster, making it more stablethough not as stable as those clusters with completely
51

Scientific American December 1997

lled electronic shells. Physicists have long recognized that many stable atomic nuclei also have deformed shapes. So some of the theoretical tools for describing deformed clusters come from previous work in nuclear physics. Supershells

he formalism of quantum mechanics fully accounts for the electronic shells of clusters and their corresponding magic numbers. But to many nonspecialists, this explanation is not very satisfying, because the rules of quantum mechanics often appear counterintuitive and are hard to visualize. Fortunately, it is possible to achieve some understanding of magic numbers using classical formulations that correspond to ones daily experience with macroscopic bodies. This approach requires the periodic orbit theory, which scientists developed in the early 1970s to serve as a bridge between quantum mechanics and classical mechanics. Periodic orbit theory allows physicists to determine, with comparatively little effort, the energies for the major shells in a system containing electrons or other small particles. To visualize how periodic orbit theory applies to clusters, one can imagine that a large metal cluster is nothing but a hollow sphere containing valence electrons that move with constant velocity

along straight lines. These electrons ricochet off the inner surface of the sphere with the angle of incidence equal to the angle of reection, a general property that holds both for particles bouncing off a wall and for light waves being reected by a mirror. This crude model works because the valence electrons are essentially free within the cluster but cannot leave it. The periodic orbits in a spherical cavity are just the trajectories that the particles can take: they shoot back and forth along a diameter or follow around various polygons with three or more corners. Physicists can derive the approximate distribution of allowable energy levels by considering only the periodic orbits that have the shortest length and occur most frequently. The three shortest orbits are diameters, triangles and squares, but for subtle reasons having to do with the geometry, the diameters occur relatively infrequently. Using only the triangles and squares, the calculation yields a pattern of energy shells that is, in a sense, periodic: when plotted as a function of energy, allowable levels appear at regular intervals. But the repetition does not occur at just one frequency. Rather two different frequencies exist, corresponding to the two dominant orbits (triangles and squares). And much as two overlapping sound waves of similar frequency interfere with each other to produce lowerfrequency beats, the two sets of shells

also interfere with each other to produce large-scale clustering of energy levels, termed supershells. Supershells should, in principle, also occur in atomic nuclei. But they do not, because nuclei are never large enough to support them: it takes some 800 to 1,000 particles to give rise to the rst supershell beat, whereas the largest nuclei created so far have fewer than 200 neutrons and 120 protons. But supershells can in fact be found in large metal clusters. In 1991 physicists at the Niels Bohr Institute in Copenhagen and at the Max Planck Institute for Solid State Research in Stuttgart observed supershells for hot sodium clusters. Shortly thereafter, researchers at the Aim Cotton Laboratory in Orsay, France, found evidence for supershells in lithium clusters, and investigators at the University of Lyon-I detected supershells in gallium clusters. The periodic orbit theory for the spherical cavity model predicts that the cube root of a magic number, which corresponds roughly to the radius of the corresponding cluster, should increase by the same amount from one shell to the next. And when researchers plot the cube roots of the magic numbers observed for various hot metal clusters against the shell number, they indeed obtain a straight line with a slope of 0.61. This result agrees with the periodic orbit theory for the spherical cavity (0.603) to within about 1 percent. The fuller

SIMPLIFIED MODEL of metal clusters treats them as hollow spheres with electrons bouncing inside (right). Using periodic orbit theory, physicists can calculate the approximate energy levels of the electrons (horizontal bars). The two most important kinds of periodic trajectories (triangles and squares) interfere to create clusterings of shells, called supershells. The supershell pattern is seen in experiments and in more sophisticated theoretical models (below).

SLIM FILMS

CLUSTER SIZE

LOW STABILITY

HIGH

TRIANGLE

INTERFERENCE

SQUARE

52

Scientific American December 1997

Metal Clusters and Magic Numbers

ELECTRON ENERGY

CLUSTER SIZE (CUBE ROOT OF NUMBER OF ATOMS)

CLUSTER SIZE, expressed as the cube root of the number of atoms present, always grows with an increasing number of shells. But the rise (slope) depends on the temperature of the cluster beam. Cold sodium clusters consist of icosahedral shells of tightly packed atoms, a geometry that gives a characteristic slope of about 1.5. In contrast, the shape of hot sodium clusters is controlled by electronic shells, which give a slope of about 0.6. (The offset seen at shell number 14 is caused by supershell effects and is well explained by the periodic orbit theory.)

25

20

COLD SODIUM CLUSTERS

15

quantum-mechanical calculation using the jellium model, which is feasible only for clusters with up to a few thousand atoms (larger clusters would require an unreasonable amount of computing time), yields exactly the slope observed. Perfect Polyhedra

he ancient Greek philosopher Plato had a vision that the building blocks of matter are regular polyhedra, the so-called Platonic solids. Curiously, in the case of cold metal clusters, he may not have been far wrong. When a large number of metal atoms (about 1,000 or more) aggregate slowly at relatively low temperatures, they form tiny solids in which the atoms pack themselves tightly, like oranges in a grocers pile, to form regular geometric shapes. Physicists can deduce the geometric conguration of these cold metal clusters by observing how their stability depends on their size. In 1991 researchers at the Max Planck Institute for Solid State Research allowed sodium atoms to coalesce at low temperatures and found that the most stable clusters formed in completely different sizes from those that occur at high temperatures. That is, an entirely new set of magic numbers operated. These investigators soon concluded that the newly observed magic numbers corresponded to collections of atoms packed into perfect icosahedra, 20-sided solids with triangular faces. In such experiments, a specic sequence of magic numbers arises as larger and larger icosahedra form through the addition of layers of sodium that are only one atom thick. These atomic shells resemble, in a crude sense, the layered skin of an onion. The simple polyhedral shapes formed in this way are more stable than irregular arrangements because the energy required to hold the atoms in place is reduced when the number of edges is minimized. Nature, it seems, likes the economy of perfect polyhedra. The magic numbers for cold metal clusters, when plotted in terms of their cube roots, increase with the number of atomic shellsbut not at the same rate
Metal Clusters and Magic Numbers

that the magic numbers of hot metal clusters increase with the number of electronic energy shells. And the sharpness of the rise, or slope, changes for various kinds of cold clusters. Different slopes, it seems, signal the presence of different polyhedral shapes. For example, cold clusters of sodium or calcium, which give a slope of about 1.5, most likely form icosahedra. And clusters containing equal numbers of atoms of sodium and iodine, or sodium and chlorine, produce a slope of exactly 1.0, indicating that they form cubes. Why one shape occurs rather than another remains something of a mystery. Cold clusters of aluminum or indium pose an especially interesting puzzle. On plotting the cube roots of their magic numbers as a function of shell number, investigators nd a slope of 0.220, which is less than the value for stable electronic shells (0.6) but different from the slopes for atoms packed into tetrahedra (0.550), octahedra (0.874), icosahedra (1.493) or cubes (1.0). In fact, it is not possible to nd a regular polyhedron of any kind that, when coated with additional layers of atoms, gives the set of magic numbers observed for these clusters. Members of the Stuttgart group have proposed one possible solutionthat such aluminum or indium clusters grow as octahedra but that in going from one magic number to the next, only enough atoms are added to cover one triangular facet at a time. This behavior would lead to the shallow slope that is seen. Their explanation seems plausible, but it immediately raises a new, as yet unanswered question: Why are complete octahedra not much more stable than those with only one or two triangular facets added? Interestingly, the magic numbers arising from atomic packing do not uniquely determine the shape of the polyhedra involved. For example, the magic numbers associated with icosahedra coincide with magic numbers that would be found for cubo-octahedra (cubes with their corners cut off). Careful study of cold calcium clusters has suggested to scientists that it is the icosahedra that

10 HOT SODIUM CLUSTERS 5

10 15 SHELL NUMBER

20

form. Yet we cannot be surenobody has seen the icosahedra directly. In principle, the largest clusters could be viewed with an electron microscope if they were at rest. The problem is that free clusters are formed in beams of swiftly moving atoms, and they cannot be stopped for a picture without altering their shapes. Competing Masters

n metal clusters composed of a large number of atoms, the two kinds of shells (electronic energy shells and atomic shells forming perfect polyhedra) compete for control over form and stability of the cluster. Size and temperature are critical in determining which type of shell prevails. But to measure the temperature of isolated clusters traveling in a beam is quite difcult; one might even ask whether a temperature can be dened at all. Strictly speaking, temperature is meaningful only for systems of many particles in thermodynamic equilibrium. But this condition is not realized in most cluster experiments. Rather each cluster formed in a beam follows its own path, unaffected by its neighbors. Such isolated clusters do not attain thermodynamic equilibrium. So, short of sticking a thermometer against a tiny cluster under studyan experimental impossibilitythere is no easy way to determine its temperature. Still, the trends are clear. By heating the nozzle of the source under carefully controlled conditions, the shells of neatly packed icosahedra in cold sodium clusters seem to disappear. This transition presumably corresponds to the
53

Scientific American December 1997

ARTHUR CHRISTIE

melting of solid clusters. The melting temperature in this case depends on the size of the cluster. Researchers have found a similar transition in clusters made of aluminum. When the temperature of the source exceeds 500 kelvins, which is still far below the melting temperature of bulk aluminum, plots of the cube root of magically stable cluster sizes versus shell number cease to exhibit the

slope of 0.220 (the number that suggests the packing of atoms on the facets of perfect octahedra). Instead they show a slope of about 0.6, which is typical of electronic shells. It seems reasonable to surmise that these clusters have become molten, even though the temperature of the nozzle is well below the melting point of the bulk metal. But that conclusion may be overly simplistic: as with temperature, the notion of a

phase transition between solid and liquid itself becomes problematic for objects of this size. Perhaps the clusters are experiencing a local melting of their surfaces that is sufcient to destroy the structure controlled by the packing of ions, allowing the stability conferred by completed electronic shells to reign. Physicists at the University of Freiburg have recently observed a somewhat different but related transition. To

Packing for the Cold

cientists can explain the exceptional stability of cold metal clusters with geometry. For example, cold clusters of sodium prove especially stable when the atoms pack neatly into an icosahedron, a 20-sided solid. Hence, charts of stability as a function of cluster size show distinct peaks when the number of atoms available is such that they form perfect icosahedra (left). Cold clusters of aluminum atoms, which pack into tiny octahe-

dra, prove somewhat more difficult to explain. Stability is high when the atoms pack into perfect octahedra (right), but maxima also occur at intermediate sizes. The reason may be that enhanced stability also arises when atoms cover individual faces of an octahedron (red layers, below). Because adding a fourth additional face completes an octahedron of the next largest size, only M.B. three intermediate stability maxima are expected.

INTERMEDIATE MAXIMA HIGH HIGH

STABILITY

LOW 2,000 2,500 NUMBER OF ATOMS 3,000

STABILITY LOW 1,800 1,900 2,000 2,100 NUMBER OF ATOMS 2,200 2,300

891-ATOM OCTAHEDRON

ONE FACE ADDED

TWO FACES ADDED

THREE FACES ADDED

1,156-ATOM OCTAHEDRON

54

Scientific American December 1997

Metal Clusters and Magic Numbers

SLIM FILMS; SOURCE: T. P. MARTIN Max Planck Institute for Solid State Research

8 7 5 Y ENC EQU Z) R ER F4 ERT LAS 101 H (x 6

35

100

120

142

164

188

209

229

249

269

380

TEMPERATURE (KELVINS)

LASER PHOTOABSORPTION SPECTRA reveal that clusters are transformed from rigid molecules to structureless masses as temperature rises. The probability of photon absorption varies markedly with laser frequency, a characteristic of molecules, for

clusters of sodium held at low temperatures (blue shades). But at high temperatures (orange shades), the spectra smooth to show just two broad humps, consistent with the so-called jellium model that physicists use for clusters lacking internal structure.

control the temperature of small sodium clusters in a better way, they used a clever trick: they embedded the clusters in helium gas, for which they could adjust the temperature quite accurately. In this mixture of helium atoms and sodium clusters, many collisions take place. But because helium is a noble gas (that is, it remains chemically inactive), it does not change the structure of the clusters at least at moderate temperatures. So the helium provides an environment in which a thermodynamic equilibrium can establish itself. What the scientists examined in this case were not magic numbers but the way in which sodium clusters of one xed size absorb the light of a laser beam. The probability for light absorption as a function of laser frequency (the absorption spectrum) can reveal much about the physical condition of the particles illuminated. In particular, the spectra for cold, rigid molecules and hot, liquid droplets differ. For temperatures below 100 kelvins, the experimenters observed a spectrum with many sharp peaks, which is typical of a rigid mole-

cule, whereas for temperatures above 380 kelvins, they found a spectrum with just two broad humps, positioned as one would predict theoretically for an unstructured jelly. For the smallest clusters, where the internal structure is always important, rigorous quantum-mechanical calculations that take into account all electrons in each atom are possible. They reveal the detailed molecular structure of the clusters and can give quite precise descriptions of their stability. For larger aggregates with some hundred atoms or more, such calculations take up too much computing time to carry out, and physicists have to invoke the jellium model. For clusters with many thousands of atoms, researchers must abandon quantum mechanics and invoke the simplied periodic orbit theory or rely on indirect information from the magic numbers found experimentally. Scientists have come a long way since 1984 in understanding and extending magic number sequences in metal clusters. Some experiments have produced icosahedral sodium clusters made up of

as many as 21,000 atoms. Yet it is clear that sodium in bulk does not contain such icosahedra. In fact, no bulk material can be built up by packing icosahedra together: just as pentagonal tiles cannot be made to cover a plane, icosahedra lack the symmetry needed to form extended crystals in three dimensions by lling out the space completely. So even the most massive sodium clusters produced so far have their atoms organized quite differently from the way they occur in bulk sodium. Part of the initial motivation for studying clusters was to determine how many atoms are required for a tiny lump of material to attain the properties of the bulk solid. Physicists have learned a great deal about metal clusters and the magic number sizes that bring stability to them. But we still do not know the answer to this fundamental question about when and how the transition to the bulk solid occurs. All we can say for now is that the clusters of metal atoms we have studied constitute a weirdly unique arrangement of matter, one that surely offers more surprises to come. SA

The Author
MATTHIAS BRACK received his education in physics at the Niels Bohr Institute in Copenhagen and at the University of Basel, where he received a Ph.D. in 1972. He has done research on nuclear physics at the Niels Bohr Institute, at the State University of New York at Stony Brook and at the Laue Langevin Institute in Grenoble. Since 1978 he has been a professor of theoretical physics at the University of Regensburg in Germany. Metal Clusters and Magic Numbers

Further Reading
Semiclassical Physics. Matthias Brack and Rajat K. Bhaduri. Addison-Wesley, 1977. Clusters, Condensed Matter in Embryonic Form. Sven Bjrnholm in Contemporary Physics, Vol. 31, No. 5, pages 309324; September 1990. The Physics of Metal Clusters. M. L. Cohen and W. D. Knight in Physics Today, Vol. 43, No. 12, pages 4250; December 1990. The Physics of Simple Metal Clusters: Experimental Aspects and Simple Models, by Walt A. de Heer, pages 611676; and Self-Consistent Jellium Model and Semiclassical Approaches, by Matthias Brack, pages 677732. In Reviews of Modern Physics, Vol. 65, No. 3, Part 1; July 1993. Shells of Atoms. T. P. Martin in Physics Reports, Vol. 273, No. 4, pages 199242; August 1996.

Scientific American December 1997

55

ARTHUR CHRISTIE; SOURCE: MARTIN SCHMIDT AND HELLMUT HABERLAND University of Freiburg

PHOTOABSORPTION LOW HIGH