APPLIED PHYSICS: Assembly and Probing of Spin Chains of Finite Size
achieved IR mode-selective chemistry. The atomic-level mechanism underlying this process is far from clear, however. The authors observe a quadratic dependence of desorption yield on laser intensity. It is tempting to implicate two neighboring H-Si units, each vibrationally excited by a single IR photon. Liu et al. correctly caution, however, that a quadratic dependence on intensity does not necessarily imply a two-photon process. The measured activation energy for thermal desorption of H 2
... sorption of H 2 from H-Si(111) is 2.4 eV (6)-far more than two 0.26-eV IR photons. The measured activation energy of a chemical reaction does not necessarily equal the height of an actual energy barrier. Nevertheless, it is difficult to see how only two photons can produce desorption. Any atomic trajectory that might be launched by IR laser excitation of two neighboring-singly excited H-Si bonds can also be produced by heating-there cannot be any secret passages that are accessed only by IR laser excitation. Therefore, if only two photons worth of energy can produce desorption in the laser experiment, then the same amount of energy can produce desorption thermally. But if so, the activation energy would be far lower than 2.4 eV. It thus appears necessary to invoke many IR photons. Perhaps there is some contribution from vibrational energy pooling. Chang and Ewing have observed this effect in physisorbed layers of CO on NaCl (7) , where dipole-dipole coupling induces many molecules excited by a single photon to pool their energy into a few highly vibrationally excited molecules (8). For the H-Si(111) system, however, this effect is expected to be much less important; the dipoledipole coupling is weaker than for CO, and the vibrational lifetime is much shorter. It appears more likely that the high intensities achievable with the free-electron laser produce multiple (sequential) photon absorption by individual H-Si bonds. Much of this energy must remain localized long enough for two H atoms to combine and surmount the 2.4-eV barrier. Scanning tunneling microscopy studies of the H-Si(100) surface have shown that energetic electrons are far more likely to induce the breaking of an Si-H bond than the breaking of an Si-D bond (9) . This difference is believed to be the reason for the increased lifetime of semiconductor devices upon deuterium substitution (10). Van de Walle and Jackson (11) have proposed that energetic electrons excite both Si-H and Si-D vibrations. However, the Si-D vibrational frequencies more closely match those of the silicon substrate, and energy dissipation should thus occur more rapidly for Si-D, leading to preferential breaking of Si-H bonds. The experiments of Liu et al. may exhibit similar behavior, where energy transferred to Si-D units quickly dissipates, leaving only the Si-H units energized. But whatever the mechanism, Liu et al. have successfully accomplished a longstanding goal: IR mode-selective chemistry in a many-atom system. T he creation, investigation, and manipulation of low-dimensional model systems is of fundamental importance in condensed-matter physics. Moreover, an understanding of the wide variety of electronic and magnetic properties of these models-and their associated phase transitions-may lead to applications in spintronics and other areas of device physics. To an increasing degree, such model systems have been created by researchers in surface science. A well-known example was achieved in 1993 by Eigler and co-workers at IBM Almaden, who used a low-temperature scanning tunneling microscope (STM) to arrange adsorbed atoms into a corral that imposed a circular boundary on the surface-state electrons of the underlying single crystal (1). The resulting quantum interference patterns exactly displayed the solution of the Schrödinger equation, which for that geometry can be given in analytical form. A second but no less exciting example is now reported from the same lab on page 1021 of this issue by Hirjibehedin et al., who have carried out low-temperature STM measurements of manganese atom chains (of up to 10 atoms), assembled by atomic manipulation on copper nitride islands that provide an insulating monolayer between the chains and a copper substrate (2). These chains are model systems representing one-dimensional (1D) Heisenberg antiferromagnets of finite size. A Heisenberg chain is a linear arrangement of spins S mutually coupled by an exchange interaction with energy J. Several order-disorder transitions are predicted for such chains. In the absence of magnetic anisotropy, for example, long-range order is predicted to be lost at any finite tem-perature (3, 4). A model system for a ferromagnetic Heisenberg chain (where all the spins are aligned parallel to each other) has been realized in the form of atomic cobalt chains created by step decoration of vicinal platinum single-crystal surfaces. In this case, anisotropy was found to stabilize small ferromagnetic spin blocks (5). 1D ferrimagnets (where neighboring spins are antiparallel but do not cancel, leaving a net moment) have been realized with molecular magnets (6) and have been found to display the predicted slow magnetization relaxation (7). However, a chain with antiferromagnetic coupling (where neighboring spins are antiparallel) is of fundamental importance in many-body physics, as it is one of the few systems where a nontrivial many-particle ground state is known exactly (8). The quantum mechanical nature of the spins gives rise to the collapse of the Néel state (the arrangement of antiparallel spins) into a single Understanding magnetic ordering at the atomic scale is essential for spintronic technology. A linear chain of manganese atoms has been created for studying one-dimensional systems.