Simon et al.

Vol. 6, No. 5/May 1989/J. Opt. Soc. Am. B

1035

Material applications of the far-infrared free-electron laser
John D. Simon, John E. Crowell, and John H. Weare
Department of Chemistry, University of California, San Diego, La Jolla, California 92093

David R. Miller Department of Applied and Mechanical Engineering, University of California,San Diego,La Jolla, California
92093 Received November 3, 1988; accepted February 14, 1989

Our research program using the free-electron laser (FEL) to study structural and dynamical properties of materials is described. Initial experiments using the FEL facility at the University of California, Santa Barbara, focus on
understanding (1) low-energy processes in bulk and composite materials, (2) the intermolecular forces in van der Waal clusters, and (3) molecular interactions with surfaces. To support the interpretation of our experimental program we have developed the theory necessary to compute the dynamics of multicomponent systems that can interact through weak van der Waal forces or through formation of covalent and metallic bonds.

INTRODUCTION Understanding the chemistry of materials requires a detailed knowledge of the intermolecular interactions between the constituent atoms and molecules. These intermolecular (interatomic) forces determine the structure of the materials as well as the kinetics and mechanism of their chemical reactions. In order to understand these processes, one needs experimental probes that can examine intermolecular
properties of materials. Infrared (IR) light is a powerful

the far-IR spectral region.

However, there are several limi-

tool for studying the structural and dynamical properties of matter, as this spectral region encompasses the vibrational transitions of the aggregate system on the ground-state potential energy surface.' IR-light wavelengths are traditionally divided into three regions. The portion closest to the
visible region is called the near IR, X = 0.8-2.5 Aim,the mid IR extends from 2.5 to 20 ,um, and the far IR ranges from 20 to 1000 Am. The majority of commercial IR spectrometers

tations characteristic of these molecular-gas far-IR lasers. Poor beam structure (mode and pulse shape), large shot-toshot instability in power, and low repetition rate hinder their application to study properties of matter. Furthermore, these laser sources are not continuously tunable, providing radiation only on rovibrational transitions of the lasing medium. In addition to advances in providing narrow-bandwidth IR light for high-resolution spectral studies, there has been significant progress in the generation of time-resolved IR light. A variety of techniques have been explored, but the
majority focus on downconversion techniques that use a

operate in the mid IR. The potential uses of IR radiation for the study of chemical and physicial properties of molecules and materials has prompted an effort in the development of high-power light throughout the IR spectral region. Several types of IR-laser sources have been developed in the past
decade. All these sources have significant limitations; how-

ever, they offer a variety of both high spectral resolution as
well as short-time resolution (from microsecond to subpico-

variety of nonlinear-optical crystals. These frequency-mixing techniques generally involve using the fundamental of a Nd+3 :YAG laser and the output of a picosecond or subpicosecond dye laser to produce IR pulses over the spectral range 4 The wavelength of the IR light is from 4 to -11 ,um. determined by the frequency of the visible dye laser; thus, unlike molecular-gas lasers, this type of IR source is continuously tunable. The major limitation of the tuning range accessible results from the phase-matching conditions and optical absorption properties of the nonlinear optics. As exemplified in later sections of this paper, there are several important properties of molecules and matter whose
study requires light in the wavelength range from 10 to 1000 Am. As mentioned above, line-tunable far-IR gas lasers

second). The combination of these light sources now permit one to carry out experiments that examine both structural (high-resolution spectroscopy) and dynamical (time-resolved studies) properties of molecules and materials.
In our program we are interested in chemical processes in the mid- and far-IR spectral region (X = 2.5-1000 gim).

Both time-resolved and narrow spectral sources are required. A variety of high-resolution infrared lasers are cur-

rently available that cover a limited part of this wavelength region. For example, F-center lasers provide tunable, narrow-bandwidth radiation over the spectral region from 0.8 to
3.4 gm.2 CO2 lasers and far-IR gas lasers 3 (i.e., an optically

pumped CH3F laser) can provide line-tunable radiation over
0740-3224/89/051035-10$02.00

exist for this region of the spectrum. If the system of interest is accessible by one of the narrow bands generated by molecular-gas lasers, experiments can be carried out. However, as these lasers cannot be tuned, it is impossible to obtain a spectrum in this region, significantly limiting the information that can be obtained. Furthermore, ultrashortpulse generation is not possible, as the temporal properties of the molecular-gas lasers are determined by the pulse width of the pump CO2 laser. One desires optical sources of good stability (both frequency and time), mode structure, and tunability. Significant advances in this direction have been made using free© 1989 Optical Society of America

b There are several lines for the CH3F laser in the wavelengthregion covered electron lasers5 (FEL's). This system is currently equipped with Auger electron spectroscopy. depending on active rovibrational transitions in the lasing medium. and adequate vacu- .5-50 usecc 1 Hz 1 :40 nsecd <1 Hz >1 by the FEL. Most of the instrumentation utilized to study the interaction of light generated by a FEL with condensed matter is somewhat conventional and not unique to use with FEL's. must be employed. The tunability of a FEL is especially important in some of the areas that we discuss below. ion beam cleaning. Gases are introduced into the system by using a separate turbomolecular-pumped gas handling system at <1 X 10-8 Torr. interfacing of the experimental vacuum system to the FEL light source is straightforward. The multilevel. librations. offer to be powerful light sources for experiments on the structure and dynamics of matter. In addition. Most of these have at one time or another been probed with conventional light sources. enables one to record optical spectra of the materials of interest. Other studies will utilize different configurations. Opt. Santa Barbara (UCSB) and the Los Alamos National Laboratory. Table 1.um 10-8 CH3Fb S250 kW 359. This property affects the focusing of the laser beam and is also important for the carrying out of a variety of nonlinearoptical experiments of interest. in which many-bodied interactions shift and broaden the spectra. The system is maintained at <1 X 10-10Torr. and thermal desorption mass spectrometry. As mentioned above. First. As shown in this table. 1) for our ongoing studies involving excitation of surface states (as described below). hindered rotations. Second.01 eV or -100 K). 5/May 1989 Simon et al. Am. As mentioned above. Furthermore. the FEL permits a much wider tunability and often much higher photon fluxes for these studies. we discuss three types of experiment that we are currently pursuing using the UCSB FEL: phonon-adsorbate dynamics following IR excitation of phonons and/or librational molecular modes of the adsorbate. below 100cm-' the absorption by water vapor is serious. the frequency bandwidth is extremely narrow. For a source such as the UCSB FEL. helium-atom scattering.g. we discuss examples of systems that are currently feasible for study at the UCSB FEL. Fraction frequency Band width (single pulse) Pulse length Repetition rate Number of modes 0. aData from Ref. Specifically. the possible uses of this light for examining properties of molecular clusters and molecular interactions with surfaces are discussed. and thin-film preparation and examination. reflecting pump laser.3 Smooth pulse shape. magnons. and vibrations) involve transitions in the IR. including internal and external IR reflection studies. In the region of the far IR and mid IR. it is obvious that molecular systems under investigation should be rotationally and vibrationally cold. which currently operates below 80 cm-' (0. capable of signal averaging. For molecular-beam or surface experiments. B/Vol. Soc. available through Port Plastics) or single-crystalline quartz can be used with good transmission.8 Owing to the low repetition rate common to many FEL's. the FEL light is single mode. the output characteristics of the UCSB laser system are compared in Table 1 with those observed for an intense line of a molecular CHF laser. TPX is easily machined. opens up the possibility for many nonlinear multiphoton experiments. using a titanium sublimation pump and a 360-L/sec turbomolecular pump backed by a diffusion pump. either the plastic poly-4-methylpent-1-ene (TPX. a boxcar integrator or related gated detector. and light pipes must either be purged with dry nitrogen or evacuated. but special considerations are sometimes required. and spectroscopy and bonding of composite clusters formed in molecular beams. No.7 5 X 10-3 In the followingsections of this paper. Many condensed-matter phenomena (e. respectively. FEL's have been demonstrated to work in various regions of the optical spectrum. the molecular-beam species must be properly chosen to match the energy range of the light for molecular-beam sources by source. Comparison of the Output Characteristics of the UCSB FEL with Those of a CH3F Molecular Gas Laser Property Peak power Wavelength FEL (UCSB)a 10-40 kW 12 0-800. Expansion ports are present for the addition of further surface diagnostic probes. this can be accomplished using free-jet expansions7 and for solid sample materials by utilizing low-temperature cold stages for sample manipulation. Data presented are for pumping the moleculargas laser with the 9(R)12line of a TEA CO2 laser. EXPERIMENTAL STUDIES IN THE FAR INFRARED After a brief description of experimental considerations and instrumentation that we have developed for surface and cluster studies. the molecular-gas laser is only line tunable. This. The latter feature. As an example of the advantages of FEL light sources. rotational energy exchange in sticking/desorption at the gas-surface interface. ultrahigh vacuum and proper surface characterization probes are a necessity in studying surface physics. d Irregular spiked pulse shape. Currently all these techniques are present on a single level (as shown in Fig.1036 J.. the FEL laser offers several significant advantages over the conventional molecular-gas laser. the FEL's at the University of California. however. multiport design of the surface facility will permit a variety of experiments to be performed. rather easily. Similar considerations must be taken into account when excitation of the surface substrate is desired. Experimental Considerationsand Instrumentation Figure 1 is a schematic diagram of the ultrahigh-vacuum surface science chamber utilized for experiments at the UCSB FEL. 6. such as the spectroscopy of solid surfaces and clusters. near 60 cm-'. coupled with the continuous tunability of the laser (not possible with the molecular-gas laser) in the far IR. For example. the window materials that permit the light to pass into the vacuum chamber depend on both the wavelength range and the desired vacuum in the experimental chamber. high intensity. Because of the low duty cycle of the FEL and because many future experiments will be operated at cryogenic temperatures. 6. surface scattering with gas-phase excitation. phonons. For example.

um seals can be made with compression 0-ring seals. 1. helium-atom scattering.g. We have used Figure 2 is a schematic of the experimental apparatus that we have constructed for cluster studies using the UCSB FEL. B Quadrupole Mass 1037 _ Spectrometer Electron ITY Metal Deposition Source Level I Auger Electron Spectrometer X.YZ. the laser it is desirable. In this case singlecrystalline quartz (which transmits between 30 and 250 cm-') is used. For these experiments we have found that TPX is the window material of choice. Schematic diagram of the experimental ultrahigh-vacuum configuration used to study the excitation of surface modes or those of gases incident upon a surface using FIR radiation. Vol. though not necessary. to utilize a pulsed-nozzle free-jet beam source. 5/May 1989/J. 6. this material cannot be baked and is therefore not suitable for the ultrahigh-vacuum conditions required for high-purity surface characterization. The crystal surface can be cleaned by ion beam sputtering and characterized by Auger electron spectroscopy. 2 depicts one such design (by Laser Technics) that uses a piezoelectric crystal to pull a plunger and permit gas flow. The quartz windows are vacuum sealed with differentially pumped viton 0-ring seals. Opt. and the upper half of Fig.0 Crystal Manipulator Sputter \Ion Gun FEL Radiation (Entrance) /UDetector Level I Shielded Quadrupole Mass Spectrometer Helium Beam Source Level II Level I Level III 360 Liter/sec Turbomolecular Pump Titanium Sublimation Pump Optical Ports Level III Fig. After the chamber is baked at 1500C. No. Soc. Am.Simon et al. Several are commercially available.this system maintains a base pressure of 10-10Torr. Because of the pulsed nature of such a design in the system constructed for studies at the UCSB FEL user center.. A solenoid valve (e. How- ever. and temperature-programmed desorption. those available through General Valve Corporation) will do .

Gough et al. For example." pumped surface plasmons in sodium clusters to stimulate desorption. 5/May 1989 Simonet al. This is a function of the AA. the photon absorption can also be followed by reflection attenuation measurements. equally well.13 used the Stanford FEL at 3. High source pressures are desirable to achieve a considerable density of large clusters or to cool internal molecular rotation states. They were able to distinguish a greater desorption yield on excitation of . Fig. Hoheisel et al. Most recently Tro et al. Am. and several examples of the type of experiment that we are discussing are available.la and many experiments will be enhanced by the tunability and intensity of the FEL. most of which use metal surfaces and irradiation by UV light from intense laser sources. B/Vol.9 have been able to adjust the source such that the chromophore is initially either inside the cluster (by mixing the gases in the nozzle) or attached to the surface of the cluster (by using a pure solvent expansion into a background of solute molecules). the solute chromophore will diffuse within the cluster and how long it takes for such an adjustment. AB. Soc. two of the interesting questions are whether. of the figure. 6. The energy deposited in the chromophore may be exchanged with the host and subsequently sublime and/ or melt and evaporate a fragment. Schematic of the experimental setup for cluster beam photon-interaction studies. Also indicated in the figure is a composite cluster made of a solvent speciesB and a solute chromophore A. and BB interaction potentials and the dynamics of the cluster. The advantage of the pulsed valve is that highpressure sources can be used with small diffusion-pumped vacuum chambers. In both cases. We are particularly interested in surface modes and the dynamical interaction of surface modes with adsorbates. The interaction of light with surfaces and adsorbates has been a subject of intense research during the past decade. respectively. Such studies in the bulk have a substantial history covering the past 30 years.3 m to examine the dynamics of adsorbed C4H1a. No. as suggested by the experimental schematic shown in Fig. and indirect excitation through excitation of a surface phonon. such as phonons and magnons. in fact. Excitation of Surface States The far-IR light at the UCSB FEL is probably most well suited for studies involving condensed-matter excitons. 2. 1 above. Budde and co-workers' 2 have pumped electronic states of NO and then characterized the translational and internal states of the desorbed species for dynamical information. followed by desorption into a mass spectrometer detector. which absorbs the photon. in the bulk and at interfaces.1038 J. They were able to resolve both thermal and nonthermal channels for desorption. Of course. followed by dynamical coupling to an adsorbate and subsequent desorption and detection by a mass spectrometer. A detailed drawing of the pulsed valve and the interaction of a composite cluster with the laser radiation are shown in the upper and lower halves. The size distribution of clusters is qualitatively controlled by the source pressure and temperature but is rarely better than 10% FWHM. Opt. There are two types of experiment that are of interest: direct excitation of an adsorbate mode.

which is raised out of state on the surface and that the surface plays a role.' who used electron-energy-loss spectroscopy to probe hydrogen physisorbed on metals. SP .2 X 1013sec'1). It has a transverse-optical vibration pattern. has not been confirmed experimentally. Soc. However. as we have found. Rotationally inelastic transitions are labeled J J. The grating that we have put into a CsI(100) crystal surface is spaced at 2-mm intervals. In this case. and would be a sensitive probe of the lattice dynamics calculation and force model. although the UCSB FEL is not currently extended to this energy. 5/May 1989/J. Vol. an arrangement is possible to permit excitation of optically coupled phonons (polaritons) or multiple acoustic phonons. using both reflection intensity and desorption as monitors of the phonon coupling. While coupling to acoustic phonons such as the surface localized Rayleigh phonon does not involve a polariton. respectively. When one is considering the UCSB FEL. These surfaces. an adsorbate such as HCl and D20 might be examined with the UCSB FEL. there should be a large apparent enhance8 ment of Si expected at both the X and M points. it does require a two-phonon excitation in order to conserve momentum. It is possible to select a frequency where the substrate will not absorb strongly.' The former that uses a prism to couple to surface polariton experiment. Sl' is an optical (or second) Rayleigh mode. At low temperatures. The sum of S. coupling of the light to the surface must be done to the resulting surface polariton. near 115 cm-' (2. B 1039 the asymmetric C-H stretch as compared with the symmetric C-H stretch. It is observed that hydrogen is in an unhindered rotational effects owingto the large polarization of the lattice atoms. a polariton is not excited. respectively. It is an obvious candidate for the the FEL. in the ortho-para spin conversion. adsorbates can be weakly physi- the surface polariton is not suitable for examining the surface and the interaction of adsorbates under ultrahigh vacuum. good candidates for substrates include the alkali halides.18 which are influenced by substantial surface relaxation sorbed to the substrate sothat such excitation by the electric field or by phonon coupling may desorb them into the mass spectrometer for detection. Therefore we are using a grating on the surface for such optical phonon experiments. This sagittal plane (SP) mode vibrates perpendicular to the surface. such an arrangement 6 is the basis of the attenuated total reflection method. 3 are the surface modes 8. Of further interest in Fig.Simon et al.e. while X and Mare the high-symmetry points at the zone edges along the [110] and [100] directions. 19. Since the optical phonon dispersion. 6. and maintained in the ultrahigh vacuum rather easily. c(k). so that direct excitation of the adsorbate can be examined together with the effect of relaxation from the adsorbate to the lattice. and Sl' of RbBr. permitting coupling to the FEL light along the [110] direction at -67 cm'. attributable to orientation affects in the adsorbed butane layer. cleaned. i. Related investigations are those of Avouris et al. not yet understood. For mass spectrometric detection it is desirable to use a molecule that has a small background concentration in the UHV chamber. Opt. The most pronounced relaxation effect is the separation of a new surface optical mode. recently calculated by de Wette et al. r is the surface Brillouin zone center. which then couple dy- RbBr relaxed U) Ln V 3 Fig. Conversely. Adapted from Ref. Angular distribution - of the reflected number density for HD scattering from Ag(111) along the (112) direction. 3. No. w = ck. consider the RbBr surface. The vibrational excitation leads to resonant heating and thermal desorption of the butane ad4 layer. Am.' Therefore a two-phonon (TA + TO) experiment should be possible with adsorbates.' 6"7 This requires that the incident light wave vector be altered by using a prism or a grating. .. can be prepared. and Sj' at X and M is 78 and 66 cm-. crosses the photon dispersion curve. and since its relative motion cannot be distinguished from the Rayleigh acoustic mode S1. such as CsI and RbBr. Figure 3 shows the phonon dispersion curves. which have well-defined optical and acoustical phonons at the surface in the frequency range accessible with the transverse optical band when the surface is allowed to relax. As another example of a system that we are currently examining.' 5 and Andersson and Harris.

Analysis of such data provides insight regarding the gas-surface potential and the collision dynamics. the translational energy with which the species desorbs. for example. it is possible to examine rotational excitation in the beam. which do not involve exchange of energy with the surface. Soc. Rotational temperatures can be varied between 5 and 30 K. into high vacuum. As an example. and the rotationally mediated selective adsorption process as a de- Both nozzle source temperature and pressure are used to adjust the beam properties. Molecular beams provide a means of probing such systems in a free state. 6. The FEL tector. by time-of-flight analysis. and involves the J = 0 initial state. has not been examined. and the effect of rotational excitation on the gas-surface interaction.. especially NO. providing additional dynamical information. if HF (with rotational constant B = 21 cm-') were cooled rotationally to 5 K. we intend to pump energy into composite clusters and then to examine the heating and fragmentation of the small clusters. Here the prototype experi- Experimental Studies As discussed further below. then essentially all molecules would be in the J = 0 state Nj=l/Nj=o = (2 X 10-5). B/Vol. the mass spectrometer indicates not only what desorbs but. Such an experiment. and Elliot and Miller24 have investigated the sticking coefficient of CO on Au(111). provided that the states of the solid surface with which the molecules interact represent an equilibrium distribution of states. for example. Recently a great deal of progress has been made in using laser spectroscopies to examine the rotational states of desorbing species. although its sticking probability on metals is not yet known.1040 J. Two studies that are closely related to our interests in the far IR are those of Gough et al. has been discussed theoretically by Reyes et al. Using far-IR FEL radiation. and they are smooth surfaces so that the AG = 0 channels are favored. Therefore such studies are affected by the relaxation rates of the lattice and the dynamics of the interaction. while at 30 K the ratios are approximately Nj=l/Nj=O = 0. In the first geometry the photons cross the beam in the nozzle chamber and may make multiple passes with properly arranged reflecting surfaces. Figure 2 indicates two geometries for studying the interaction of far-IR photons with the molecular beam. Rotational to translational energy exchange opens up well-defined inelastic channels. 1 This is accomplished. photons. and those of Vernon et al.01. especially the attractive portion. The bound states in the attractive potential well can significantly alter the intensity of elastic. During the past decade many studies involving lasers using pulse-probe techniques have been applied to condensed phase systems to investigate energy transfer.9 who examined SF6/Ar composite clusters. using polariton-induced desorption. Again.2 6 who studied small water clusters. Molecular systems that can be probed include HCl and HF as chromophores in Ar or D20 clusters. Opt. and the velocity distribution translational temperature can be varied from well below 1 to 10 K. including multiphoton processes. using a mass spectrometer. a more direct measure is to use helium-atom scattering as a monitor of surface coverage. The reverse process. One such candidate. The two processes would be coupled by detailed balancing. Rotational Exchange with Surfaces Molecules can be prepared in low-rotational-state popula- tion in a high-pressure free-jet molecular-beam expan72 sion. In this case the HD molecules were cooled to the J = 0 state in a free-jet expansion and the rotational-translational transitions followed by measuring scattered intensity as a function of angle from the surface normal. Detec- . It is seen that most of the scattering is not coupled to a diffraction channel. 9 ments are those of Whaley et al. While sticking probabilities can be studied by mass spectrometric analysis of scattered molecules. . AG = 0. there is considerable interest in understanding the dynamics of small condensed systems or clusters. by mixing the molecular species of interest into a carrier gas such as helium and then expanding the mixture from high pressure. In particular. This probe is especially sensitive for surface coverages below 10%. we will be interested in the effects of rotational states in the sticking and scattering from welldefined single-crystal surfaces. The carrier gas serves to prevent condensation or clustering but provides the rapidly cooled bath of background atoms that cools the molecular species. The molecular species must be light enough to exhibit quantummechanical scattering and to scatter preferentially without interactions with phonons at the surface. Naturally the lattice absorption can be followed by reflection intensity. Ag system. and provides strong resonant features in the scattered differen- namically to the adsorbate and induce desorption.20 Candidates for the UCSB FEL would be physisorbed krypton or HD. 22 One of the interesting results is the underpopulation of high rotational states in the desorbing species compared with the lowJ states. from smooth metals. No.4 and Nj= 2 /Nj=o = 0. or solid surfaces. Both Ag(111) and Au(111) are good candidates for the experiment because they can be maintained clean in ultrahigh vacuum. which are greatly influenced by the nature of the gas-surface interaction potential. as compared with examining clusters on surfaces where the interaction of the substrate can be significant. Adsorbates attenuate the intensity of coherent helium-atom 23 scattering.. The beams and associated fragments are then monitored in the mass spectrometer chamber. through a nozzle or an aperture. on CsI or RbBr. scattering. whose states are accessible with the UCSB FEL. 4 for the HD/ of enhanced electric field from mechanical coupling to the phonon motion.1 Light species such as HD exhibit strong quantum-mechanical features such as diffraction. An example is shown in Fig. Structure and Dynamics of Clusters can then be used to prepare a rotationally excited beam for further interaction studies with other gases. 25 and inelastic. Comsa and Poelsema have used this technique to examine adsorbates and defects on several surfaces. is HF. This is now understood in terms of straighforward collision dynamics. Most of these have been in the UV portion of the spectrum and have used electronic state spectroscopy. Am. By using the UCSB FEL for the beam excitation. Another type of experiment is rotationally mediated selective adsorption processes. sticking of prescribed rotational states. A few near-IR experiments have been reported that probe the dynamics of vibrational energy transfer in clusters. while the coupling to the adsorbate is followed by mass spectrometric analysis of the desorption rate. More detailed considerations will likely be necessary to separate clearly the effects tial cross section. 5/May 1989 Simon et al. for HD on Ag(111).

The pairwise additivity of the interaction potential between atoms in van der Waals bonded systems. the problems of simulation 27 The interactions in these become much more complex. 18.. At present our programs are fully operational.Simon et al. However. ). most notably those derived from the symmetry of the icosahedral point group (Mn. which provides a measure of their translational energy as they leave the parent cluster. Opt. 4. tion can be either on the beam center line. These simulations can be used to support interpretations of experi- does not appear to provide a viable improvement. These studies have shown that certain size clusters are especially stable. Since both beams are pulsed it is straightforward to arrange the appropriate timing. Only surface modes with vibrations perpendicular to the surface have been identified. clus- mental data. Adapted from Ref. 6. because of the lack of a dipole moment in a pure van der Waals cluster. For example. Doppler broadening is larger in this geometry. Am. For a velocity distribution translational temperature of 10 K. Such a measurement is obviously much more precise in the crossed- beam geometry. compared with the 10-8 mode purity per pulse of the FEL (see Table 1). For example. Adding higher-body phenomenologicalcontributions The study of the dynamics of these systems has led to some interesting results. However. Av/v0 = 3 X 10-7. so that a decrease in intensity is observed as the cluster fragments. . B 1041 4 C\ 2 Can 01 ! I I 30 II I I I i 0 60 90 Fig. Because of the possible relationship of this phenomenon to bulk transitions it has become a topic of considerable theoretical interest. Vol. which display more interesting chemistry. Unfortunately. In order to simulate such systems we have developed software to implement an ab initio molecular dynamics approach similar to that recently suggested by Car and Parrinello.with the availability of large computers it is now possible to simulate complicated dynamical processes theoretically with considerable accuracy. and we are using them to calculate the dynamics of systems of interest to the experimental program. The method requires no empirical information other than the nuclear charges of the constituent atoms and therefore can provide predictions about systems with unknown chemistry. simulations of melting in mixed and pure thin films may be of value to the development of damage-resistant materials for laser mirrors. Simulations of Energy Transfer in Doped van der Waals Clusters: The spectroscopy of molecular and atomic clusters is currently an area of intense experimental and theoretical investigation. makes these systems easily simulated. for systems with metallic and covalent bonding. where the fragments give an increase in signal and the noise of the primary beam is avoided. 2 9 From a theoretical point of view. to increase the interaction density substantially and thereby favor multiphoton processes.2 8 In this approach the potential interactions are calculated from an approximate solution to the many-electron Hamiltonian for the system. 5/May 1989/J. SP11 and SP 1 indicate longitudinal and transverse sagittal-plane vibrations. systems are not satisfactorily described by two-body interactions. together with the spherical symmetry of the potential for rare gases. respectively. Soc. In the second geometry the photon flux is brought in along the axis of the molecular beam. constant energy simulations have identified a sharp liquid-solidlike phase change in clusters as small as seven atoms. the methods of simulation are straightforward. or off the axis of the primary molecular beam. In both cases it is possible to obtain dynamical information by measuring the time of flight of the fragments to the detector. 55. experimental data for these . n = 13. The generality of the programs that we have developed is such that the programs can be used in other studies of direct interest to the development of the FEL. The dynamical behavior of these systems is considerably more complicated than that of the molecular systems that are more commonly studied by chemists. No. For this reason our group has developed the theoretical tools and software to simulate accurately systems of complex chemistry. In what follows we briefly describe our progress with our simulation programs. Theoretical Studies A primary objective of our FEL research program is the study of the structural and dynamic properties of condensed-matter systems. Phonon dispersion curves for 15-layer unrelaxed and relaxed slabs of RbBr with free (001) surfaces: S1 and S1' indicate the acoustical and optical Rayleigh modes. For systems interacting by means of van der Waals forces ters of rare-gas atoms are certainly the most studied. 147.

and N2Ar 46 .9 studied argon clusters doped with IRactive SF6 . indicating solidlike vibration in the solvation shell. No. N2Ar 2 .0 lower than the predicted three-dimensional transition. The impurity forms a spherical surface on which the pure atoms form a solidlike ordered phase at low temperatures and a liquidlike phase at higher temperatures.3 temperature 0. 0 E 2. Other experiments with a similar objective have been published by Hahn and Whetten3 0 and Bosiger and Leutwyler. As the temperature is raised the bond-length MSD remains roughly constant. using various 40.0 4. . By studying the shifts from the gas-phase IR spectra with the addition of argon atoms. also using molecular dynamics simulations. these workers hoped to obtain information about the dynamics of the cluster system.32 For example. indicating a two-dimensional melting in the incomplete solvent shell of the SF6. 5. 5/May 1989 Simon et al. 6.1 0. N2Ar5 4. In this figure the open circles show the MSD's for the bond length from the impurity atom (SF6) to the solvating argon atoms. However.0 0 0. We have studied the energy transfer dynamics in the three smallest icosohedral structures. Calculated MSD of an A12B cluster. On the other hand the bond-angle MSD shows a sharp change with temperature. Temperature dependence of various shells of the Ar 54N2 cluster after excitation of the embedded nitrogen molecule. our more recent simulations suggest that the situation in the mixed systems is considerably more complicated.0 1. Other supporting evidence for this transition can be obtained from 32 the simulations.0 X 30.9 we have found a sharp two- dimensional phase transition at a temperature considerably 50. . Soc.0 2.. The filled circles show angular MSD's for the same bonds.3 1 Superficially. Because there are not enough atoms the impurity is incompletely solvated. 6. second shell. First shell. The filled circles are systems are not available. 5. B/Vol. We have begun research along these directions.0 5. Am.0 3. Opt. -----. the flow of vibrational energy from the excited molecule into the cluster can be examined.1042 J. The best evidence for this abrupt change in the dynamics of this system is given by the calculated mean-square displacements (MSD's) as shown in Fig. The open circles are the MSD of the motion perpendicular to the shell. these experiments appear to support an interpretation in terms of the sharp three-dimensional transition apparent in the simulations of pure clusters. 0.4 the MSD of the motion within the solvation shell.0 a) 10.0 CU E Ca2 d) 20. -- .2 0.0 Time (psec) Fig. Extensions to experimental studies using the FEL light at UCSB are planned. third shell. The structure of this system is dominated by the strong interactions between the impurity atom and the pure atoms. Gough et al.4 0 (n. Our calculations on vibrational relaxation have focused on rare-gas clusters (xenon and argon) doped with molecular nitrogen. In order to circumvent this difficulty. The goals of this aspect of our research are to understand (1) vibrational relaxation in clusters and (2) the dynamics of photodissociative systems in rare-gas clusters. Clusters of this type are easily formed by using molecularbeam techniques. Even less is known about either energy transfer in clusters or the dynamics of simple chemical reactions in clusters. in systems meant to simulate the IR experiments of Gough et al.6 c08 E @ 0 Fig. By monitoring the potential and kinetic energy of the nitrogen and argon atoms followingvibrational excitation of the diatomic.0 0.

The data required. Vol. but it has been calculated to a high approximation by numerical methods. The Hamiltonian for the equivalent problem contains a potential. By using the maxima in the temperature curves for the first and outer shells. An effective method of updating the LDA wave longer sufficiently accurate. numerous calculations have shown that the results approach the accuracy of configuration interaction calculations. for studies of systems of more than a few metal atoms. is a difficult computational task. however. The temperature data also show that neous electron gas. Soc. The length varies strongly with bond coordination. an estimate of the propagation velocity of the thermal wave is found to be -2000 m/ sec. the dynamics of the system can be adequately described by interactions on one Born-Oppenheimer surface. even for a diatomic by a configuration interaction approach. In this case. and density as the original many-electron problem. Results for our simulation of vibrationally excited nitrogen in the rare-gas cluster of Ar 54 are presented in Fig. Recently there has been considerable interest in an alternative approximate approach to the quantum-mechanical problem. In the LDA. that is a function of the electron density in a way similar to the Hartree-Fock potential.Kohn and Sham36 have shown that the solution to the equivalent problem will have the same energy Fig. 5/May 1989/J. the local density approximation (LDA). Furthermore. calculating the solution to the full Schrodinger equation for a single potential energy. Although they are approximate. 6. however. Calculated structure of the Be 13 cluster. for covalent and metallically bound systems the usual approach to simulation in which the interactions between molecules are described by analytical potentials such as the Lennard-Jones potential is no to the extent that limited dynamical simulations are now possible. In this ab initio molecular dynamics approach the electronic states and atomic motions are computed simultaneously. 32 Since energy propagates from the inner shell to the outer shells in a wavelike manner. to parameterize such an approach are also quite extensive and are not generally available for systems that exhibit interesting chemistry. Unfortunately. . g(r)N 2 -Ar. the potential for the equivalent problem is known only in principle.34 These data indicate that immediately following the excitation of the diatomic. Opt. The average bond length is 3. In this method the many-electron quantum mechanical problem is replaced by an equivalent one-electron problem. B 1043 excitation energies of the nitrogen. In this case. use of three-body 27 Therefore theoretical inerate electronic configurations. Changes are also observed in the pairwise radial distribution functions. The shells are defined by peaks in the N2 -Ar pairwise distribution function. Recently Car and Parrinello2 8 showed that the LDA method can be implemented in a way that improves its efficiency Simulations of Covalently and Metallically Bound Systems: As mentioned above. This coupled with the proven accuracy and efficiency of the LDA approach is proving to be a powerful and effective tool for predicting properties of quite large systems. the experimental data are more complicated than those found for rare-gas clusters because of the presence of a number of nearly degenput becomes even more important for interpretation. g(r)ArAr and g(r)N2 Ar. This is in excellent agreement with the speed of sound in solid argon. as illustrated. the method can be applied to an arbitrary system simply by specifying the number of electrons and the nuclear charges of the constituent atoms.Simon et al. 6. the exact exchange correlation potential for a particular problem is replaced by the exchange correlation potential for a homoge- The data in this figure correspond to an excitation equal to 35 quanta of the stretching vibration. which provides a highly accurate and practical approximation. No. Am. Unfortunately. 35 the only approximation in the theory is the replacement of the potential with the homogeneous electron gas potential. In this plot the time dependence of the temperature of various shells of the clus- ter is plotted. For most systems the forces between atoms can be calculated from a single solution to the time-dependent manyelectron Schrodinger equation.1 A. approximate approaches such as that of Hartree-Fock are seldom accurate enough to be relied on to predict chemical behavior. A possible solution to this dilemma. On the other hand. This work. was suggested by Kohn and Sham. the first shell temperature increases rapidly. the dynamical information is important and applicable to understanding other systems. Vari- ous efforts have been made to simulate these systems by the potentials to improve the force calcula- tions. This potential is also not exactly known. either computed or measured. The melting temperature of the Ar55cluster is 40 K. the exchange correlation potential. 7.33 Although this particular system cannot be experimentally probed by using the UCSB FEL. The structures and dynamics predicted from such approaches disagree strongly with accurate calculations. has not been successful. 1800-2400 m/sec.

18. Fenn. 29. H. 40. Phys. Wharton. Amsterdam. Adv. J. Gough. Smalley. Phys. (to be published). C. M. 69. Arthur. X. 4. Chem. Lett. Chem. 88. McGinty.. Lett. Andersson and J. T. Kwok. 1970). J. G. ed. Sci. ACKNOWLEDGMENTS The ongoing studies reviewed in this paper represent the work of a number of students and research associates. 24. G. 83. 151. Hohler. New York. N. Chem. Spears. 6. and F. Surface Polaritons (NorthHolland. Rev. 19. 12. New York. Am. J. Gough. New York. 77. Y. SF6 on Ar. and F. Ontario. L. QE-23. F. T. Phys. K. R. F. 66. Long. Chem. M. D. 4. p. and L. Chem. Krajnovich. M. Rev. 349. eds. 17. 84. Phys. Engel and K. 48. New York. A. R. Zhu. 70 of Advances in Chemical Physics (Wiley. P. 16. Knight. Khana. 8. See the special issue on free electron lasers. 15. 1985). A. J. G. 5/May 1989 Simon et al. Time Resolved Vibrational Spectroscopy (Academic. S. Briant and J. 9. Ryoichi Kawai. Rev. Rothschild. P. A. and G. Sci. 3 7 Because of its closed s2 atomic configuration. New York. in Structural Studies of Surfaces. J. ed. J. Tamkin. P. R. London. L. T. Recently several attempts to quantify this progression of behavior as a function of system size have been published.1044 J. J. S. Kulkarni. G. W. Berry. L. Phys. 221 (1985). Lee. QE-23. functions as the atomic positions are updated in the simulation further increases the efficiency of the LDA calculation in the Car-Parrinello approach. As noted above. C. 59. Simon is a National Science Foundation Young In- vestigator (1985-1990) and an Alfred P. Demuth. David Smith. Sibener. Phys. 167 (1988). J. 47 (1982). in- sonicjets. R. T. 1975. Quantum Electron. Davis. 55.. L. W. 385 (1987). 64. J.11 eV. Rev. George. 369 (1988). Phys. 63. M. 1982). it is well known that systems interacting through van der Waals forces adopt a highly stable 13-atom icosahedral structure. B. Yu. D. Far-Infrared Spectroscopy (Wiley-Interscience. J. Kappas. "Infrared spectroscopy at the surface of clusters-a new type of surface spectroscopy: small clusters of beryllium are weakly bound. 36. Phys. Phys. L. Our calculated results suggest that there is considerable sp interaction in quite small cluster sizes. Phys. Phys. L. Schmeisser. and S. Hogg. Quantum Electron. Phys. Chem. J. 49 (1979). N. Soc. and H. J. Gross. C. Mitra and S. R. E. 83. Rev. New York. 60. Ertl. 377 (1987). and G. Keilmann. Burton. in Rarified Gas Dynamics. in Evolution of Size Effects in Chemical Dynamics. L. Car and M. 20. 14. Roundtree. 5. D. D. Wang. 5783 (1987). Bratos and R. R. R. Yang. K. Vibra- tional Spectroscopy of Molecular Liquids and Solids (Plenum. Weidenauer. Kiess. J. Scoles. C. "Laser spectroscopy in super- The complexity of the structure in Fig. F. M. L. Vol. Phys.. New York. 26. This research is supported by a grant from the Strategic Defense Initiative Material and Medical Free Electron Laser Program. A. T. Tro. J. 63. in The REFERENCES 1. IEEE J. and M. Commun. Jena. Miller. P. and H. 1 (1984). Schroder. New York. The structure in this figure is fully optimized with respect to bond angle and lengths. . London. J. Opt. Chem. Canada. and J. 88. Cole. Nudelman. Lett. Beck and R." J. 1518 (1988). Rep. H. Barker and D. F. Phys. (Academic. Weide. deWette. 153 (1985). Appl. We have developed the theory and software necessary to implement such an algorithm. Elias. Sham. Cohn. 199 (1982). D. D. 1981). J. 7 illustrates the expected difficulty of describing such a system with phenomenological force equations. 1989). T. Phys. Button. K. 10. D. Chem. Hanza. in Chemical and Biochemical Applications of Lasers. Rice.. 7.1190 (1988). M. Press. Vol. B 35. 61. Mengel. 2045 (1975). G. Beck. 1988). and S. Roundtree. R. Rev. 1977). Vol. X. Phys. G. Venables. Proceedings of the cluding Shyh-Gang Su. X. The size of the largest beryllium cluster calculated by other authors is five atoms. Far-Infrared Properties of Solids (Plenum. Levine. A. Scoles. 48. and references therein. Chem. Levy. Phys. 34. Keilmann. 6. Leutwyler. Avouris. R. (North-Holland. W. Burton. 1895 (1987). K. Molecular Spectroscopy New York. p. To illustrate some of our results we show in Fig. eds. Lett. Garzone. 1971). U.. 413 (1977). Vollmer. 4958 (1985). Shen. 91 of Springer Tracts in Modern Physics (Springer-Verlag. Chem. 11.4733 (1973). ed. Chem. Reyes. Su and J. 4235 (1985). Miller. Additional calculations that we have done have shown that for Be13 the hp structure is already close in energy to the icosahedral structure. P. and S. 33. 1975). Simon.Quantum Electron. 31. Agranevich and D. Chem. Waldman. IEEE J. 1982). 1983). Opt. Kress. QE-23. R. 13. Trager. Andresen. Tokyo Press. S. Freund. Quantum Electron. 86. Mengel. Gough. and Xiping Long. Z. R. Kawai. Hoare. 140. Chem. A1133 (1965). 3266 (1976). Budde. A. D. 1976). 35. L. L. in Atomic and Molecular Beam Methods. J. Tokyo. Rev. Atkinson. Wharton. Rieder. 28. T. and G. Phys. K. p. Stegeman." to be submitted to J. 1. Moller and W. Drozdowicz. The stable structures for small metallic systems are essentially unknown. 27. Briant and J. H. No. and Y. S. 1987).2783 (1986). Phys. M. Bulk beryllium adopts a hexagonal close-packed structure. M. QE-13. H. Wallis and G. CP-2716 system is increased the bonding changes from van der Waals to metallic. J. Goodman. 97. Whaley. V. Rep. Elliot and D. E. T. Bruce Andrein. (Oxford U. Berlin. and W. 1470(1987). R. K. R. New York. Quantum Chemical Physics of Atomic and Molecular Clusters. 1588 (1978). R. A 38. Demtroder. Weare. J. 545 (1982). J. 2471 (1985). Berry. B. Electromagnetic Surface Excitations (Springer-Verlag. D. Whetten. "Energy relaxation in nitrogen-doped Ar clusters. Isobe. 3. Lett. and J. H. 2 (Academic. Scoles. Vol. G. Light. R. Harris. Vol. Sloan Fellow (19881990). I. P. Gough. Laser Spectroscopy: Basic Concepts and Instrumentation (Springer-Verlag. Oguchi. Miller. Phys. and J. and L. Hoheisel. T. S. S. K. I. 1649 (1988). S. 30. J. Smalley. IEEE J. Rev. Morse. T. 25. Metallic behavior is present even in systems as small as five atoms. V. Rev.2045. M. Auerbach. New York. J. Hahn and R. Moore. Beck and R. Bosinger and S. We thank Vincent Jaccarino and the UCSB FEL Group for their help in carrying out experiments. For example. Theiler. Comsa and B. M. I. 58. Phys.Scoles. 2476. G. II. QE-23(9) (1987). Berry. ed. International Symposium of the Physics and Chemistry of Small Clusters (Plenum. and F. B/Vol. Poelsema. Mayer. Phys. 1 (Academic. As the size of the (University of Waterloo. 60. Lett. Parrinello. Proceedings of the 14th International Symposium on Rarified Gas Dynamics (U. Wessel. Surf. D. Mills. and G.. Rao. Kohn and L. Picks. G. 1980). "Infrared resonant desorption of butane from A1 2 0 3 (1120): evidence for an ordered adlayer from vibrational mode selectivity. Lisy. Waterloo. Rare Gas Solids. Levy. Prigogine and S. J. Electron. eds. G. Phys. 3910 (1988). R. Phys. Chem. D. the bond energy of the dimer is 0. 23. Lett. Scoles. IEEE J. 155 (1983)." Chem. Bob Ning. C. L. Jungmann. Surf. 22. 2. IEEE J. Phys. 21. (Wiley. 32. P. A. Scoles. Woskoboinikow. Ph. New York. Klein and J. Lett. 1984). D. M. 1988). Chem. ed. Chem. Vernon. L. 1977). 1986). D.-G. 1049 (1986). D. Rev. 337. J. 7 the calculated structure of the Be13 system. Rev.