It has been appreciated for years that the reaction order, activation energy and preexponential factor for reactions on surfaces often show a coverage dependence when the rate is expressed in simple Polanyi-Wigner form, R = n e^-E/RT qⁿ. This complex behavior originates in the lattice nature of the adsorbed phase and the existence of lateral interactions among adsorbed species. Lattice gas and Monte Carlo treatments of surface reactions indeed indicate that such complex behavior is to be expected. Moreover, even for some simple surface reactions of low molecularity, such as the reaction of preadsorbed oxygen on platinum with CO or H₂, complex reaction kinetics due to islanding of the reactants occurs, and an accurate treatment of the reaction kinetics requires an understanding of the spatial organization of reactants on the surface. Indeed, a recent STM study shows conclusively that the reaction of CO and oxygen adsorbed on Pt(111) near 250 K is proportional to the linear boundary between the two separated phases of the reactants. Under such circumstances it is obvious that the reaction is not proportional to the average concentration of reactants over the surface. However, even when lateral interactions are included in the kinetic analysis, assumptions are often made, such as equilibrium among the surface phase that dictate the form of the kinetic expressions. In view of our STM results for the CO oxidation on Cu(110), it appears that such simplifying assumptions can be quite incorrect in some regions of temperature and pressure. In fact, STM studies have shown many examples of surface reconstructions accompanying adsorption of simple reactants, and their reactions must necessarily involve complex surface structure changes which possess their own characteristic time scales. Indeed, macroscopic kinetic oscillations in surface reactions have been at least in part attributed to such effects. The pursuit of the understanding of the origin of complex surface reaction kinetics is the motivation for the core of the research proposed here.

We have recently initiated a new program devoted to the study of surface reactions on an atomic scale utilizing scanning tunneling microscopy. We first designed, constructed and successfully tested a novel controlled environment, variable temperature scanning tunneling microscope that would allow the use of scanning tunneling microscopy in combination with other important surface diagnostic tools without removal of the single crystal specimen from the crystal manipulator (see figure below). This design allows us to prepare surfaces with reactive intermediates using normal ultrahigh vacuum dosing procedures, apply selected analytical methods to their ensure their identity and then examine them in atomic detail at selected temperatures without altering the surface temperature due to transfer to the STM. We have used this microscope to study the heterogeneous oxidation of carbon monoxide and ammonia, as well as to discern the evolution of structures of adsorbed phenoxide, on Cu(110).

FIGURE 1: A cross-section of the UHV STM chamber showing the chamber, the beam frame and the air legs. The top half contains surface science anaylsis equipment and the sample manipulator. The inset shows a schematics diagram of the STM head.

The variable temperature UHV STM is capable of monitoring surface reactions between 120 and at least 500 K. Though we have not exceeded 500 K with our instrument, temperatures of 700 K have been reached with in an instrument using nearly the same scanning head design. The system incorporates a scanning tunneling microscope of the "Johnnie Walker" type into a versatile UHV system equipped with other diagnostic equipment necessary for studies of surface reactivity. In principle, the design allows the combination of a wide variety of surface science measurements with STM. Our system incorporates scanning tunneling microscopy, rearview low energy electron diffraction, Auger electron spectroscopy, argon ion bombardment, temperature programmed reaction spectroscopy (TPRS), and gas dosing facilities. Atoms within the p(2x1) islands of -Cu-O- rows on Cu(110) are clearly visible at all temperatures of operation (see images below); intermediate low temperatures were achieved with a combination of continuous liquid nitrogen cooling and radiative heating of the sample.

FIGURE 2: STM images taken at various sample temperatures from 120 to 400 K showing atoms within the p(2x1) islands of -Cu-O- rows on Cu(110).

After completion of the initial performance tests, we examined the oxidation of CO on Cu(110). This reaction was studied because oxygen is known to form islands of added-row O-Cu-O strings in a (2x1) structure on Cu(110), and site- and direction-specific reactivity was possible. Furthermore, the previously studied macroscopic kinetics could be compared to the STM observations. In order to obtain observable rates of reaction of CO with preadsorbed oxygen by STM, a surface temperature of 400 K and a CO pressure of 10^-4 was required. At 400 K CO reacts with the oxygen islands most rapidly along the -O-Cu-O- rows (seel images below). Figure (a),(b), and (c) show sequential topographic STM images of oxygen on Cu(110) initially in the state depicted by figure above reacting with CO at 400 K. The oxygen rows on the island perimeter react away in a sequential fashion, leaving two individual rows shown in (c). Figure (d) is a close-up image of the resulting clean surface obtained at 400 K showing individual Cu atoms, verifying that the surface is clean after the reaction. If the defect density at the perimeter of the p(2x1) islands is low, the reaction appears to be initiated by penetration into the side of an oxygen (2x1) island along the (11¯0) direction, followed by rapid reaction in the (001) direction. This effect is evident in figure 3b, which clearly shows fragments of a single oxide chain formed by reaction along the (001) direction (arrow B) without significant further penetration into the island in the (11¯0) direction. Monte Carlo simulations suggest defect sites in the island boundary are 500-1000 times more reactive than non-defect sites.

FIGURE 3: Sequential STM images of p(2x1)-O reduced by CO. The images between each frame is 150 s. In (b) arrow "A" marks a row that fragmented and shifted one row to the left; arrow "B" marks a row at an island edge that fragmented; arrow "C" marks a transient row not observed in either (a) or (c). Image (d) shows a topographic image of the clean surface after the oxygen is completely reduced.

Furthermore, the highly defective nature of the islands during reaction clearly indicates that equilibrium between oxygen islands is not maintained during the reduction of the surface oxygen by CO at 400 K. This fact introduces a major complication into the description of surface reaction kinetics which must be addressed by theorists. Specifically, equilibrium statistical mechanics does not always describe the distribution of reactants on the surface and transport rates must be incorporated into theories of surface reactions in order to properly describe the reactant distribution over time. We also noted that the copper atoms released by reduction of the surface oxide with CO always appear to migrate to the step edge under the conditions of our experiments; no Cu atom aggregation on terraces to form Cu islands was observed, and oxygen adsorption followed by reaction with CO does not produce surface roughening.

In order to better simulate catalytic conditions we also monitored the reaction under isosteric conditions with both oxygen and CO present in the gas phase at fixed partial pressures. Islands maintain their general size and shape; preferential reaction and reaggregation of oxygen occurs at kink sites on the perimeter of the (2x1) oxygen islands. The continual fluctuation in the structure of the kink sites at the islands perimeter suggests that reduction at kink sites is followed by readdition of oxygen to the island at kink sites. Indeed, oxygen addition to defect sites in the p(2x1) overlayer during adsorption has been established in previous studies. In addition, mobile copper adatoms are always present on the terrace; copper atoms are continuously released from the reaction of oxygen with CO and reincorporated into the p(2x1) island structure by reaction with adsorbing oxygen. A continuously fluctuating island structure is typifies the surface during the steady state reaction.

The absence of reaction of the p(2x1) O structure with CO at 300 K we observed seems to contradict reports of reaction between coadsorbed CO and oxygen near 100 K This apparent anomaly spurred our interest in studying the reaction at low temperature. STM revealed that when an oxygen-precovered copper (110) surface partially covered with (2x1) islands is exposed to an additional 10 Langmuir of oxygen at 150 K, short, isolated, immobile Cu-O chains oriented in the (001) direction form on the areas of clean surface between the islands. When dosed onto this surface, carbon monoxide forms striped structures in the interstitial regions which are oriented in the (110) direction, perpendicular to the -Cu-O- chains and span the (2x1) islands. This CO structure is stable for at least 30 minutes under STM observation, and the adsorbed CO does not react at 150 K with either the (2x1) oxygen or short Cu-O chains between these islands.

When the (2x1)O surface is predosed with carbon monoxide to form this striped structure and then exposed to oxygen at 145 K, however, carbon dioxide is evolved. Sequential images of the mixed CO-O adlayer continuously exposed to 4x10^-9 Torr of oxygen show that the CO stripes all disappear and are replaced by short, interstitial Cu-O strings. The clear implication of these results is that CO reacts readily with adsorbed oxygen that is not the most stably adsorbed form. We believe this form to be dissociatively chemisorbed oxygen which has not yet combined with migrating copper atoms and is thus more susceptible to reaction. Isolated oxygen atoms have indeed been imaged at lower temperatures on copper; they appear to be very mobile above 100 K.

Publications arising from this research

1. W.W. Crew and R.J. Madix, "A Variable Temperature Scanning Tunneling Microscope for the Study of Surface Reactions in Ultrahigh Vacuum," Rev. Sci. Instr. 66 (1995) 4552

2. W.W. Crew and R.J. Madix, "Monitoring Surface Reactions with Scanning Tunneling Microscopy: CO Oxidation on p(2x1)O-Precovered Cu(110) at 400 K," Surf. Sci. 319 (1994) L34

3. W.W. Crew and R.J. Madix, "A Scanning Tunneling Microscopy Study of the Oxidation of CO on Cu(110) at 400 K: Site Specificity and Reaction Kinetics," Surf. Sci. 349 (1996) 275

4. W.W. Crew and R.J. Madix, "CO Adsorption and Oxidation on Oxygen-precovered Cu(110) at 150 K: Reactivity of Two Types of Adsorbed Atomic Oxygen Determined by STM," Surf. Sci. 356 (1996) 1

5. X.-C. Guo and R.J. Madix, "Monolayer Structures of Phenoxy Species on Cu(110): an STM Study," Surf. Sci. 341 (1995) L1065

6. X.-C. Guo and R.J. Madix, "Site-specific reactivity of oxygen at Cu(110) step defects: an STM study of ammonia dehydrogenation," Surf. Sci. 367(1996)L95

7. X.-C. Guo and R.J. Madix, "In-situ STM imaging of ammonia oxydehydrogenation on Cu(110): The reactivity of preadsorbed and transient oxygen species," Surf. Sci. 387(1997)1

8. X.-C. Guo and R.J. Madix, "Atom-resolved investigation of surface reactions: ammonia and oxygen on Cu(110) at 300 and 400 K," Faraday Discussions 105(1996)139

9. X.-C. Guo and R.J. Madix, "Non-uniform product inhibition in surface reactions: spatial organization effects in ammonia oxydehydrogenation on Cu(110)," J. Chem. Soc., Faraday Trans. 93(1997)4197

10. R.J. Madix, W.W. Crew and X.-C. Guo, "Chemical Reactions on solid surfaces: atomistic observations by scanning tunneling microscopy", Appl. Surf. Sci. 113/114 (1997)539

Please visit our Image Gallery for additional STM images.