J. Phys. Chem. B 2004, 108, 14663-14670
14663
Growth of Copper on Reconstructed Pt(100) B. Schaefer, M. Nohlen, and K. Wandelt* Institut fu¨r Physikalische und Theoretische Chemie, UniVersita¨t Bonn, 53115 Bonn, Germany ReceiVed: February 19, 2004; In Final Form: May 24, 2004
The growth of copper on the hexagonally reconstructed Pt(100) surface has been studied in UHV by scanning tunneling microscopy (STM). The copper film grows anisotropically and initially almost in a bilayer fashion due to a high diffusion barrier at copper island edges and a reduced mobility on the Cu islands. The substrate reconstruction is lifted only locally by Cu island formation. This lifting leads to an ordered incorporation of platinum excess atoms into the first layer of the growing copper film: At 340 K, parallel, monatomically wide platinum chains are formed, whereas at 650 K, broader line structures appear in which the platinum atoms are again hexagonally arranged. Up to nine monolayers, the Cu films remain pseudomorphic with the Pt(100) substrate but undergo an fcc f bct transition beyond 4 monolayers.
1. Introduction Metal heteroepitaxy on anisotropic substrates is of particular interest as it opens up the possibility to produce nanostructured layers by mere selforganization of the deposited material. Natural anisotropic substrates are reconstructed fcc(110)(n × 1) or well prepared vicinal metal surfaces which are often used as templates for unidirectional growth structures. Another class of anisotropic substrates are the reconstructed (100) surfaces of Pt, Ir, and Au, which exhibit a quasi-hexagonal reconstruction of their outermost surface layer.1-10 The atom density of this reconstructed layer is about 20% higher than that of an unreconstructed surface, i.e., the (100)(1 × 1) phase.11 Furthermore, due to the misfit between the hexagonal surface layer and the unreconstructed second layer, the surface atoms occupy a range of inequivalent sites varying from 4-fold hollow-sites to on-top sites. As a result, along the 〈110〉 direction, six surface atoms are resting on five second layer atoms leading to a wavy superstructure with a periodicity of ∼14 Å. The height modulation of this superstructure along the 〈110〉 direction amounts to ∼0.4 Å giving rise to pronounced troughs and ridges perpendicular to this direction. The driving force for the reconstruction is the increase in coordination within the reconstructed surface layer and the concomitant gain in binding energy.9-12 In the present work, we report on the growth of copper on the reconstructed Pt(100) surface. There exist actually two different hexagonal reconstructions of the Pt(100) surface. A metastable, “unrotated” hexagonal surface layer forms after annealing at temperatures below 1100 K. This reconstruction is usually denoted (5 × 20) or Pt(100)-hex phase. Annealing at 1100 K leads to the stable Pt(100)-hex-0.7° structure which is rotated by 0.7° with respect to the Pt(100)-hex phase.4,5,8,9 In the present context, the difference between both structures is considered to be minor, so that the surface will only be termed Pt(100)-hex throughout this paper. Previous studies on the adsorption of gases such as NO, H2, CO, O2, or C2H4 have shown that they cause a lifting of the reconstruction.8,13-15 The same has been observed in the few studies on metal deposition, namely Pt, Sm, Au, Ag, and Cu on the reconstructed Pt(100) surface16-24 published to date. * To whom correspondence should be addressed.
The diffusion and nucleation behavior of Pt adatoms during homoepitaxial growth on Pt(100)-hex has been studied by Linderoth et al. using STM.16,17 The Pt adatoms diffuse strongly anisotropically along the reconstruction troughs and form rectangular islands with their main axis aligned parallel to the troughs. The characteristic Pt(100) reconstruction is seen to persist on these islands, from which the authors conclude that the reconstruction underneath the islands has been lifted. Furthermore, the reconstruction appears already on very small Pt islands, indicating that even small adatom clusters are sufficient to induce the hex f (1 × 1) transformation. Berg et al.22 investigated the growth of gold on the Pt(100)hex-0.7° surface and found a strongly anisotropic growth accompanied by a local lifting of the substrate reconstruction, both proceeding much faster parallel to the troughs than perpendicular to them. As the hexagonal layer is more densely packed than the evolving (1 × 1) structure underneath the Au islands, the transition to the unreconstructed surface produces excess Pt atoms, which are expelled from the transforming Pt layer and are found to be incorporated in a partly ordered fashion into the (1 × 1) Au overlayer. Likewise Batzill and Koel23,24 studied the growth of silver on the Pt(100)-hex-0.7° surface and found a preferential growth along the direction of the troughs accompanied by lifting of the reconstruction and incorporation of Pt clusters into the growing Ag islands. The system Cu/Pt(100) has already been studied before but mainly with LEED and AES. Barnes et al.25 report on an initial layer-by-layer growth. The Pt(100) reconstruction was seen to be lifted already by submonolayer coverages of Cu, and the Cu found to grow pseudomorphically up to 2 ML. Li et al.26 observed a gradual lifting of the reconstruction with increasing Cu deposition until a coverage of 2-3 ML was reached. According to Radnik et al.,13 the lifting is completed by one pseudomorphic Cu monolayer on the unreconstructed Pt(100)(1 × 1) substrate. A relaxation of the Cu film occurs only for higher Cu coverages. Our present investigation of copper growth on the reconstructed Pt(100) surface completes the series of STM studies on the deposition of “coin metals” on Pt(100)-hex and differs from the previous Au/Pt and Ag/Pt systems in that Cu atoms are smaller than Pt atoms, whereas both Au and Ag are larger
10.1021/jp0401527 CCC: $27.50 © 2004 American Chemical Society Published on Web 07/24/2004
14664 J. Phys. Chem. B, Vol. 108, No. 38, 2004
Schaefer et al.
Figure 1. Pt(100)-hex substrate is strongly anisotropic and consists of troughs which appear to be ∼0.45 Å deep and are ∼14 Å wide. 55 × 55 Å2, UB ) 33 mV.
than Pt. Our STM studies, indeed, prove that also Cu deposition lifts the Pt reconstruction only locally and demonstrate that the anisotropy of the substrate leads to a highly anisotropic growth of the first Cu layer. The concomitant lifting of the reconstruction induces an incorporation of Pt atoms into this first Cu layer, which in turn causes an anisotropic growth of higher Cu layers. The initial growth mode is found not to be layer-by-layer. In this paper, we concentrate on measurements carried out after Cu deposition at substrate temperatures of 340 and 650 K, respectively. 2. Experimental Section The experiments were performed in a UHV system (base pressure 1 × 10-10 mbar) equipped with STM, AES, and a Kelvin probe for work function measurements. The homemade STM is described elsewhere.27-29 The Pt(100) crystal was prepared by sputtering with 600 eV Ar+ ions and subsequent annealing at 1000 K for 10 min. Segregated carbon could be removed via exposure to an oxygen atmosphere (pO2 ) 1 × 10-6 mbar) at 800 K for 2 min. Finally, the sample was flashed to ∼1100 K in order to desorb remaining oxygen and to produce the reconstructed surface. The Cu films were deposited by evaporating Cu from a resistively heated tungsten basket. During Cu evaporation, the pressure never exceeded 5 × 10-10 mbar. Deposition temperatures were chosen to lie between 340 and 650 K. The sample was allowed to cool to room temperature before the STM measurements were carried out. The deposition rate was ∼0.5 ML/min. The tunneling current was always 1 nA. 3. Results and Discussion 3.1. Submonolayer Growth at 340 K. Figure 1 shows an atomically resolved image of the clean Pt(100)-hex substrate. The first layer atoms are quasi-hexagonally rearranged. As the underlying layers are unreconstructed and of square symmetry, the first layer is not flat but exhibits a wavy pattern in the form of parallel troughs and ridges. The wavelength between two troughs is ∼14 Å and corresponds to 6 Pt atom rows, which
Figure 2. Long narrow Cu islands on Pt(100)-hex. Island formation mainly proceeds after heterogeneous nucleation at substrate steps (a) and at substrate domain boundaries (b). Td ) 340 K, ΘCu ∼ 0.2 ML. (a) 4300 × 4300 Å2, UB ) 500 mV. (b) 3200 × 3200 Å2, UB ) 500 mV.
together form one “reconstruction row”. In the STM images, they appear to be ∼0.45 Å deep. This value, however, depends on the shape of the tip and was found to vary by (0.15 Å Figure 2a gives an impression of the typical film morphology after deposition of ∼0.2 ML Cu at 340 K. Predominantly after heterogeneous nucleation at ascending as well as at descending substrate steps, long, narrow islands have formed. However, islands were not found to nucleate at steps that run parallel to the reconstruction rows. The Pt(100)-hex surface also exhibits different reconstruction domains which are rotated relative to each other by 90° due to the square symmetry of the Pt substrate underneath. Figure 2b demonstrates that nucleation of Cu islands takes place also at the corresponding domain boundaries. Such a boundary is running vertically through the image (apparently undisturbed by substrate steps) and obviously serves as a nucleation center for Cu adatoms. To both sides of this boundary Cu islands grow into two orthogonal directions which are given
Growth of Copper on Reconstructed Pt(100)
J. Phys. Chem. B, Vol. 108, No. 38, 2004 14665
Figure 3. Cu islands run parallel to the surrounding Pt(100)-hex channels which remain undisturbed. The atomic structure of the Cu islands is of quadratic nature (inset). Td ) 340 K, ΘCu ) 0.2 ML, 580 × 580 Å2, UB ) 400 mV, high pass filtered. Inset: 20 × 35 Å2, UB ) 22 mV.
by the orientation of the reconstruction rows in the respective domain. This alignment is generally encountered and becomes very clear in Figure 3 where the reconstruction rows of the uncovered substrate are visible just next and parallel to the Cu islands. Close to Cu islands the Pt reconstruction remains completely undisturbed even on the atomic scale. (This is in contrast to deposition at temperatures higher than ∼500 K where “darker” atoms appear on substitutional sites within the reconstructed layer next to the islands (not shown). These are assigned to Cu atoms incorporated into the surface). The brighter protrusions on the islands in Figure 3 provide evidence that nucleation in the second layer sets in long before completion of the first layer. We will return to this point later. The inset in Figure 3 shows a close-up of the atomic structure of the Cu islands. The Cu atoms exhibit a square (100) arrangement without any buckling. This proves that the underlying Pt substrate locally adopts a (1 × 1) structure upon Cu island formation. The long edges of the Cu islands are always located in the middle of the reconstruction troughs; that is, no reconstruction rows of partial width are ever observed. Hence, Cu induced lifting of the reconstruction proceeds only in terms of full reconstruction rows of 14 Å width (or multiples thereof) very much like with Au22 and Ag.23,24 The microscopic diffusion and nucleation processes leading to the described submonolayer island morphology appear to be relatively complicated, especially because nucleation of Cu clusters is accompanied by the lifting of the substrate reconstruction. Nevertheless, it is evident that the deposited Cu atoms mainly diffuse parallel to the reconstruction channels. As their coordination will be lower on top of the Pt ridges, the adatoms are presumably diffusing within the troughs. Given a sufficiently high adatom diffusivity, the first Cu atoms eventually become bound at a substrate step (or a domain boundary). Further adatoms approach and a small adatom island forms leading to a local lifting of the underlying reconstruction. The Cu island continues to grow anisotropically because the majority of additional Cu atoms arrives within the reconstruction rows antiparallel to the islands’ growing direction. However, upon closer inspection, the situation appears to be a little more
Figure 4. (a) Cu island exhibiting parallel bright lines with a height of 0.2 Å above the Cu layer. The white patches (marked by circles) represent second layer clusters. Td ) 340 K, ΘCu ) 0.5 ML, 115 × 115 Å2, UB ) 25 mV. (b) The lines are monatomically wide chains of Pt atoms built into the Cu islands. 55 × 55 Å2, UB ) 6 mV.
complicated than this. According to the above picture of island growth and given a spatially homogeneous deposition of Cu atoms, a rather uniform seam should develop at substrate steps with Cu islands in all troughs. Instead, isolated Cu “needles” are observed which are separated by completely uncovered reconstruction rows. This can only be understood if a nonvanishing diffusion perpendicular to the troughs is allowed, so that troughs can also be depopulated of Cu atoms which then attach to existing Cu islands in neighboring troughs, where they seem to diffuse easily toward the tips of these islands. Otherwise, no needlelike but broader structures would arise. As in the case of gas adsorption, the Cu induced lifting of the Pt(100) reconstruction produces excess atoms which are displaced from the less densely packed Pt(1 × 1) layer. These are incorporated into the growing Cu layer giving rise to the structures seen in Figure 4a which consist of parallel bright lines running along the preferential growth direction of the Cu islands,
14666 J. Phys. Chem. B, Vol. 108, No. 38, 2004
Schaefer et al.
Figure 5. Model visualizing Cu island growth and how Pt chains could possibly form within the first Cu layer.
Figure 6. Early nucleation and anisotropic growth in the second Cu layer. Td ) 340 K, ΘCu ∼ 1 ML, 600 × 600 Å2, UB ) 500 mV.
i.e., parallel to their long axis. These lines are imaged as being 0.2 Å higher and represent monatomically wide chains of Pt atoms (Figure 4b). Their linearity and striking parallel arrangement seem to be of kinetic origin; that is, the structures are not thermodynamically favorable ones but are caused by the microscopic dynamics during film growth. Otherwise, ∼50% of the Pt chains should point also into the (locally) crystallographically equivalent direction perpendicular to the long axis of the Cu islands, or substitutional alloy islands should form altogether. Figure 5 gives an idea of how these Pt chains possibly form. It shows a model of three neighboring reconstruction troughs separated by ridges that are, for the sake of simplicity, indicated as a (brighter) single atomic row. Cu atoms are diffusing in the troughs, and a Cu island is growing from the bottom. The Cu island atoms sit pseudomorphically on the locally unreconstructed Pt surface. The attachment of further Cu atoms to the island (right trough) leads to a continued lifting of the reconstruction and the displacement of excess Pt atoms onto the surface where they become incorporated into the growing Cu island. In Figure 5, it is assumed that only ridge atoms are expelled and built into the film. This has also been proposed for the system Au on Pt(100)-hex-R0.7° by Berg et al., who observed similar Pt chains.22 According to these authors, ridge atoms are less strongly bound than other substrate atoms within the trough and are thus more easily moved out of the surface. In that case, indeed, monatomically wide, almost linear Pt chains within the Cu islands are to be expected. However, they would also be roughly equidistant having a spacing of ∼14 Å corresponding to the separation of the original reconstruction ridges. The chains we observe are not always equidistant, though their average separation of 15 Å is close to the value of 14 Å. In fact, as suggested in Figure 5 ridge atoms may be displaced either to the left or to the right side of their original position depending on the direction they are “attacked” by an attaching Cu atom. As a consequence, the separation between Pt chains incorporated into the Cu islands may vary between four and six interatomic distances (see Figure 5). This is exactly what is seen in Figure 4b. It should be noted that the clusters of Pt excess atoms, which were seen for instance by Borg et al. after a CO induced lifting of the reconstruction,8 are also often
elongated and point into the direction parallel to the surrounding reconstruction troughs. A third system showing such a chainlike incorporation of substrate atoms into the growing film is Cu on the reconstructed Ir(100)-(5 × 1) surface. The (5 × 1) reconstruction is similar to the Pt(100) reconstruction in that again a hexagonal top layer sits on a quadratic substrate, also showing parallel channels on the surface. Gilarowski et al.30 found by STM that Cu grows again anisotropically and lifts the reconstruction more or less locally, with liberated surplus Ir atoms being embedded into the Cu film. 3.2. Second and Higher Layer Growth at 340 K. As was indicated before, the growth mode of the first two Cu layers on Pt(100) is not two dimensional (see Figure 3) in contrast to the results of earlier AES studies. Instead, small Cu clusters form in the second layer immediately after formation of monolayer islands. The same was observed for Au growth on Pt(100)-hexR0.7°.22 Figure 6 shows the film morphology after deposition of the equivalent amount of ∼1 ML. Thermodynamically, a wetting of the Pt surface by Cu is expected.31 However, as is seen in Figure 6, the second layer tends to grow anisotropically just like the first layer pointing toward nonthermodynamic conditions. This anisotropy is induced by the Pt chains in the first Cu layer: As is visible in Figure 4a, the chains act as heterogeneous nucleation centers for Cu adatoms diffusing on the first layer leading to adjacent cluster formation (marked by white circles). The existence of these nucleation centers is one of the reasons for the observed deviation from 2D growth. Additionally, there exists a large step edge barrier for Cu adatoms on Cu/Pt(100) islands as can be concluded from annealing experiments which caused second layer clusters to coalesce but did not remove them by making the second layer atoms diffuse down onto the substrate surface. Finally, the pseudomorphic nature of the Cu film is possibly a further cause for early nucleation in the second layer: According to theoretical and experimental studies,32,33 the diffusion length of adatoms is reduced on surfaces which are under tensile stress compared to relaxed ones. This can be understood qualitatively by a rougher diffusion potential since the surface atoms are further apart. The pseudomorphic first Cu layer is under tensile stress and thereby contributes to the short diffusion length for second layer atoms.
Growth of Copper on Reconstructed Pt(100)
Figure 7. (a) Local coverage ∼1.5 ML. Besides Pt chains in the first layer dark lines on the second layer can be seen. 230 × 230 Å2, UB ) 45 mV, high pass filtered. (b) Profile along the white line in (a).
Upon approaching a total coverage of ∼2 ML the first and second Cu layer are closed almost simultaneously. Only at this stage, the third layer begins to grow. This growth behavior shows that coverage calibration via AES cannot be trusted, as a possible first break in AES intensity plots will most likely correspond to a true coverage of 2 ML. Although such a break might actually occur, it will mislead to the assumption of a layerby-layer growth as in former AES studies.13,25,26 Third layer islands are again long and narrow which points toward a remaining anisotropy, now in or on top of the second layer. This remnant anisotropy in the second layer becomes clear from Figure 7a which corresponds to a local coverage of ∼1.5 ML. On the first Cu layer, the familiar bright Pt chains are found besides second layer clusters. In contrast, the second layer to the right shows characteristic parallel dark lines. These dark lines are assigned to “ditches”, as concluded from a model described below. These ditches are imaged by the STM tip as being not deeper than ∼0.2 Å as shown by the height profile in Figure 7b taken along the white line in Figure 7a.
J. Phys. Chem. B, Vol. 108, No. 38, 2004 14667
Figure 8. (a) Three ditches in the second Cu layer in near-atomic resolution. 29 × 29 Å2, UB ) 13 mV. (b) Model of the ditch structure. Top view. (c) Cross-sectional view.
A hint to the nature of these ditches is given by images such as the one shown in Figure 8a. Along the vertical direction, close-packed rows in the second Cu layer can be seen. Unfortunately, atomic resolution along this direction was not achievable. Three dark ditches are present in this image. They are formed by a local spreading of two neighboring close-packed Cu rows. The ball model in Figure 8b,c tries to account for this observation. In (b) the surface is seen from the top. The black atoms represent the unreconstructed Pt(100) substrate. The dark gray atoms in the next layer stand for the first pseudomorphic Cu layer into which a Pt chain is embedded in the center. As Pt atoms are larger than Cu atoms, the two Cu atom rows in the second layer adjacent to the Pt chain are locally shifted outward from their regular 4-fold hollow positions. Consequently, a gap in the second layer, i.e., a ditch, opens up above every Pt chain in the first layer (Figure 8c). This simple geometric model is in accordance with the observation that the average distances between Pt chains in the first layer and ditches in the second layer are equal (∼15Å). It is furthermore in accordance with a theoretical treatment of the system by
14668 J. Phys. Chem. B, Vol. 108, No. 38, 2004
Figure 9. (a) Ditches in the fourth monolayer. Td ) 340 K, ΘCu ) 4.4 ML, 45 × 60 Å2, UB ) 20 mV. (b) Cross-sectional view of the corresponding structure modelshowing why higher layer ditches are broader.
Wortmann et al.,34 whose calculations show that the positions of Cu atoms in the second layer on top of Pt chains do indeed relax, leading to a local buckling. However, even if relaxation is not allowed in the calculations, a corrugation of the LDOS above buried Pt chains results. So the ditches, as visible by STM, are due to electronic and geometric effects. The anisotropic growth of the third Cu layer is induced by these ditches, as they act again as diffusion barriers for adatoms in the third layer. This property is concluded from the fact that each third layer island is bordered by ditches along its long edges. The fact that ditches are diffusion barriers can also be understood with the help of the above arguments. Figure 8c presents a cross-sectional view: The displaced Cu atoms forming the spread rows are shifted not only sideways from the Pt chain but also a little upward. To cross a ditch, an adatom will then have to overcome these displaced atoms where it experiences a reduced coordination. This produces an energetic barrier for lateral diffusion across a ditch. Ditches are visible in even higher layers, where they continue to cause anisotropic Cu growth. The ditch structure, however, changes in that they widen in higher layers of the growing Cu film. Figure 9a presents two atomically resolved ditches in the fourth layer. Obviously, they do no longer consist of a single gap but of two to three lowered atomic rows. An extrapolation of the above purely geometrical ball model to thicker films is in accordance with this effect (Figure 9b).
Schaefer et al. 3.3. Thicker Copper Films Grown at 340 K. The Cu film grows pseudomorphically with respect to the local Pt(100)-(1 × 1) substrate as seen in the inset of Figure 3. As a consequence, it experiences a tensile strain due to the misfit of 8% between Pt(100) and Cu(100). Calculations by Vogtenhuber in fact favor this epitaxial (1 × 1) growth.31 Interestingly, our STM results indicate that this pseudomorphic growth persists even up to the ninth monolayer of Cu: Within the experimental error, the lattice constant (measured as the average interatomic separation along close-packed rows of at least 15 atoms) does not change when going from the first monolayer to thicker films. Normally, such thick pseudomorphic films with a comparable misfit to the substrate will accumulate a considerable strain energy. Therefore, after growth of a few pseudomorphic monolayers, an epitaxial film tends to relax toward its natural lattice constant in order to reduce the inherent strain. Our results were thus not to be expected and actually do contradict the findings of Radnik et al.,13 who observed a complete relaxation beyond two monolayers using LEED. A possible explanation for this discrepancy may be a wrong coverage calibration in this earlier work. Also Barnes et al.,25 reported on an increased LEED background for coverages above 3 ML which they associated with a possible relaxation of the Cu film. However, a higher background can also be explained by an increased film roughness; the latter one was actually revealed in our STM studies for coverages higher than ∼7 ML. Li et al.26 performed I-V-LEED measurements and favor the pseudomorphic growth of a strongly strained and tetragonally distorted fcc Cu film up to a coverage of at least 15 ML (room temperature). This supports our findings, and in fact, we neither observed any dislocations nor a moire´ pattern in our STM images of thick films, though the formation of one or the other should accompany the formation of a relaxed film. Instead, a reduction in the Cu island step heights is seen above ∼4 ML. Apart from the first layer, where the observed step height is influenced by electronic interactions between film and substrate, the first three to four layers exhibit step heights of (1.9 ( 0.1) Å (layer distance in the bulk Cu: 1.8 Å). The corresponding value at the surface of films thicker than ∼4 ML is only (1.6 ( 0.2) Å. This observation is in line with the existence of thick pseudomorphic films not necessarily experiencing large strain which was originally pointed out by Li et al.35 Earlier, Morrison et al.36 had given theoretical evidence for the existence of a body-centered-tetragonal (bct) Cu phase (a ) 2.76 Å, c ) 3.09 Å, d ) c/2 ) 1.55 Å; a being the lattice constant in the (001) plane, c the height of the bct cell, and d the interplanar spacing) and had suggested that bct Cu films could possibly be grown on a substrate with the same lattice constant 2.76 Å. Li et al.35 chose Pd(100) (a ) 2.75 Å) as a substrate for Cu deposition and were able to grow films with a ) 2.75 Å and d ) 1.62 Å once a thickness of more than 4-5 ML was reached. They assumed that these films had bct structure. Hahn et al.37 examined Cu growth on Pd(100) with STM. Their measurements also revealed a reduction of the step height in higher Cu layers. On the basis of the above-mentioned studies the authors inferred a transition from a pseudomorphic fcc film structure to a likewise pseudomorphic bct structure induced by the film strain energy. In our case of Cu/Pt(100), the reduced step height for films thicker than 4 ML also points to an fcc f bct transition of the growing Cu film. As a bct film exhibits an intralayer lattice constant of 2.76 Å,36 a very small misfit with respect to the Pt(100)-(1 × 1) substrate (2.77 Å) results, so that a thick pseudomorphic film may in fact exist. The calculated bct interlayer spacing of 1.55 Å36 is in very good agreement with
Growth of Copper on Reconstructed Pt(100) the measured step height of (1.6 ( 0.2) Å of our thick Cu/ Pt(100) films. 3.4. Growth at 650 K. The morphology of Cu films deposited at 650 K is, at submonolayer coverages, comparable to the one of 340 K films. Despite the high deposition temperature, anisotropic diffusion is still governing the film growth resulting in the formation of fingerlike islands. Heterogeneous nucleation now takes place only at ascending substrate steps, whereas in the case of growth at 340 K, also descending steps serve as nucleation centers. This shows that the binding of Cu adatoms to ascending steps is stronger, the binding energy at descending steps being too low to allow for binding at the higher deposition temperature. Similar to the growth at 340 K the second layer starts to grow already on the smallest first layer islands despite the higher thermal energy of adatoms at 650 K. This is again partly caused by line structures on the first Cu layer, which, however, now have a width and height of 13 Å and 0.9 Å, respectively, and reach lengths of up to 1000 Å (Figure 10). Similar structures develop upon annealing a film, that was deposited at 340 K, to 650 K. The orientation of these “hot lines”, as we will call them, is predominantly parallel to the long axis of an Cu island. However, a small percentage of lines are running perpendicular to this preferred direction, this percentage increasing as broader and broader Cu host islands exist. A very broad island is imaged in Figure 10a. The first layer is covered with “hot lines”, their majority running along the preferential growth direction of the island (horizontally). The island is actually crossing a domain boundary of the original Pt(100) reconstruction (indicated by the white line) which gives its upper left part a different preferential orientation. The second layer does not consist of very small clusters as in the case of 340 K films but has the form of nearly round islands which have obviously not nucleated at “hot lines”. This prevents the second layer from growing anisotropically as it does at 340 K. The “hot lines” nevertheless seem to be diffusion barriers for adatoms in the second Cu layer just like the Pt chains found at 340 K. In contrast to those, “hot lines” are not monatomically but 5-6 atomic rows wide as can be seen in Figure 10b. Though the atomic structure is not perfect, one can clearly identify its hexagonal symmetry. There is, indeed, a structural similarity to the rows of the Pt(100) reconstruction, and it is plausible to suppose that “hot lines”, just like the Pt chains at 340 K, consist of Pt excess atoms expelled from the substrate surface undergoing hexf(1 × 1) transformation and grouping together to form the “hot lines”. This interpretation is supported by a CO adsorption experiment. After a CO dosage of 30 L, small clusters appear on the Cufree Pt substrate as well as on the “hot lines” because the hexagonal atomic structure of both transformes into a square one, producing clusters of excess atoms. Therefore we conclude that the “hot lines” consist of Pt, also because their hexagonal atomic structure is energetically the most favorable for Pt on Pt(100)-(1 × 1). Why are “hot lines” broader than the Pt chains produced at a lower temperature? In case there is an attractive interaction between Pt atoms within the Cu layer, one could argue that the mobility of Pt atoms in the growing Cu film will be higher at 650 K compared to 340 K. So the excess atoms could diffuse more easily through the film aggregating to broader structures. However, this raises the question why no two dimensional Pt patches are formed instead of these long and 5-6 atomic rows wide lines. Additionally, Pt and Cu mix exothermally, so that no segregation should be expected, at least from bulk consid-
J. Phys. Chem. B, Vol. 108, No. 38, 2004 14669
Figure 10. (a) Upon deposition at 650 K bright line structures appear on the first Cu layer which are broader than those obtained at 340 K. ΘCu ) 0.4 ML, 1800 × 1800 Å2, UB ) 500 mV, high pass filtered. The white hatched line marks a domain boundary in the reconstruction of the substrate. (b) These lines consist of atoms in a hexagonal arrangement and resemble the Pt(100) construction rows. 580 × 580 Å2, UB ) 20 mV.
erations. The mechanism responsible for the formation of the “hot lines” remains unclear. It is interesting to note, however, that the width of the “hot lines” corresponds to the width of pure Pt reconstruction rows, which are a stable building element as suggested by the row-wise lifting of the reconstruction due to Cu deposition. Interestingly, the “hot lines” are not stable. Films of higher coverages than ∼1 ML do not show “hot lines” anymore. These thicker films were produced with the same deposition rate, and accordingly, an extended deposition time was used; that is, the film remained at the elevated temperature for a longer time. This obviously leads to a dissolution of the incorporated Pt structures. Films of a coverage of ∼2 ML do no longer show any dominant anisotropy but instead do consist of a Pt-Cu mixture because interdiffusion sets in above ∼500 K and becomes important at longer deposition times.
14670 J. Phys. Chem. B, Vol. 108, No. 38, 2004 4. Summary The growth of Cu on Pt(100)-hex has been studied in UHV using STM. Cu is found to lift the substrate reconstruction locally. This leads to the incorporation of Pt excess atoms into the anisotropically growing first Cu layer resulting in line structures the atomic structures of which depend on the deposition temperature, namely monatomic chains at 340 K but stripes 5-6 atom rows wide at 650 K. At 340 K, the first two layers nearly grow in a bilayer fashion, not layer-by-layer, and the Cu film grows pseudomorphically up to high coverages, whereas the growth remains anisotropic even in higher layers due to the embedded Pt lines. Cu films thicker than 4 ML exhibit bct structure. References and Notes (1) Hagstrom, S.; Lyon, H. B.; Somorjai, G. Phys. ReV. Lett. 1965, 15, 491. (2) Fedak, D. G.; Gjorstein, N. A. Surf. Sci. 1967, 8, 77. (3) Heilmann, P.; Heinz, K.; Mu¨ller, K. Surf. Sci 1979, 83, 487. (4) Van Hove, M. A.; Koestner, R. J.; Stair, P. C.; Biberian, J. P.; Kesmodel, L. L.; Bartos, I.; Somorjai, G. A. Surf. Sci. 1981, 103, 189. (5) Bickel, N.; Heinz, K. Surf. Sci. 1985, 163, 453. (6) Gibbs, D.; Ocko, B. M.; Zehner, D. M.; Mochrie, S. G. J. Phys. ReV. B 1990, 42, 7330. (7) Albernathy, D. L.; Mochrie, S. G. J.; Zehner, D. M.; Gru¨bel, G.; Gibbs, D. Phys. ReV. B 1992, 45, 9272. (8) Borg, A.; Hilmen, A.-M.; Bergene, E. Surf. Sci. 1994, 306, 10. (9) Fiorentini, V.; Methfessel, M.; Scheffler, M. Phys. ReV. Lett. 1993, 71, 1051. (10) Ritz, G.; Schmid, M.; Varga, P.; Borg, A.; Rønning, M. Phys. ReV. B 1997, 56, 10518. (11) Norton, P. R.; Davis, J. A.; Creber, D. K.; Sitter, C. W.; Jackman, T. E. Surf. Sci. 1981, 108, 205. (12) Chang, C. S.; Su, W. B.; Wei, C. M.; Tsong, T. T. Phys. ReV. Lett. 1999, 83, 2604. (13) Radnik, J.; Gitmans, F.; Pennemann, B.; Oster, K.; Wandelt, K. Surf. Sci. 1993, 287/288, 330.
Schaefer et al. (14) Pasteur, A. T.; Dixon-Warren, S. J.; King, D. A. J. Chem. Phys. 1995, 103, 2251. (15) Ritter, E.; Behm, R. J.; Po¨tschke, G.; Wintterlin, J. Surf. Sci. 1987, 181, 403. (16) Linderoth, T. R.; Mortensen, J. J.; Jacobsen, K. W.; Laegsgaard, E.; Stensgard, I.; Besenbacher, F. Phys. ReV. Lett. 1996, 77, 87. (17) Mortensen, J. J.; Linderoth, T. R.; Jacobsen, K.; Laegsgaard, E.; Stensgard, I.; Besenbacher, F. Surf. Sci. 1998, 400, 290. (18) Stru¨ber, U.; Kastner, A.; Ku¨ppers, J. Thin Solid Films 1994, 250, 101. (19) Attard, G. A.; Price, R. Surf. Sci. 1995, 335, 63. (20) Sachtler, J. W. A.; Van Hove, M. A.; Biberian, J. P.; Somorjai, G. A. Surf. Sci. 1981, 110, 19. (21) Venvik, H. J.; Berg, C.; Borg, A.; Raan, S. Phys. ReV. B 1996, 53, 16587. (22) Berg, C.; Venvik, H. J.; Strisland, F.; Ramstad, A.; Borg, A. Surf. Sci. 1998, 409, 1. (23) Batzill, M.; Koel, B. E. Surf. Sci. 2002, 498, L85. (24) Batzill, M.; Koel, B. E. Surf. Sci. 2004, 553, 50. (25) Barnes, C. J.; Lindroos, M.; Pessa, M. Surf. Sci. 1985, 152/153, 260. (26) Li, Y.; Quinn, J.; Li, H.; Tian, D.; Jona, F.; Marcus, P. M. Phys. ReV. B 1991, 44, 8261. (27) Wilms, M. diploma thesis, University of Bonn, 1994. (28) Schmidt, M. Ph.D. Thesis, University of Bonn, 1995. (29) Wilms, M.; Schmidt, M.; Bermes, G.; Wandelt, K. ReV. Sci. Instr. 1998, 69, 2696. (30) Gilarowski, G.; Me´ndez, J.; Niehus, H. Surf. Sci. 2000, 448, 290. (31) Vogtenhuber, D. Philos. Mag. B 1999, 79, 269. (32) Brune, H.; Bromann, K.; Ro¨der, H.; Kern, K. Phys. ReV. B 1995, 52, 14380. (33) Schroeder, M.; Wolf, D. E. Surf. Sci. 1997, 375, 129. (34) Wortmann, D.; Blu¨gel, S.; Bihlmayer, G. private communication. (35) Li, H.; Wu, S. C.; Tian, D.; Quinn, J.; Li, Y. S.; Jona, F.; Marcus, P. M. Phys. ReV. B 1989, 40, 5841. (36) Morrison, I. A.; Kang, M. H.; Mele, E. J. Phys. ReV. B 1989, 39, 1575. (37) Hahn, E.; Kampshoff, E.; Wa¨lchli, N.; Kern, K. Phys. ReV. Lett. 1995, 74, 1803.