This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 5007 Cite this: Phys. Chem. Chem. Phys., 2013, 15, 5007 Formation and decay of a compressed phase of 4,4 0 -biphenyldicarboxylic acid on Cu(001) Daniel Schwarz,*w Raoul van Gastel, Harold J. W. Zandvliet and Bene Poelsema The molecular arrangement of 4,4 0 -biphenyldicarboxylic acid (BDA) on Cu(001) has been studied at high coverage and relatively high temperature (B400 K) using Low Energy Electron Microscopy, LEEM, and selected area diffraction, mLEED. Next to the previously reported c(8 8) structure, we also observe a compressed phase with a 5 7 5 4 superstructure in matrix notation. All four equivalent (rotational and mirror) domains are equally populated. Both the c(8 8) and the compressed phase are confined to the first layer and the latter has a 14% higher density compared to the c(8 8) phase. Remarkably, this compressed phase is stable only during deposition and decays after interruption of the deposition. Apparently, the density of physisorbed admolecules on top of the c(8 8) layer has to be above a relevant threshold to allow the formation of the compressed phase. 1 Introduction The structure in which large molecules order on metal surfaces depends strongly on the interplay between intermolecular (intra- layer) energies, binding (interlayer) energies and entropy. 1–5 For example, molecules without any functional groups, which could mediate an attractive molecular interaction, often interact repulsively with each other and try to maximize their lateral distance. At low coverages this results in the formation of a disordered gas phase, while at high(er) coverages the molecules order without forming islands. 5–7 At high coverage, usually various compressed structures are formed with increasing density. On the other hand, molecules with functional groups that allow the formation of molecule–molecule bonds form ordered structures (islands) already at low coverage, due to an attractive interaction force. A typical example for the latter are benzoic-acids, where the functional carboxylic acid groups are actively involved in the formation of hydrogen-bonds between adsorbed molecules. In this case, the formation of large, well-ordered supramolecular networks is regularly observed. 8–17 In previous studies on molecular self assembly on surfaces it was often found that also the detailed structure formed by the molecules depends strongly on coverage. 18–22 With increasing coverage the molecules order in increasingly denser structures. By squeezing more molecules into the layer, the system gains total free energy, even though this decreases the average binding energy of the molecules, both to the substrate and to neighboring molecules. For example, Ye et al. 22 observed different structural phases of trimesic-acid on Au(111). At low coverage the molecules formed open molecular networks, which converted into more closely packed structures with increasing coverage. The molecules interacted less and less through the formation of energetically favorable dimeric hydrogen bonds, and formed rather cyclic hydrogen bonds in increasingly denser structures. Another recent example is a significant compression (20%) of the first layer formed by terephthalic acid (TPA) on Cu(001) at high coverages. 18 The structural change is enormous: instead of a parallel alignment of TPA molecules in the relaxed layer, neighboring molecules are perpendicular to each other in the compressed layer. Here, we present a combined low-energy electron micro- scopy (LEEM) and selected-area diffraction (mLEED) study on a compressed first layer phase of 4,4 0 -biphenyldicarboxylic acid (BDA) on Cu(001). LEEM allows us to follow the formation and the subsequent decay of the compressed phase in situ and at variable temperature. With these data, we propose a mecha- nism for the emergence of the compressed phase. The linear BDA molecule (length E 1.15 nm) is a simple molecule built out of two phenyl rings and two functional carboxylic acid groups, each at opposite ends of the molecule. The molecule is of high interest as a building block for the formation of supramolecular or metal-coordinated networks on metal surfaces. 16,17,23–28 On Cu(001) at or above room tempera- ture, BDA adsorbs in a (almost) flat lying geometry with the Physics of Interfaces and Nanomaterials, MESA + Institute for Nanotechnology, University of Twente, Enschede, Netherlands. E-mail: [email protected]; Fax: +31 53 489 1101; Tel: +31 53 489 3106 † Present address: Peter Gru ¨nberg Institut (PGI-3), Forschungszentrum Ju ¨lich, 52425 Ju ¨lich, Germany, and Ju ¨lich Aachen Research Alliance (JARA)-Fundamen- tale of Future Information Technology, 52425 Ju ¨lich, Germany. Received 7th January 2013, Accepted 6th February 2013 DOI: 10.1039/c3cp00049d www.rsc.org/pccp PCCP PAPER Downloaded by University of Illinois - Urbana on 02/05/2013 07:36:33. Published on 07 February 2013 on http://pubs.rsc.org | doi:10.1039/C3CP00049D View Article Online View Journal | View Issue
Transcript
This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 5007
Cite this: Phys. Chem.Chem.Phys.,2013,15, 5007
Formation and decay of a compressed phase of4,40-biphenyldicarboxylic acid on Cu(001)
Daniel Schwarz,*w Raoul van Gastel, Harold J. W. Zandvliet and Bene Poelsema
The molecular arrangement of 4,40-biphenyldicarboxylic acid (BDA) on Cu(001) has been studied at
high coverage and relatively high temperature (B400 K) using Low Energy Electron Microscopy, LEEM,
and selected area diffraction, mLEED. Next to the previously reported c(8 � 8) structure, we also observe
a compressed phase with a5 �75 4
� �superstructure in matrix notation. All four equivalent (rotational
and mirror) domains are equally populated. Both the c(8 � 8) and the compressed phase are confined
to the first layer and the latter has a 14% higher density compared to the c(8 � 8) phase. Remarkably,
this compressed phase is stable only during deposition and decays after interruption of the deposition.
Apparently, the density of physisorbed admolecules on top of the c(8 � 8) layer has to be above a
relevant threshold to allow the formation of the compressed phase.
1 Introduction
The structure in which large molecules order on metal surfacesdepends strongly on the interplay between intermolecular (intra-layer) energies, binding (interlayer) energies and entropy.1–5 Forexample, molecules without any functional groups, which couldmediate an attractive molecular interaction, often interactrepulsively with each other and try to maximize their lateraldistance. At low coverages this results in the formation of adisordered gas phase, while at high(er) coverages the moleculesorder without forming islands.5–7 At high coverage, usually variouscompressed structures are formed with increasing density. On theother hand, molecules with functional groups that allow theformation of molecule–molecule bonds form ordered structures(islands) already at low coverage, due to an attractive interactionforce. A typical example for the latter are benzoic-acids, wherethe functional carboxylic acid groups are actively involved in theformation of hydrogen-bonds between adsorbed molecules. Inthis case, the formation of large, well-ordered supramolecularnetworks is regularly observed.8–17
In previous studies on molecular self assembly on surfaces itwas often found that also the detailed structure formed by themolecules depends strongly on coverage.18–22 With increasingcoverage the molecules order in increasingly denser structures.
By squeezing more molecules into the layer, the system gainstotal free energy, even though this decreases the averagebinding energy of the molecules, both to the substrate and toneighboring molecules. For example, Ye et al.22 observeddifferent structural phases of trimesic-acid on Au(111). At lowcoverage the molecules formed open molecular networks,which converted into more closely packed structures withincreasing coverage. The molecules interacted less and lessthrough the formation of energetically favorable dimerichydrogen bonds, and formed rather cyclic hydrogen bonds inincreasingly denser structures. Another recent example is asignificant compression (20%) of the first layer formed byterephthalic acid (TPA) on Cu(001) at high coverages.18 Thestructural change is enormous: instead of a parallel alignmentof TPA molecules in the relaxed layer, neighboring moleculesare perpendicular to each other in the compressed layer.
Here, we present a combined low-energy electron micro-scopy (LEEM) and selected-area diffraction (mLEED) study on acompressed first layer phase of 4,40-biphenyldicarboxylic acid(BDA) on Cu(001). LEEM allows us to follow the formation andthe subsequent decay of the compressed phase in situ and atvariable temperature. With these data, we propose a mecha-nism for the emergence of the compressed phase.
The linear BDA molecule (length E 1.15 nm) is a simplemolecule built out of two phenyl rings and two functionalcarboxylic acid groups, each at opposite ends of the molecule.The molecule is of high interest as a building block for theformation of supramolecular or metal-coordinated networks onmetal surfaces.16,17,23–28 On Cu(001) at or above room tempera-ture, BDA adsorbs in a (almost) flat lying geometry with the
Physics of Interfaces and Nanomaterials, MESA + Institute for Nanotechnology,
University of Twente, Enschede, Netherlands. E-mail: [email protected];
Fax: +31 53 489 1101; Tel: +31 53 489 3106
† Present address: Peter Grunberg Institut (PGI-3), Forschungszentrum Julich,52425 Julich, Germany, and Julich Aachen Research Alliance (JARA)-Fundamen-tale of Future Information Technology, 52425 Julich, Germany.
Received 7th January 2013,Accepted 6th February 2013
5008 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 This journal is c the Owner Societies 2013
carboxylic acid groups deprotonated.9,29,30 At low BDA coverage,the molecules form a 2D dilute phase with no long range order.If this dilute phase reaches a critical (temperature dependent)density, molecules start to form large supramolecular networks,which can extend over entire Cu terraces.17,23,28,31 In thesenetworks, the molecules form a c(8 � 8) superstructure on theCu(001) lattice, where one molecule occupies 16 Cu(001) unitcells (1.04 nm2). Neighboring molecules in the structure arerotated by 901 and interact through the formation of hydrogenbonds between the deprotonated carboxyl groups and the phenylrings of the adjacent molecules. We prefer to comply with thisnon-primitive notation, in accordance with the main body ofliterature on BDA/Cu(001). Moreover the c(8 � 8) cell allows adirect comparison also with the 4 molecules containing unitcell of the compressed phase (see below). In a previous study,we determined the 2D cohesive energy of this network to beEC = 0.35 eV.17 At the substrate temperatures employed in thisstudy, ranging from 370 K to 420 K, the equilibrium density ofmolecules in the dilute phase is between 10% and 30% of thedensity in the c(8 � 8) structure. The molecular networkscannot grow across Cu steps, due to registry shift on the fcc(001) surface.44 In contrast to this, steps are permeablefor individual molecules under the experimental conditionshere.17 For a wide range of BDA coverage and temperatures, thec(8 � 8) structure is the only solid phase we observe.
Here, we show that at high coverages and high substrate
temperatures a compressed phase emerges with a5 �75 4
� �
superstructure. The compression starts by a conversion of thec(8 � 8) structure in the center of large domains duringdeposition of BDA, i.e., the coverage is locally sufficientlyincreased to favor the formation of the compressed structure.The compressed phase is not stable, it transforms slowly backinto the c(8� 8) phase upon interruption of the deposition. Ourresults are closely related to the recent publication on first layercompression of TPA on Cu(001) from Tait et al.18 The TPAmolecule is the shorter equivalent to BDA with one instead oftwo phenyl rings. Still, the relaxed TPA layer is very differentfrom the relaxed c(8 � 8) BDA layer. On the other hand, thestructures of the compressed TPA and that of the here proposedcompressed BDA layers are very similar. This is showingthe delicate interplay between the different energies: at lowcoverage the adsorption geometry of individual molecules isoptimized, while at higher coverages a close packing is favored.The latter results in an increase of the total number of mole-cules in contact with the substrate.
2 Experimental
Experiments were performed using an Elmitec LEEM III micro-scope32 with a base pressure of about 1 � 10�10 mbar. ACu(001) single crystal with a miscut angle less than 0.11 wasused.33 Prior to mounting in LEEM, the crystal was annealed at1170 K in an Ar–H2 mixture for a prolonged period to depletethe bulk of the crystal from sulphur contamination. After
insertion into the microscope, the sample surface was furthercleaned by cycles of sputtering with hydrogen,34 argon, andannealing at 900 K. BDA in powder form (purity >0.97, TCIEurope, CAS: 787-70-2) was deposited from a Knudsen cell.Deposition rates between 2 and 5 � 10�5 monolayers (ML)per second were used, where a monolayer is defined as oneBDA molecule per Cu(001) unit cell. With this definition, thec(8 � 8) structure completely covers the surface at 0.0625 ML.LEEM images were background corrected by applying a flatfield correction. No indications for beam induced damage tothe molecular networks were observed, e.g., at the edge of theilluminated surface area. We inserted illumination apertures inthe incoming electron beam path for selected area diffraction(mLEED). Apertures with diameters of either 1.4 mm or 25 mmwere used to spatially resolve the surface structure. The mLEEDpatterns were corrected for a cloud formed by secondary electronswith a procedure described in a previous publication.17 Duringdeposition, the sample temperature was kept at the valuesmentioned in the text, always well below the temperature of450 K, at which we start to observe degradation of the films,presumably due to substrate assisted decomposition.
3 Results and discussion3.1 Formation of the compressed phase
We start by describing an experiment in which we have grown aclosed layer of the c(8 � 8) phase at a substrate temperature of
Fig. 1 LEEM images showing the completion of the first BDA layer on Cu(001)during growth at T = 420 K. Cu terraces appear light gray and c(8 � 8) BDA islandsdark. Thin, curved lines in (a) and (b) are Cu steps, thicker, curved lines step bunches.BDA coverage is increasing from (a)–(d). The arrow in (c) shows the formation of asecond phase within the existing BDA islands upon closing of the c(8 � 8) layer. Thenew phase appears lighter than the surrounding island under the given imagingconditions. The start voltage was 2 eV for (a)–(c) and 2.3 eV for (d). The dashed circlemarks a defect on the micro-channel plate detector which appears bright. Imagecontrast was adjusted for each image. Field of view (FoV) is 10 mm.
This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 5009
420 K with an average BDA deposition flux of 4 � 10�5 ML s�1.A LEEM image sequence of the completion of the c(8 � 8) BDAlayer is presented in Fig. 1. The light gray areas in Fig. 1(a) and(b) are Cu terraces covered with a BDA dilute phase17 and thedark areas are BDA c(8 � 8) islands, which was verified withmLEED. The thin curved lines are Cu steps and thicker curves arestep bunches. In the first two images (Fig. 1(a) and (b)), we cansee how the first BDA domains get blocked in their growth by Custeps, and the islands follow the shape of the underlying Cuterraces.44 In the third image (Fig. 1(c)) almost the entire surfaceis covered with BDA c(8 � 8) domains. The arrow in Fig. 1(c)marks a triangular area, where a light contrast developed in oneof the larger domains. The contrast first appeared in the centerof that domain, whose expansion was already terminated bysteps earlier in the growth process. The shape of the lighter arearoughly mimics the shape of the surrounding domain. In the lastimage (Fig. 1(d)) the c(8 � 8) layer is closed and all of the largerdomains developed the light contrast in the center. The lighterareas are growing in size with further BDA deposition. It is thusevident that the lighter areas represent either a second layeron-top of the c(8 � 8) layer, or a compressed surface confinedphase that formed due to the pressure exerted by admoleculesadsorbed on-top of the domains. Our further findings strongly hintat a compressed phase. For clearness, we will thus denote the newlyemerging structure as the ‘‘compressed’’ phase in the following.
The contrast between the compressed and the regular c(8 � 8)areas depends strongly on electron energy (i.e., start voltage). InFig. 2, we analyzed the electron reflectivity of the two phases as afunction of start voltage (LEEM I–V). At first sight, the two curvesfall almost perfectly on top of each other. This is consistent withthe presence of a compressed phase. For an emerging secondlayer we would expect clear differences with respect to the c(8� 8)phase spectrum. The latter is nicely illustrated for the growth ofsexiphenyl (6P) on graphene.35 Even though the crystallinestructures of the first, second and third monolayers of 6P arealmost identical, the different layers can be discerned easily. It isimportant to note that we actually never observed the nucleationof an ordered second layer both with mLEED and LEEM.
Minute differences between the curves give rise to the observedcontrast in Fig. 1. This is more clear in the plot of the differencebetween both LEEM I–V curves (green dashed line in Fig. 2). Thedifference in electron reflectivity of both phases is largest for startvoltages of 0.8 V, 2.4 V and 4.8 V. While the compressed phaseappears dark for 0.8 V and 4.8 V, the contrast is inverted for 2.4 V.Exemplary LEEM snapshots taken at the three respective startvoltages are shown in the bottom part of Fig. 2.
3.2 lLEED structure analysis
After stopping the deposition of BDA molecules, we analyzed theemerging compressed phase using mLEED. We restricted theelectron beam to an area of 1.4 mm diameter on the surface bymeans of an illumination aperture. When we selected the darkerareas of the layer in Fig. 1(d) with the aperture, we observed theregular c(8� 8) superstructure without any additional superspots.28
An example of this mLEED pattern is presented in Fig. 3. However, ifwe selected one of the lighter areas (i.e., the compressed phase), weobserved a new, highly complex pattern, which does not develop atlower coverages and neither at temperatures below B350 K. InFig. 3(a) we present a (background corrected) cumulative mLEEDpattern, which we obtained by summing up individual patterns forelectron energies from 5 eV to 36 eV at 1 eV spacing. The pattern isfour-fold symmetric and the underlying structure is commensurate,like the c(8 � 8) structure. This can be seen by the repetition of thepattern around each of the four first order Cu(001) substrate spots.While the pattern appears highly complex at first sight, it can benicely described by the superposition of four equivalent domains.
One of the domains can be described with a5 �75 4
� �super-
structure, with the [110] and [1%10] substrate directions as a base. Acalculation of the corresponding pattern is shown in Fig. 3(b) and(d), with the spots belonging to each of the four rotational domainsin the same color (gray tone). Fig. 3(c) and (d) show the inner partof the pattern at a higher magnification. Additionally, we super-imposed the expected position of spots on the diffraction pattern inFig. 3(c). All spots that carry intensity were traced back in thecalculated diffraction pattern.
The diffraction pattern is quite complex and since weobserve the superposition of four equivalent domains, alreadya slightly different overlayer structure would produce a verydifferent pattern. Interestingly, we observed no intensity at the
Fig. 2 Image intensity versus start voltage (LEEM I–V) plots obtained fromdifferent areas: c(8 � 8) (solid), compressed phase (long dashed), and cleanCu(001) (short dashed). The difference between the first two is plotted as thedotted line at two times magnification. The LEEM images on the bottom showthe surface at start voltages for which the difference is maximal (FoV 4 mm). For0.8 eV and 4.8 eV the compressed phase is darker, and for 2.4 eV lighter than thesurrounding c(8 � 8) phase. The white dashed line in the first snapshot outlinesan example of the compressed phase.
5010 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 This journal is c the Owner Societies 2013
expected positions of the c(8 � 8) structure. Together with thesimilar LEEM I–V characteristics (cf. Fig. 2), this is strong prooffor another surface confined layer phase and not a second layeron-top of the c(8� 8) domains. Since this phase forms at higherBDA coverages, molecules are packed in a denser way in the
emerging structure. The5 �75 4
� �unit cell covers an area of
55 Cu(001) unit cells, which is 14% smaller than the unit cell ofthe c(8 � 8) structure.
In the c(8 � 8) structure, the BDA molecules form only onerotational domain on the surface (though, there are 32 differenttranslational domains). This is because the molecules in thisstructure are aligned along a high symmetry direction of the
substrate lattice. In the5 �75 4
� �structure, this symmetry is
lifted. The molecules form four equivalent domains, of whichtwo, respectively, have the same handedness, i.e., the domainscan be translated into each other by a rotation. In other words,the achiral c(8 � 8) structure is converted into a chiral structureupon increasing coverage. This is a well known effect andcaused by the close packing of molecules, which induces theformation of chiral structures.36 The four equivalent structuresare equally occupied: the intensities of the respective equivalent
diffraction spots are identical, even when using the smallestillumination aperture of 1.4 mm diameter available on theinstrument. The individual domains are very small and cannotbe resolved here, neither in mLEED nor LEEM.
To clarify the question, whether the conversion into thecompressed phase occurs through a gradual compression of thec(8 � 8) structure or through the nucleation of the compressedphase, we also followed the conversion in situ with mLEED. Fig. 4shows the development of the diffraction pattern with increasingBDA coverage at an electron energy of 6 eV and at a substratetemperature of 400 K. The patterns were obtained with a largeillumination aperture that limited the beam to an area with 25 mmdiameter on the surface (compared to 1.4 mm in Fig. 3). In the firstpattern of the sequence, the only ordered structure observable onthe surface was the c(8 � 8) phase. As we started to deposit moreBDA, the coverage increased, and the spots which belong to thecompressed phase appeared and gained intensity. At the sametime, the c(8 � 8) diffraction spots lost intensity, until theycompletely vanished for the highest coverages.
The lower part of Fig. 4 shows line scans through thediffraction pattern as a function of BDA coverage. Fig. 4(a) showsline scans through the (1=4,0) and (0,0) c(8 � 8) spots, Fig. 4(b)
along the same direction but through the (1=4,1=4) c(8 � 8) spot.
Fig. 3 mLEED analysis of Fig. 1(d). The darker areas produce the c(8 � 8) pattern (bottom left), while the lighter (compressed) areas produce a new diffraction pattern(a–d). (a) The cumulative (background corrected) mLEED pattern obtained from the compressed phase by summing up individual patterns obtained for electronenergies from 5 eV to 36 eV at 1 eV spacing. The complex diffraction pattern can be described with a superposition of four equivalent domains, of which one is a
5 �75 4
� �overlayer in matrix notation. A calculated pattern is shown in (b). Each color (gray tone) represents one of the four domains and crosses mark spots
belonging to the Cu(001) substrate. (c) and (d) show the inner part of (a) and (b) enlarged. The calculated positions of spots are superimposed in (c) as black circles.Without exception we found experimentally finite intensity at these positions, even when sometimes very weak. Black arrows indicate the unit vectors of the
5 �75 4
� �superstructure, red (gray) squares in (c) and (d) are guides to the eye.
This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 5011
In Fig. 4(a), the center of the (1=4,0) spot seems to move con-tinuously with increasing coverage to a position that belongs to thepattern of the compressed phase. This would point at a gradualcompression of the c(8 � 8) overlayer with increasing coverage.However, this is actually a coincidental effect, since both theoriginal and the emerging spot(s) are close to each other and
overlap partially. In the line profile through the (1=4,1=4) spot
(Fig. 4(b)), we can see that the (1=4,1=4) c(8� 8) spot is not movingat all as it disappears. The closest spots belonging to the com-pressed phase appear simultaneously, also at a fixed position.Thus, both structures form distinct phases and the conversion isnot proceeding through the continuous compression of the c(8� 8)phase, but rather through the nucleation of the compressed phasewithin the c(8 � 8) domains.
3.3 Molecular structure in the compressed phase
We will now discuss the possible arrangement of the BDAmolecules in the compressed structure. From the mLEED
patterns we found a5 �75 4
� �overlayer structure (cf. Fig. 3).
Compared to the c(8 � 8) structure, the unit cell of thisoverlayer is both rotated and compressed (see Fig. 5(a)). Thecompressed unit cell covers 55 Cu lattice cells, compared to 64for the c(8 � 8). In the c(8 � 8) unit cell one molecule occupies16 Cu(001) unit cells (1.04 nm2). Assuming that all molecules
adsorb in a flat lying geometry, which is expected given theanticipated strong interaction of the deprotonated carboxylategroups with the substrate, four molecules fit into the compressedunit cell. This corresponds to a 14% compression, i.e., 13.75Cu(001) unit cells (0.89 nm2) per molecule.
The four molecules can fit into the compressed unit cell inseveral different ways and orientations. While in principle, we
Fig. 4 Top: conversion of the c(8 � 8) phase into the compressed phase with increasing BDA coverage as seen in mLEED at an electron energy of 6 eV and atemperature of 400 K. At first (leftmost pattern), only the c(8 � 8) phase is visible, which completely vanished upon conversion in the rightmost image at higher BDAcoverage. The dashed rectangles mark the two areas used for line scans in the lower part of the figure. (a) Shows line profile versus coverage plots through the (0,0)and (1=4,0) spots (white dashed box), (b) parallel to the same direction through the (1=4,1=4) – c(8 � 8) spot (pink (gray) dashed box). The (relative) intensity color codeis explained on the right hand side.
Fig. 5 (a) Sketch of the uncompressed c(8 � 8) BDA structure. The red (lightgray) dashed square shows the c(8 � 8) unit cell, the blue (dark gray) dashed
rectangle the5 �75 4
� �unit cell. (b) Tentative sketch of the compressed
structure: similar to the c(8 � 8) structure, but with the molecules slightly rotatedand at smaller distances. The corresponding molecules in both structures arelabeled A–E. The molecular interaction is thought to be similar in both structures.
5012 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 This journal is c the Owner Societies 2013
should be able to get more information on the orientation of themolecules from the mLEED intensity distribution, this analysis iscomplicated here possibly also by the coherent superposition offour equivalent domains (see Fig. 3). The intensity distributionis to a large degree determined by the addition of intensitiesfrom different rotational domains. Equivalent spots from allfour domains have identical intensity, even with the smallestillumination aperture of 1.4 mm diameter. Thus, the compressedareas consist of several small equivalent domains, with noapparent preference for a specific orientation.
Without microscopic information on a molecular scale, wecan only speculate how the molecules are oriented in the unitcell. A possible and reasonable structure is shown in Fig. 5(b).The four molecules in the unit cell are not at identical positionson the Cu(001) lattice, so a reduction to a smaller unit cell is notpossible. The compressed structure is commensurate withhigher order, while the c(8 � 8) structure is commensuratewith first order, i.e., all molecules are adsorbed on identicalCu(001) lattice sites.
The proposed structure is very similar to the c(8 � 8)structure. Thus, the conversion between both phases onlyinvolves a minimal rotation and movement of molecules. Thebasic building blocks, two molecules rotated now by approxi-mately 901, are the same in both the relaxed and compressed
structures. Also, the carboxylate groups are still pointingtowards the slightly positively polarized hydrogen atoms ofthe phenyl rings of the neighboring molecules, which allowsfor a remaining mutual interaction between molecules. Ourmodel is also consistent with the compressed TPA layer onCu(001).18
Irrespective of the actual structure in the compressed phase,the higher order commensurability will reduce the averagebinding energy per molecule in the compressed, surface con-fined layer compared to the c(8 � 8) layer. However, thedensification still reduces the total free energy, due to the increasednumber of molecules that can make contact to the substrate, whichensures a stronger binding than in the physisorbed state on top ofthe c(8 � 8) layer.
3.4 Dynamics of the phase conversion
Without an incoming flux of BDA molecules, the compressedphase is not stable and slowly converts back to the c(8 � 8)structure. The formation and subsequent decay of the com-pressed phase is shown in a clear way in the image sequencepresented in Fig. 6. The experiment was performed at a sub-strate temperature of 378 K and a start voltage of 0.8 V, forwhich the compressed phase appears darker than the sur-rounding c(8 � 8) phase. The first image in Fig. 6 (t = 0 s)
Fig. 6 LEEM image sequence (FoV 4 mm, start voltage 0.8 V) illustrating the formation and decay of the compressed phase. We deposited additional BDA on a surfacecovered with large c(8 � 8) domains. The temperature was 378 K. The initial BDA deposition was stopped approximately 1000 s before the first image. White areas areCu terraces with the dilute phase, gray areas are the c(8 � 8) structure and the dark gray areas represent the compressed phase. The arrow in the first image showsremainders of the compressed phase from the initial growth. The compressed area for this island is additionally marked with a white dashed line in the sequence,where clearly distinguishable. The shutter of the BDA source was opened at t = 0 s and closed at t = 60 s, with an estimated deposition rate of 5 � 10�5 ML s�1.Immediately after opening the shutter, the contrast of the islands turns dark, the compressed areas increase and several new BDA domains nucleate on the Cu terrace.After closing the shutter the contrast on the islands slowly lightens, starting from the outside rim of the domains until the end the compressed areas almost completelyvanished again.
This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 5013
shows the surface after deposition of an initial amount of BDA(B0.5 BDA molecules per nm2) and waiting for >1000 s with theshutter of the BDA source closed. At this point, slightly lessthan 40% of the surface is covered by solid c(8 � 8) domains.Some of the domains show the compressed phase in the center(see e.g. island indicated by the arrow). At t = 0 s we openedthe shutter again for 60 s and molecules are deposited atan estimated rate of 5 � 10�5 ML s�1. The images directlyafter opening the shutter show two effects: several new, smallc(8 � 8) domains nucleated within the dilute phase on the Cuterraces due to the increased supersaturation. At the same time,the contrast of the c(8 � 8) islands turned dark, and thecompressed areas grew, which appear dark for the start voltageused here (cf. Fig. 2).
After closing the shutter (t = 60 s), the contrast increasedagain, beginning from the rim of the domains. At t = 160 s wecan clearly distinguish the two different phases again: thecompressed phase is in the center of larger islands, surroundedby a rim of the c(8 � 8) phase. The interface between bothphases appears sharp – no contrast gradients are resolved. Theshape of the compressed phase follows the shape of the island,with the distance to the outside border of the surroundingisland being almost constant. The compressed phase comple-tely decayed after about 1500 s. Several of the smaller c(8 � 8)islands in the vicinity grew at the same time. Also the largerislands grew, though, only where they were not bordered byunderlying Cu steps. The material from the compressed phaseis apparently incorporated into the c(8 � 8) domains in thesurrounding areas. We also observed indications for materialdesorbing back into vacuum, however, with desorption rateswhich were very low (a small fraction of a monolayer per hour).
An obvious next step is a comparison of the c(8 � 8) domainsize increase relative to the area initially occupied by the com-pressed phase. We did so for several different experiments andfound densities varying between 1.1 and 2 times the c(8 � 8)density for the compressed structure. These numbers roughlycorrespond to the expected 14% difference. Unfortunately, a moreprecise determination of the density is not feasible in this way. Theproblem with this analysis are the many very small BDA domainsthat nucleated on smaller Cu terraces and whose areal increase ishard to quantify with LEEM. The latter is due to electron lensingeffects at the edge of the BDA domains caused by the workfunction differences between the Cu substrate and BDA.37
The conversion (or decay) of the compressed phase is furtheranalyzed in Fig. 7. In this figure, we plotted the area of thecompressed phase versus time after stopping BDA depositionfor five different domains in the same experiment. Thesubstrate temperature was kept constant at 378 K. The indivi-dual decay curves were shifted in time for maximal overlap; thetime scale is thus not directly related to the closing of the BDAsource shutter. Even though the five BDA domains have quitedifferent shapes (see insets in Fig. 7) and the compressed areashave different relations to the domain boundaries, all five decaycurves follow roughly the same trend. The decay curves deviatefrom a linear decrease: the conversion slows down for smallerremaining compressed areas. This is pointing to one of two
things: either the conversion is limited by the area, or by theperimeter of the compressed phase. In the first case, we wouldexpect an exponential decay of the area:
dAðtÞdt¼ �C � AðtÞ (1)
A(t) = A(t0) � exp(�Ct), (2)
where A(t) is the area of the compressed phase at time t and C isa parameter. In the case of decay limited by the length of theinterface between compressed and c(8 � 8) area, and assumingfor simplicity a circular geometry, we would expect a decaycurve following a second order polynomial:
dAðtÞdt¼ �C�
ffiffiffiffiffiffiffiffiffiAðtÞ
p(3)
AðtÞ ¼ A t0ð Þ � 2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA t0ð Þð Þ
pC�tþ C�tð Þ2; (4)
where C* is a parameter with dimensionm
s. We fitted the decay
curves with both models in Fig. 7. The black short dashed lineis an exponential fit using eqn (2), the red long dashed line apolynomial fit using eqn (4). The polynomial model gives theslightly better fit. The adjusted R2 of the polynomial fit is 0.989,versus 0.958 for the exponential fit. Additionally, the plot of theresiduals for the exponential fit shows a systematic deviation(see Fig. 7(b)). Therefore, we conclude that the decay is limitedby the perimeter length of the compressed areas.
Fig. 7 (a) Area of the compressed phase embedded in five different BDA islandsversus time at T = 378 K and after stopping BDA deposition. The curves are shiftedin time for maximum overlap. The insets show the five islands that were analyzedand the dashed lines therein mark the respective area of the compressed phase.Colors (gray tones) correspond to the colors (gray tones) of the data points. Thedifferent decay curves are identical within the uncertainty of the size determina-tion. An estimate for the error bar is shown for two times. The black short dashedline is an exponential fit (eqn (2)), the red (gray) dashed line a polynomial fit ofsecond order (eqn (4)). (b) Plot of the residuals for the two fit curves (black datapoints exponential fit, red (gray) data points polynomial fit).
5014 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 This journal is c the Owner Societies 2013
The concave shape of the decay curves in Fig. 7 is actuallyquite unusual. The decay of 2D clusters typically involvesconvex shaped decay curves, i.e., accelerated decay for smallercluster sizes, or no curvature at all.38–43 The first case isobserved when the decay process is limited by the diffusionof entities away from the edge of the island (diffusion limiteddecay). Also the decay of BDA c(8 � 8) islands follows thisbehavior.17,28 A linear decay curve is expected for decay that islimited by the detachment rate of entities from the island edge(interface limited decay). In both cases, the driving forcebehind the decay is a reduction of the free energy associatedwith the interface. The decay dynamics arise from the increasingfree energy cost for smaller clusters, i.e., the Gibbs–Thomsoneffect. Here, this is apparently not the case; the decay rate isproportional to the perimeter length irrespective of its curvature.
The decay rate of the compressed areas depends strongly onthe substrate temperature: the higher the temperature, thefaster the decay process. Fig. 8(a) shows decay curves obtainedfor three different substrate temperatures: 378 K, 398 K and420 K. The respective BDA islands at the different temperatures areshown as insets in the same figure. An Arrhenius plot of the threeinitial decay rates at A(t0) = 2 � 105 nm2 is presented in Fig. 8(b).
The three data points lie nicely on a straight line. From theslope we find an activation energy of EA = 0.63 eV for theprocess that is limiting the decay. From the interception ofthe fit with the ordinate, we find a prefactor of 5 � 1010 nm2 s�1.To determine the frequency prefactor n0 of the decay limitingprocess, we need to take into account our choice of A(t0). Thelatter comes into play through the number of molecules alongthe perimeter of the compressed areas. For A(t0) = 2 � 105 nm2,this corresponds to B1600 molecules. Therefore, the frequencyprefactor is n0 = 3 � 107 s�1.
Despite the good fit in Fig. 8(b), the error bar attached tothe activation energy is probably quite large, but difficult toestimate here. The three islands in Fig. 8(a) have differentgeometries and relations to the compressed areas. Preparingan experiment with exactly the same, or even only similar,geometry is almost impossible due to the fact that the overallshape is determined by the shape of the hosting Cu(001)terraces. Given the fact that in this particular case the mostcompact form is found for the high temperature data a slightlylower activation energy might be more realistic.
3.5 Transformation mechanisms
We now address possible mechanisms behind the formationand decay of the compressed phase. To summarize our find-ings: the compressed phase grows in the center of islands,whose expansion is terminated by the steps bordering the hostCu(001) terrace. The shape follows the shape of the c(8 � 8)islands, i.e., the host Cu terrace, and the rims maintain theirc(8 � 8) structure.44 The formation of the compressed phaseonly occurs while additional BDA is deposited from vacuumand at elevated substrate temperatures (E>350 K). Without anincident flux, the compressed phase is unstable and decays.The decay of the compressed phase proceeds from the divisionline between the compressed and the c(8 � 8) islands. Thecompressed phase does not emerge in small c(8 � 8) domainsand is not observed at the lower end of our experimentaltemperature window below about 350 K.17,28,31 For instance,after BDA deposition far beyond a macroscopically closedc(8 � 8) layer at room temperature, we did not observe theemergence of the compressed structure or any other orderedstructure (e.g. a second ordered BDA layer) both with mLEED orLEEM. We expect that the molecules form a 2D gas phase ontop of the c(8 � 8) domains, but with a density that was too lowto access with LEEM. The latter observations hint at theinvolvement of a thermally activated compression process.Unfortunately, we have no access to processes at the molecularscale and, therefore, we can only provide the most plausiblecourse of events for formation and decay of the compressedareas based on the information presented before.
3.5.1 Formation mechanism. Our basic hypothesis is thatthe formation of the compressed phase requires a local densityof physisorbed admolecules on-top of the c(8 � 8) domainsbeyond a certain threshold value. The emerging admoleculedensity profile will have its maximum at the center of thec(8 � 8) domains, since admolecules will descend from thedomain at its edge subject to a finite Ehrlich–Schwoebel barrier.
Fig. 8 (a) Decay curves of the compressed phase at three different substratetemperatures (cf. those in Fig. 7). LEEM images of the respective BDA domains areshown as insets at the same scale and with the compressed phase indicated with whitedashed lines. The image at 420 K was recorded with a start voltage of 2.4 V, hence thereversed contrast. (b) The Arrhenius plot of the initial slope at A(t0) = 2 � 105 nm2 inarea per second of the three curves in (a) (solid lines). From the slope, we find anactivation energy for the decay limiting process of EA = 0.63 eV.
This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 5015
These molecules will become part of the surrounding dilutephase and finally become aggregated in the edges of stillgrowing c(8 � 8) patches. The extent of the lower density nearthe rims will be determined by the interplay between diffusion(thus temperature and diffusion barrier) and Ehrlich–Schwoebelbarrier. In this framework, the persistent c(8 � 8) rims arenaturally explained by the fact that there the admolecule densityremains below the threshold value needed for compression. Thisexplains, why compression does not occur in small c(8 � 8)patches. It also explains why the compression starts on smallerdomains at lower substrate temperatures and identical BDAdeposition rates: due to the lower mobility across step edges,the required threshold density can be reached even onsmaller domains.
The compression process is now thought to occur by areduction of free energy: the physisorbed admolecules on topof the c(8 � 8) domains can gain binding energy by makingcontact to the substrate, which is possible with the relativelyopen c(8 � 8) structure. A local transformation from thestructure sketched in Fig. 5(a) to that of Fig. 5(b) requires arelatively mild rearrangement: molecules have to move slightlyand leave their preferred adsorption geometry, and also rotateslightly to optimize the intermolecular interaction in the newsituation. However, the weak hydrogen bonds in the c(8 � 8)structure have to be (partially) broken, which would explainwhy the process appears to be thermally activated. In short, thetotal binding to the substrate has increased at the cost of lessattractive lateral interaction between molecules. The mLEEDobservations reveal equal spot intensities for all four rotational/mirror domains. This is consistent with the simultaneousnucleation of the compressed phase at several locations nearthe center. We note that these domains cannot be resolved withLEEM and mLEED. The large number of diffraction spots effec-tively makes it impossible to perform dark field measurementsand the illumination aperture used to do mLEED is just too large(1.4 mm). Since the molecular rearrangement occurs easiest at theedges of the compressed areas, the nuclei will grow and coalescein the center of gravity of the c(8 � 8) domains.
3.5.2 Decay mechanism. After interrupting the BDAdeposition, the compressed areas slowly decay back into thec(8 � 8) phase. This implies that this phase is not stable in theabsence of the enhanced admolecule density on top of the c(8� 8)domains, which is the result of the incoming flux of molecules.As during growth, the interface between the two phases is sharpand the contrast difference remains constant (cf. Fig. 6). Fromthis observation, and the shape of the decay curves (cf. Fig. 7),we conclude that the decompression also takes place at theinterface between the two phases, which is consistent with aprocess reverse to compression: the c(8 � 8) structure isrestored and this happens most naturally at the interfacebetween both phases. The excess molecules are then expelledon top of the c(8 � 8) domain, where they form mobileadmolecules. Eventually, the admolecules will descend fromthe domain and be incorporated in growing c(8 � 8) domainsnearby or perhaps even desorb into vacuum. Indeed weobserved the very slow loss of material into vacuum at elevated
temperatures (above E 380 K), which means that even a fullydeveloped compressed layer would not be stable and willeventually convert back into the c(8 � 8) structure. However,this will happen on a longer time scale, which is limited by thedesorption of molecules from the layer, as compared to thedecay shown for example in Fig. 7 and 8, where the c(8 � 8)phase can still expand onto terraces which are merely coveredwith a diluted BDA phase.
We mention that the sketched mass transport events wouldsatisfy micro-reversibility requirements. However, the situationmay be more complicated, since we cannot exclude that thediffusing particles are different: the arriving BDA molecules willbe protonated, while the expelled molecules may be (partially)deprotonated. As mentioned above, we will stick to this simpledescription since we have no access to detailed molecularprocesses. The described formation and decay processes areillustrated in Fig. 9.
We found that the decay process is activated with 0.63 eV(cf. Fig. 8). An obvious assumption would be that this is theenergy required to push out one molecule from the layer at theboundary between the compressed and the c(8 � 8) phases.
4 Conclusions
We have demonstrated with LEEM and mLEED the formation ofa compressed BDA phase on Cu(001). The compressed phasenucleates in the center of large (uncompressed) c(8 � 8)domains during BDA deposition at elevated temperatures(B>370 K). The interface separating both phases is very welldefined. We determined the structure in the compressed phase
with mLEED and found a5 �75 4
� �overlayer, with four equally
populated domains (two rotational domains, each split into two
Fig. 9 Sketch of the compressed phase formation (a) and decay (b). (a) Duringdeposition of molecules on the existing layers, an admolecule phase is formed.Physisorbed admolecules have two ways to leave the layer: diffusion down onto aCu terrace (dashed arrow) or incorporation into the compressed phase (solidarrow). The density ram(x) of the admolecule phase will be highest in the centerof the domain, where the compressed phase nucleates first. (b) A proposedmechanism that leads to the decay of the compressed phase after stopping thedeposition: molecules are expelled at the interface between the compressed andthe relaxed phase (solid green arrow), where they form again admolecules thatcan diffuse down (or desorb; not shown).
5016 Phys. Chem. Chem. Phys., 2013, 15, 5007--5016 This journal is c the Owner Societies 2013
mirror domains). The structure is commensurate with higherorder and the molecules are packed 14% denser than in theuncompressed c(8 � 8) structure.
The emergence of a compressed phase shows the delicatebalance of the different energies in the growth of supramole-cular networks on metal surfaces. The interaction of thechemisorbed molecules with the substrate is much strongerthan the mutual interaction between adsorbed moleculesthrough relatively weak hydrogen bonds. Therefore, the mole-cules in the networks are only very loosely bound to eachother. At the relatively elevated temperatures used here,the presence of admolecules on the layers is apparently alreadyenough to break the molecular networks and allow the incor-poration of additional molecules, which reduces the total freeenergy.
References
1 D. E. Hooks, T. Fritz and M. D. Ward, Adv. Mater., 2001, 13,227–241.
2 L. Kilian, U. Stahl, I. Kossev, M. Sokolowski, R. Fink andE. Umbach, Surf. Sci., 2008, 602, 2427–2434.
3 M. Stohr, M. Wahl, C. H. Galka, T. Riehm, T. a. Jung andL. H. Gade, Angew. Chem., Int. Ed., 2005, 44, 7394–7398.
4 L. Kilian, A. Hauschild, R. Temirov, S. Soubatch, a. Scholl,A. Bendounan, F. Reinert, T.-L. Lee, F. Tautz, M. Sokolowskiand E. Umbach, Phys. Rev. Lett., 2008, 100, 136103.
5 C. Stadler, S. Hansen, I. Kroger, C. Kumpf and E. Umbach,Nat. Phys., 2009, 5, 153–158.
6 G. Ertl, M. Neumann and K. Streit, Surf. Sci., 1977, 64, 393–410.7 S. Soubatch, I. Kroger, C. Kumpf and F. Tautz, Phys. Rev. B,
2011, 84, 195440.8 A. Dmitriev, N. Lin, J. Weckesser, J. V. Barth and K. Kern,
J. Phys. Chem. B, 2002, 106, 6907–6912.9 S. Stepanow, T. Strunskus, M. Lingenfelder, A. Dmitriev,
H. Spillmann, N. Lin, J. V. Barth, C. Woll and K. Kern,J. Phys. Chem. B, 2004, 108, 19392–19397.
10 S. Clair, S. Pons, A. P. Seitsonen, H. Brune, K. Kern andJ. V. Barth, J. Phys. Chem. B, 2004, 108, 14585–14590.
11 T. Classen, M. Lingenfelder, Y. Wang, R. Chopra,C. Virojanadara, U. Starke, G. Costantini, G. Fratesi,S. Fabris, S. de Gironcoli, S. Baroni, S. Haq, R. Raval andK. Kern, J. Phys. Chem. A, 2007, 111, 12589–12603.
12 A. Langner, S. L. Tait, N. Lin, C. Rajadurai, M. Ruben andK. Kern, Proc. Natl. Acad. Sci. U. S. A., 2007, 104, 17927–17930.
13 M. Ruben, D. Payer, A. Landa, A. Comisso, C. Gattinoni,N. Lin, J.-P. Collin, J.-P. Sauvage, A. De Vita and K. Kern,J. Am. Chem. Soc., 2006, 128, 15644–15651.
14 S. Stepanow, M. Lingenfelder, A. Dmitriev, H. Spillmann,E. Delvigne, N. Lin, X. Deng, C. Cai, J. V. Barth and K. Kern,Nat. Mater., 2004, 3, 229–233.
15 M. E. Canas Ventura, F. Klappenberger, S. Clair, S. Pons,K. Kern, H. Brune, T. Strunskus, C. Woll, R. Fasel andJ. V. Barth, J. Chem. Phys., 2006, 125, 184710.
16 W. D. Xiao, Y. H. Jiang, K. Ait-Mansour, P. Ruffieux, H.-J.Gao and R. Fasel, J. Phys. Chem. C, 2010, 114, 6646–6649.
17 D. Schwarz, R. van Gastel, H. J. W. Zandvliet andB. Poelsema, Phys. Rev. B, 2012, 85, 235419.
18 S. L. Tait, H. Lim, A. Theertham and P. Seidel, Phys. Chem.Chem. Phys., 2012, 14, 8217–8223.
19 J. Liu, T. Lin, Z. Shi, F. Xia, L. Dong, P. N. Liu and N. Lin,J. Am. Chem. Soc., 2011, 133, 18760–18766.
20 A. Langner, A. Hauschild, S. Fahrenholz and M. Sokolowski,Surf. Sci., 2005, 574, 153–165.
21 U. Stahl, D. Gador, a. Soukopp, R. Fink and E. Umbach, Surf.Sci., 1998, 414, 423–434.
22 Y. Ye, W. Sun, Y. Wang, X. Shao, X. Xu, F. Cheng, J. Li andK. Wu, J. Phys. Chem. C, 2007, 111, 10138–10141.
23 S. Stepanow, N. Lin, F. Vidal, A. Landa, M. Ruben, J. V. Barthand K. Kern, Nano Lett., 2005, 5, 901–904.
24 S. Stepanow, N. Lin, J. V. Barth and K. Kern, J. Phys. Chem. B,2006, 110, 23472–23477.
25 N. Zhu, T. Osada and T. Komeda, Surf. Sci., 2007, 601, 1789–1794.
26 Y.-F. Zhang, N. Zhu and T. Komeda, Surf. Sci., 2008, 602,614–619.
27 F. S. Khokhar, R. van Gastel and B. Poelsema, Phys. Rev. B,2010, 82, 205409.
28 F. S. Khokhar, R. van Gastel, D. Schwarz, H. J. W. Zandvlietand B. Poelsema, J. Chem. Phys., 2011, 135, 124706.
29 B. Immaraporn, P. Ye and A. J. Gellman, J. Phys. Chem. B,2004, 108, 3504–3511.
30 Q. Chen, C. Perry, B. Frederick, P. Murray, S. Haq andN. Richardson, Surf. Sci., 2000, 446, 63–75.
31 D. Schwarz, R. van Gastel, H. J. W. Zandvliet andB. Poelsema, Phys. Rev. Lett., 2012, 109, 016101.
32 E. Bauer, Rep. Prog. Phys., 1994, 57, 895–938.33 U. Linke and B. Poelsema, J. Phys. E: Sci. Instrum., 1985, 18,
26–27.34 R. Bouwman, J. Vac. Sci. Technol., 1978, 15, 91.35 G. Hlawacek, F. S. Khokhar, R. van Gastel, B. Poelsema and
C. Teichert, Nano Lett., 2011, 11, 333–337.36 K.-H. Ernst, Phys. Status Solidi B, 2012, 249, 2057–2088.37 S. A. Nepijko, N. N. Sedov and G. Schonhense, J. Microsc.,
2001, 203, 269–276.38 G. Rosenfeld, K. Morgenstern, I. Beckmann, W. Wulfhekel,
E. Laegsgaard, F. Besenbacher and G. Comsa, Surf. Sci.,1998, 402–404, 401–408.
39 G. Schulze Icking-Konert, Surf. Sci., 1998, 398, 37–48.40 P. A. Thiel, M. Shen, D.-J. Liu and J. W. Evans, J. Phys. Chem.
C, 2009, 113, 5047–5067.41 K. Morgenstern, G. Rosenfeld and G. Comsa, Phys. Rev. Lett.,
1996, 76, 2113–2116.42 K. Morgenstern, Surf. Sci., 1999, 441, 289–300.43 J. Ikonomov, C. H. Schmitz and M. Sokolowski, Phys. Rev. B,
2010, 81, 195428.44 D. Schwarz, R. van Gastel, H. J. W. Zandvliet and B. Poelsema,
