Coverage-dependent adsorption superstructure transition of C
60Õ Cu(001)
Sheng-Syun Wong
Department of Physics, National Taiwan University, Taipei 106, Taiwan Woei Wu Pai
*
Center for Condensed Matter Sciences, National Taiwan University, Taipei 106, Taiwan Chia-Hao Chen
National Synchrotron Radiation Research Center, Hsinchu 300, Taiwan Minn-Tsong Lin†
Department of Physics, National Taiwan University, Taipei 106, Taiwan and Institute of Atomic and Molecular Sciences, Academia Sinica, Taipei 106, Taiwan 共Received 4 April 2010; revised manuscript received 4 July 2010; published 24 September 2010兲 We have investigated the growth and structure of a C60monolayer film on Cu共001兲 with scanning tunneling microscopy at room temperature and 100 K. We discovered that the equilibrium adsorption structure of annealed C60films depends sensitively on the initial deposition coverage; for a coverage of 0.5 monolayer C60 orders in an one-bright-and-one-dim共1B1D兲 row sequence along the 关110兴 direction whereas for a coverage close to one monolayer C60orders in a two-bright-and-one-dim共2B1D兲 sequence. At the transition region of the bright and dim row segments, C60often appears “frizzy” at room temperature. This indicates that a C60 rotates and adopts molecular orientations with inequivalent symmetry. Upon annealing, the C60film exhibits a high thermal stability before C60 fragmentation and desorption occur at ⬃880–960 K, depending on its adsorption superstructure. The duality of equilibrium superstructure in C60/Cu共001兲 is unique among studied C60monolayers on metals. We argue that different boundary energy of the 1B1D and 2B1D phases offers a plausible explanation on the observed tunability of superstructure versus coverage.
DOI:10.1103/PhysRevB.82.125442 PACS number共s兲: 68.55.ap, 68.35.bp, 68.37.Ef, 68.35.Rh
Since the discovery of C60buckminsterfullerene by Kroto et al. in 1985,1 great efforts have been made to understand the physical and chemical properties of fullerene materials.
In particular, the nucleation, growth, and structure of ultra- thin C60 films on various surfaces draw significant attention in the past two decades.2–9A general picture emerged prima- rily from scanning tunneling microscopy 共STM兲 studies is that a C60film often adopts close-packed hexagonal or quasi- hexagonal structure, with a C60 nearest-neighbor共NN兲 dis- tance at most slightly changed from the C60 bulk NN value of 10.0 Å. This is because the intermolecular van der Waals adsorbate-adsorbate interactions often dominate over the adsorbate-substrate interactions. The latter interaction plays a more subtle role; when a C60molecule adsorbs on a preferred adsorption site, it can often lead to C60-induced substrate reconstruction and result in a dimmer C60 appearance in STM.10–17 Therefore, the often observed bright-and-dim-C60 superstructures in C60films are a consequence of incommen- surability between a close-packed C60film and the substrate.
For example, when a close-packed C60 film adsorbs on a substrate of different symmetry, e.g., Ag共001兲, it leads to incommensurability mostly in one direction only and forms a so-called “aperiodic incommensurate” phase18 showing mo- lecular contrast with merely short-range order. For another noble-metal Cu共001兲, a previous STM study19 has revealed another superstructure with alternating sequence of “bright”
and “dim” C60 rows. Here we report two classes of super- structure, one has a two-bright-and-one-dim 共2B1D兲 se- quence ordering along关110兴 and another has an one-bright- and-one-dim 共1B1D兲 sequence instead. We discovered that
these two types of ordering at equilibrium 共after sufficient annealing兲 depend on the initial C60 surface coverage, e.g., for coverage of ⬃1 monolayer 共ML兲 the preferred type is 2B1D and for lesser coverage共e.g., ⬃0.5 ML兲 the preferred type becomes 1B1D. This result is unexpected because C60 aggregates into close-packed islands. Therefore, although the overall surface coverage decreases, the local packing density inside a C60island is not expected to change and one would not expect the superstructure to change. We shall argue that boundary energy of C60 islands plays a plausible role to
“tune” the superstructure type as C60coverage changes.
It must be noted that the 2B1D and 1B1D superstructures we studied are thermal equilibrium phases, rather than meta- stable disordered structure. Figure1共a兲is a room-temperature STM micrograph of a full C60monolayer on Cu共001兲 post- annealed to 648 K. The 2B1D sequence is seen in the regime marked “2B1D.” Ordered row sequence appears regularly after annealing to a temperature共TA兲ⱖ650 K. This ordered phase appears to have a rather high thermal stability; we found that fragmentation and desorption became evident only when TAexceeded 950 K. When the initial surface C60 coverage is 0.5 ML, the annealed 共at 600 K兲 equilibrium superstructure changes from 2B1D to 1B1D. This is shown in Fig. 1共b兲. This phase has a lower thermal stability, with fragmentation and desorption at TAⱖ880 K. We thus refer the equilibrium 2B1D共1B1D兲 superstructures as the ordered structures produced by annealing to 650共600兲ⱕTA
ⱕ900共800兲 K. Furthermore, higher annealing temperatures lead two phases to better ordering.
We conducted measurements at a temperature共Tm兲 of 300
or 100 K. The experiments were conducted with a commer- cial Omicron variable-temperature scanning probe micros- copy 共SPM兲 and a room-temperature SPM housed in two separate home-built ultrahigh vacuum chambers with a base pressure⬃5⫻10−11 Torr. The Cu共001兲 crystal was cleaned by repeated cycles of 2 KeV neon or argon sputtering and annealing to ⬃800 K for 30 min. The cleanness of surface was judged by atomically flat terraces with a nominal terrace widthⱖ100 nm and the lack of step bunches and step pin- ning contaminants. Powder C60 with purity higher than 99.9% was deposited from heated alumina crucibles with a nominal rate of 0.005 ML/s. The sample was held at room temperature during deposition. Well-ordered C60共full or sub兲 monolayer films were prepared by subsequent annealing at proper TA for 1 h. The sample temperature quoted in this work was calibrated against a dummy Cu sample with iden- tical geometry and was estimated to be accurate to⫾10 K . Figure 1共c兲 shows 0.5 ML C60 deposited on Cu共001兲 at room temperature without further thermal annealing. The molecular layer seems to comprise regions with different layer thickness. The ratio of bright C60 to dim C60 exceeds
⬃10, indicating that room-temperature deposition has not yet led to extensive interface reconstruction.11,18 Therefore, the ordering of the C60 phase is very poor due to kinetic hin- drance. The STM contour line depicted in Fig. 1共f兲 reveals not only the ⬃0.9 Å height difference between the bright and dim C60but also the⬃1.8 Å height of a Cu共001兲 single
step. Similar disordered phase was also observed for unan- nealed 1 ML C60.
The 2B1D structure of Fig. 1共a兲has two orthogonal do- mains due to the square symmetry of the Cu共001兲 substrate.
This superstructure is identical to what Abel et al.19reported in an earlier study. A high-resolution view 关Fig. 1共a兲lower right inset兴 reveals that the dim C60共denoted as D-C60here- after兲 has a three-lobe intramolecular feature at room tem- perature. The three-lobe C60 shape is typical of the lowest unoccupied molecular orbital for a C60adsorbed with its car- bon hexagon facing up. The intramolecular feature of the bright C60共denoted as C60兲 cannot be resolved at room tem- perature, however. Recent low-temperature 共⬃8 K兲 STM studies20,21have found at least five different C60orientations on Cu共001兲, indicating that the C60 adsorption geometries in this system vary and are by no means uniform. The multitude of possible adsorption configurations will likely lead to C60 switching between orientations with nearly degenerate en- ergy. Here we report a distinct type of C60 共denoted as
⬃10%兲 contrast in addition to the D-C60 and B-C60. The F-C60molecule has a frizzy appearance; this is typically in- terpreted in STM as dynamical switching of configuration 共such as frizziness of step edges22兲. We may compare the C60/Cu共001兲 and the C60/Ag共100兲 system.18,19 In C60/Ag共100兲, three C60 contrast types共bright, dim, and me- dium兲 are static. The fluctuating contrast of F-C60 suggests that the superstructure may be prone to change, as discussed
[110]
B D
F
A
A'
(a)
1 nm
T = 648 K &T = 300 KA m
1 nm 1 nm
2B1D
(e)
0 3 6 9 12
0.0 0.3 0.6 0.9 1.2 1.5 1.8
A-A' ( nm )
Height(A)
0.9 A
0.4 A 0.9 A
H-H' ( nm )
0.00 3 6 9 12
0.3 0.6 0.9 1.2
1.5 0.4 A
Height(A)
(b) T = 600 K &T = 300 KA m
B D F H
H'
[ 110 ]
1 nm
1B1D (f)
9 6 3 0 4.0
3.0
2.0
1.0
0.0
Height(A)
Q-Q' ( nm ) 1.8 A
0.9 A
(c) T = 300 K
m
1 nm 2 nm
Q
Q'
[110]
(d)
FIG. 1. 共Color online兲 共a兲 and 共b兲 are the STM images of full and half coverage after postannealing, respectively. All the STM images were taken at room temperature. The inset in 共a兲 shows a higher resolution STM image. 共c兲 is an STM image of 0.5 ML C60/Cu共001兲 prepared and imaged at room temperature without postannealing. The molecular layer is obviously poorly ordered. Note some region 共circled兲 already exhibits the 1B1D ordering. 共d兲, 共e兲, and 共f兲 show the line profiles from 共a兲, 共b兲, and 共c兲, respectively.
in the 2B1D-1B1D transition later. The C60/Ag共100兲 mol- ecules were most frequently observed in the regime when the contrast type along a C60row changes from the B-C60to the C60or vice versa. We hereafter call such a transition region as “kink” sites. The presence of kinks is clearly due to the offset of the B-C60 and D-C60 rows along 关11¯0兴. For the 1B1D film as shown in Fig. 1共b兲, the 2B1D ordering was replaced by either 1B1D or some 1B1F共one-bright-and-one- frizzy兲 ordering. Similar to the 2B1D structure, the B-, D-, and F-C60species are also present in 1B1D. Interestingly, the F-C60 species appears much more abundant in the 1B1D phase.
We first discuss the topographic contrast of the three C60 types in both the 2B1D and 1B1D phases, and at 300 and 100 K. Figure1共d兲shows a 2B1D line profile共AA⬘兲 of Fig.
1共a兲at room temperature. The height difference between the B-C60 and D-C60 is ⬃0.9⫾0.1 Å. In addition, this differ- ence is nearly bias independent from −2 to +2 V 共not shown兲. The height difference between the F-C60and B-C60 is ⬃0.4⫾0.2 Å. Figure 1共e兲 shows a 1B1D line profile 共HH⬘兲 of Fig.1共b兲. The height difference between the B-C60 and D-C60 is ⬃0.9⫾0.1 Å, and between the B-C60 and F-C60is⬃0.4⫾0.1 Å. Both are the same as those observed in the 2B1D phase 关Fig. 1共d兲兴. Upon cooling, the overall 2B1D structure does not change but the F-C60species disap- pears. Figure2共a兲is a 2B1D area taken at 100 K. No frizzy F-C60can be seen anymore. Figure2共c兲shows the line pro- file共CC⬘兲 of Fig.2共a兲. A distinct C60contrast共denoted as M兲 was seen. Most of the M-C60molecules locate at one or two kink sites. The height difference of the B-C60 and D-C60 remains at ⬃0.9⫾0.1 Å, same as that measured at room
temperature. The height difference of the B-C60and M-C60is
⬃0.4⫾0.1 Å. While this value is similar to that of the B-C60
and F-C60, only a small fraction of room temperature F-C60 turned into M-C60 upon cooling. The behavior of 1B1D phase after cooling at 100 K is similar to the 2B1D phase, see Fig.2共b兲. The height differences between the B-, D-, and F-C60species of the 1B1D phase共section LL⬘兲 are shown in Fig. 2共d兲. The height difference between the B-C60 and the D-C60is still⬃0.9⫾0.1 Å, and between B-C60and M-C60is
⬃0.3⫾0.1 Å 关Fig. 2共d兲兴. All the above-quoted bright-and- dim height differences are likely of topographic nature. It is now accepted that C60 in different orientations can give slight height difference 共⬃0.2 Å or less兲 in topographic STM images. For larger height difference 共⬃0.5 Å or above兲, it indicates the presence or difference of interface reconstruction.23In addition, all observed height differences were nearly bias independent, supporting the C60contrast as of topographic origin. This suggests that the B-, D-, and F-C60 have different adsorption interface structure but are identical in either the 2B1D or 1B1D phases. It is also noted that the bright rows of the 2B1D and 1B1D phases are often
“offset,” i.e., they shift sideway near the kink sites. Such a row offset is shown by the dashed lines in Figs. 2共a兲 and 2共b兲.
We further examine the ratios of B-, D-, F- and M-C60of 2B1D at room temperature and 100 K关Fig.3共a兲兴. Upon cool- ing, the B-C60ratio is nearly unchanged. The disappearance of the F-C60leads to increased ratios of both the D-C60 and M-C60with about an equal probability. The ratios of the B-, D-, F- and M-C60species of 1B1D at room temperature and 100 K were counted in Fig.3共b兲. At 100 K, the F-C60 ratio reduces from 45.3% to 0%. Most of F-C60 molecules turn into D-C60, as the D-C60ratio increases markedly from 5% to 33.2% and the M-C60ratio to 6.7%. As mentioned earlier, the F-C60 species indeed is much more abundant in the 1B1D phase and the ratios of M-C60 correspond to the density of kink sites. Therefore, M-C60 is not simply the “static” con- trast of the frizzy F-C60. The M-C60and F-C60must be con- sidered as distinct as they are likely to have different adsorp- tion structures.
Annealing experiments showed that both 2B1D and 1B1D are equilibrium structures. Figure4共a兲shows the evolution of the B-, D-, and F-C60 ratios in the 2B1D structure after an- nealing to different temperatures for 1 h. The statistics were measured at 300 K. The 2B1D structure remains the same up
(a) (c)
2 nm
B D
M C
C'
T = 860 K &T = 100 KA m
[110]
(b) T = 720 K &T = 100 KA m (d)
B D M
2 nm
L
L'
[110]
0 3 6 9 12
C-C' ( nm )
Height(A)
0.5 1.0 1.5 2.0
1.0 A 0.4 A
0 3 6 9 12 15
Height(A)
0.0 0.5 1.0 2.0
1.5
L-L' ( nm )
0.9 A 0.3 A
FIG. 2. 共Color online兲 共a兲 and 共b兲 are the 2B1D and 1B1D phases taken at 100 K. No frizzy C60can be seen anymore, and a distinct C60contrast共denoted as M兲 was observed. The dashed lines shown in 共a兲 and 共b兲 indicate that the 2B1D and 1B1D ordering often has offset. In 共c兲 and 共d兲, the line profiles from 共a兲 and 共b兲, respectively, are depicted.
(a) (b)
B D F M
20 40 60 80
Ratio(%) 70.9% 71.6% 22.3% 25.5% 6.8% 0.0% 0.0% 2.9%
: T = 300 Km : T = 100 Km
0
B D F M
0 20 40 60 80
Ratio(%) 49.7% 60.1% 5.0% 33.2% 45.3% 0.0% 0.0% 6.7%
: T = 300 Km : T = 100 Km
Phase : 1B1D Phase : 2B1D
FIG. 3.共Color online兲 The ratios of B-, D-, F-, and M-C60in the 2B1D and 1B1D phases at room temperature and 100 K are shown in共a兲 and 共b兲, respectively.
to TA⬃890 K. At TA⬃960 K, C60desorbs or decomposes, reducing the 2B1D phase coverage but otherwise leaving the surface contaminated with carbon fragments关see Fig. 5共a兲兴.
Without annealing 共TA= 300 K兲, most C60 appears bright.
Upon annealing to TAⱖ650 K, the B-C60 ratio decreases, indicating thermally activated interface reconstruction simi- lar to, e.g., C60/Cu共111兲.8,11The ratios of B-, D-, and F-C60 reach static values and are averaged at 63.7%, 21.2%, and 15.1%, respectively, with only insignificant variation. Figure 4共b兲 shows the ratios of B-, D-, and F-C60 of 1B1D post annealed from 600 to 800 K. The ratios remain approxi- mately fixed. Similar to the 2B1D case, at room temperature the B-C60is the dominant species. The decomposition of C60 already occurs at a lower temperature 共⬃880 K兲 as com- pared with the temperature⬃960 K for 2B1D, see Fig.5共b兲.
Comparing the various C60 ratios of the 1B1D and 2B1D superstructures, we note that the B-C60and D-C60ratios de- crease in 1B1D but the F-C60 ratio increases dramatically from ⬃10% to ⬃40%. Since for both annealed 1B1D and 2B1D phases the ratios of B-, D-, and F-C60are independent of thermal treatment if heated in the proper temperature range, it indicates that both phases are thermal equilibrium phases. We also note the quality of 2B1D and 1B1D phases improves after annealing at higher temperature. Figures5共c兲 and5共d兲show the 2B1D and 1B1D phases annealed at 890 K and 800 K, respectively. Clearly, the superstructures are much better ordered when compared with Figs.1共a兲and1共b兲 共annealed at 650 K and 600 K兲, respectively. This also sug- gests that the pure 2B1D and 1B1D phases are more thermo- dynamically stable.
The main unexpected finding is the two different super- structures at full and half ML coverage, respectively. There are quite a few other cases of coverage-dependent super- structure transitions in thin films. For example, alkali adsorp- tion often leads to different ordering superstructures at dif- ferent coverages.24–26In molecular films, such transitions are also known.27–29 This is mainly due to the molecule- molecule interactions and possibly affected by molecule- substrate interactions as well. Therefore, the different super- structure phases often have varied packing density as the molecular interactions stabilize the superstructure in different ways. For C60 films, as observed in all other studies and the present one, strong C60-C60interactions simply lead to close- packed islands. While there could exist various metastable packing structures,11 the well-annealed films often exist in one particular ordering only and it is expected not to observe
any coverage-dependent thermal equilibrium states. There- fore, the observed 2B1D-1B1D superstructure transition ap- pears to fall into a different category. We will offer a plau- sible explanation.
Thermal annealing does not drive the superstructure tran- sition; it only stabilizes either the 2B1D and 1B1D phases as shown in Figs. 5共c兲and 5共d兲. Instead, we consider the rel- evant energy terms to stabilize these two phases. As a first- order approximation, it is natural to treat the C60island en- ergy as a linear function of its size. This sole energy term clearly cannot explain the superstructure transition. Further refinement must take island boundary energy and corner en- ergy into consideration. This has been often done, e.g., for homoepitaxial metallic islands.30 Here for C60 islands with superstructure and unknown details of C60 adsorption geom- etry inside an island and at the island step, it is impossible to take a theoretical analysis much further. Instead, we give a heuristic argument, explaining that the energy of C60 island boundary共between C60and bare Cu兲 drives the 2B1D-1B1D transition. At full coverage, the boundary energy is irrelevant because they are almost eliminated. For half coverage, the C60 island boundary length is much increased 共likely to be the maximum for all coverage less than 1 ML兲. If the 2B1D or 1B1D C60 islands have distinct boundary geometry, they will have different boundary energy. We first examined whether the boundary structure is simply a truncated termi- nation of the superstructure phase. Figure 6共a兲 shows the boundary structure of the 1B1D phase. Interestingly, along the C60 关11¯0兴 row direction 共P兲 we found in all cases the islands are terminated with a bright C60 row instead of the expected 1:1 stoichastic B:D ratio. Along the approximately orthogonal direction 共P⬘兲, all terminated C60molecules also appear bright. For the 2B1D phases, the coverage is nearly full and we only observed boundary structure in small vacan- cies as shown in Fig.6共b兲. The 2B1D boundary structure is clearly more irregular. The 2B1D and 1B1D phases therefore indeed have different boundary structures. Because the 2B1D phase is the preferred phase at full coverage, the areal energy term共Ea兲 must be smaller. When the C60overall coverage is reduced, boundary becomes present and the total energy E must include the step energy term Eb as well. For a fixed island shape with area A and boundary length L, the total energy E = A⫻Ea+ L⫻Ea. To optimize E, the island shape can alter 共Wulff construction31兲 or, in our case, the super- structure can change. The latter is most likely to occur when at least two superstructures are nearly degenerate in energy.
(a) (b)
Ratio(%)
100
300 600 700 800 900
T ( K )A
B-C60
D-C60
F-C60
80
40
0 20 60
300 600 700 800
0 30 60 90
Ratio(%)
T ( K )A
B-C60
D-C60
F-C60
T = 300 K & Phase : 2B1Dm T = 300 K & Phase : 1B1Dm
FIG. 4. 共Color online兲 共a兲 and 共b兲 show, respectively, the ratios of B-, D-, and F-C60in the 2B1D and 1B1D phases with different annealing temperatures. The ratios become approximately fixed above proper temperatures.
We believe the 2B1D and 1B1D phases satisfy this condi- tion. For the transition to occur, there should be a threshold
“surface-volume ratio” L/A, which in terms depends on the coverage. Therefore, the presence of the lower-energy 1B1D boundary compensates the energy cost for converting the 2B1D phase to the 1B1D phase. For a surface coverage of 0.7 ML共not shown兲, we observed patches of area with either the 1B1D phase or a mixed 2B1D-1B1D phase. The latter appears somewhere between that of Figs.1共a兲and1共b兲. This observation is also consistent with our heuristic argument
because different C60 islands have different L/A ratios and thus may stabilize at different superstructures. For 0.5 ML, the predominant superstructure is 1B1D, in particular, after sufficient annealing. Because our argument is heuristic, we have omitted other subtle factors such as island corner en- ergy and C60domain boundary.
Finally, we comment on the observed thermal stability of the 1B1D and 2B1D phases. There have been many studies for C60 adsorption on metal surfaces, with film desorption temperatures reported in some cases, e.g., Ni共110兲
(c)
T = 960 K &T = 300 KA m4 nm
(a) (c)
1.5 nm
T = 890 KA & T = 300 Km
[ 110 ] T = 960 K & T = 300 KA m
[ 110 ]
(b) (d)
T = 800 KA & T = 300 Km1.5 nm [ 110 ] T = 880 KA & T = 300 Km
2.5 nm [ 110 ]
FIG. 5. 共Color online兲 共a兲 At TA⬃960 K, C60desorbs or decomposes in the 2B1D phase. 共b兲 The decomposition of C60was also observed in the 1B1D phase albeit with a lower TA⬃880 K. In 共c兲 and 共d兲, we observed that after more extensive annealing, the 2B1D and 1B1D phases become much more ordered when compared with that of Figs.1共a兲and1共b兲, respectively.
共⬃760 K兲,32 Pt共111兲 共⬃560 K兲,32 Cu共110兲 共⬃730 K兲,9 Au共111兲 共⬃773 K兲,33and Ag共111兲 共⬃773 K兲.32,33These de- sorption temperatures are all lower than what was observed in C60/Cu共001兲, e.g., 880–960 K. A recent study8,11 has shown that C60induces extensive reconstruction on Cu共111兲.
For C60 on Cu共001兲, the bonding could also be strong and lead to a high desorption or decomposition temperature. For the frizzy appearance of F-C60, it could due to fast C60rota- tion. Altman et al.34 have reported contrast switching from bright C60 to dim C60 and vice versa on Au共111兲 and attrib- uted it to C60 rotation. There exists another scenario, i.e., a C60can also reside in a “dynamic” adsorption geometry that changes due to short-range substrate mass flow. This was observed in the C60/Ag共100兲 system18in which the contrasts of three different species共B-, M-, and D-C60兲 can switch at
room temperature. Local microscopic detailed balance was established, indicating short-range mass transport. Since we argue that the F-C60and D-C60are more alike关Fig.4共b兲兴 but the topographic height between the F-C60 and D-C60 is
⬃0.5 Å, it appears unlikely that this is purely due to differ- ent C60 orientations. Instead, a short-range mass flow changes the Cu structure beneath C60, with possibly a con- comitant change in C60orientation, is more likely. In contrast to C60/Ag共100兲, the much faster contrast switching rate seen here is related to a more facile adsorption structure through adding or extracting Cu atoms below the F-C60 species.
Upon cooling, even a short-range substrate mass transport is quenched, explaining why the frizzy F-C60 disappears. We surmise that the facile nature of F-C60 may initiate the tran- sition between 2B1D and 1B1D.
In summary, we have studied a coverage-dependent su- perstructure transition of C60/Cu共001兲 by STM at room tem- perature and 100 K. For a coverage of 0.5 ML, C60orders in an 1B1D row sequence whereas for an one ML C60film the preferred ordering is a 2B1D sequence. We give a heuristic argument that the transition is driven by C60island boundary energy. A fast-contrast-switching C60species共F-C60兲 and the high thermal film stability were also characterized. Our study shows that C60adsorption on a substrate with an incommen- surate symmetry and lattice can lead to a variety of compli- cated adsorption phenomena. Furthermore, the previously unknown coverage-dependent superstructures of a C60 film may offer a possibility to make functional molecular thin films.
This work was supported in part by the National Science Council of Taiwan through Grants No. NSC 95-2120-M-002- 015, No. NSC 95-2112-M-002-051-MY3, and No. NSC 98- 2112-M-002-013-MY3.
1H. W. Kroto, J. R. Heath, S. C. O’brien, R. F. Curl, and R. E.
Smalley,Nature共London兲 318, 162 共1985兲.
2R. J. Wilson, G. Meijer, D. S. Bethune, R. D. Johnson, D. D.
Chamblis, M. D. de Vries, H. E. Hunziker, and H. R. Wendt, Nature共London兲 348, 621 共1990兲.
3Y. Z. Li, J. C. Patrin, M. Chander, J. H. Weaver, L. P. F. Chiba- nte, and R. E. Smalley, Science 252, 547 共1991兲; 253, 429 共1991兲.
4Y. Z. Li, M. Chander, J. C. Patrin, J. H. Weaver, L. P. F. Chiba- nte, and R. E. Smalley,Phys. Rev. B 45, 13837共1992兲.
5T. Hashizume, K. Motai, X. D. Wang, H. Shinohara, Y. Saito, Y.
Maruyama, K. Ohno, Y. Kawazoe, Y. Nishina, H. W. Pickering, Y. Kuk, and T. Sakurai,Phys. Rev. Lett. 71, 2959共1993兲.
6Y. Kuk, D. K. Kim, Y. D. Suh, K. H. Park, H. P. Noh, S. J. Oh, and S. K. Kim,Phys. Rev. Lett. 70, 1948共1993兲.
7E. I. Altman and R. J. Colton,Phys. Rev. B 48, 18244共1993兲.
8W. W. Pai, H. T. Jeng, C.-M. Cheng, C.-H. Lin, X. Xiao, A.
Zhao, X. Zhang, G. Xu, X. Q. Shi, M. A. Van Hove, C.-S. Hsue, and K.-D. Tsuei,Phys. Rev. Lett. 104, 036103共2010兲.
9P. W. Murray, M. Ø. Pedersen, E. Lægsgaard, I. Stensgaard, and F. Besenbacher,Phys. Rev. B 55, 9360共1997兲.
10W. W. Pai, C.-L. Hsu, C. R. Chiang, Y. Chang, and K. C. Lin, Surf. Sci. 519, L605共2002兲.
11W. W. Pai, C.-L. Hsu, M. C. Lin, K. C. Lin, and T. B. Tang, Phys. Rev. B 69, 125405共2004兲.
12J. K. Gimzewski, S. Modesti, and R. R. Schlittler, Phys. Rev.
Lett. 72, 1036共1994兲.
13J. Weckesser, C. Cepek, R. Fasel, J. V. Barth, F. Baumberger, T.
Greber, and K. Kern,J. Chem. Phys. 115, 9001共2001兲.
14J. Weckesser, J. V. Barth, and K. Kern,Phys. Rev. B 64, 161403 共2001兲.
15A. J. Maxwell, P. A. Brühwiler, D. Arvanitis, J. Hasselström, M.
K.-J. Johansson, and N. Mårtensson, Phys. Rev. B 57, 7312 共1998兲.
16W. W. Pai and C.-L. Hsu,Phys. Rev. B 68, 121403共R兲 共2003兲.
17X. Zhang, W. He, A. Zhao, H. Li, L. Chen, W. W. Pai, J. Hou, M.
M. T. Loy, J. Yang, and X. Xiao, Phys. Rev. B 75, 235444 共2007兲.
18C.-L. Hsu and W. W. Pai,Phys. Rev. B 68, 245414共2003兲.
19M. Abel, A. Dmitriev, R. Fasel, N. Lin, J. V. Barth, and K. Kern, (a) T = 704 K & T = 300 KA m (b) T = 720 KA & T = 300 Km
1 nm 3 nm
[ 110 ]
P
P'
1 nm 1.5 nm
[ 110 ]
FIG. 6. 共Color online兲 P and P⬘in 共a兲 show the directions and boundary structure of the 1B1D phase. Along the P edge, the islands are exclusively terminated with bright C60rows. Along the P⬘edge, all terminated C60also appears bright. For the 2B1D phase in共b兲, the boundary structure is more irregular.
Phys. Rev. B 67, 245407共2003兲.
20N. Néel, J. Kröger, L. Limot, T. Frederiksen, M. Brandbyge, and R. Berndt,Phys. Rev. Lett. 98, 065502共2007兲.
21N. Néel, J. Kröger, L. Limot, and R. Berndt,Nano Lett. 8, 1291 共2008兲.
22W. W. Pai and J. E. Reutt-Robey, Phys. Rev. B 53, 15997 共1996兲.
23W. W. Pai, C. L. Hsu, K. C. Lin, L. Y. Sin, and T. B. Tang,Appl.
Surf. Sci. 241, 194共2005兲.
24W. Ernst, C. Tegenkamp, H. Pfnür, K.-L. Jonas, V. von Oeyn- hausen, and K. H. Meiwes-Broer,Surf. Sci. 540, 303共2003兲.
25Z. Y. Li, K. M. Hock, and R. E. Palmer, Phys. Rev. Lett. 67, 1562共1991兲.
26Z. P. Hu, N. J. Wu, and A. Ignatiev, Phys. Rev. B 33, 7683 共1986兲.
27C. Stadler, S. Hansen, I. Kröger, C. Kumpf, and E. Umbach,Nat.
Phys. 5, 153共2009兲.
28Z. H. Cheng, L. Gao, Z. T. Deng, N. Jiang, Q. Liu, D. X. Shi, S.
X. Du, H. M. Guo, and H.-J. Gao,J. Phys. Chem. C 111, 9240 共2007兲.
29G. E. Poirier and E. D. Pylant,Science 272, 1145共1996兲.
30Z. Zhang and M. G. Lagally,Science 276, 377共1997兲.
31G. Wulff, Z. Kristallogr. 34, 449共1901兲.
32M. Pedio, K. Hevesi, N. Zema, M. Capozi, P. Perfetti, R. Gout- tebaron, J.-J. Pireaux, R. Caudano, and P. Rudolf,Surf. Sci. 437, 249共1999兲.
33E. I. Altman and R. J. Colton,Surf. Sci. 295, 13共1993兲.
34E. I. Altman and R. J. Colton, in Atomic and Nanoscale Modifi- cation of Materials: Fundamentals and Applications, NATO Ad- vanced Study Institute, Series E: Applied Sciences Vol. E239, edited by P. Avouris共Plenum, New York, 1993兲, p. 303.
