**3.1 Background**

110 Atomic Force Microscopy – Imaging, Measuring and Manipulating Surfaces at the Atomic Scale

The constant shear strength that we observe indicates that the mechanism of energy dissipation for this system does not change in this pressure range. Thus, the increase in friction with load is attributable to the increase in contact area, *i.e.* more atoms in contact, as opposed to a change in the frictional dissipation per interfacial atom. This may not be so surprising given that the nominal stress is only increasing as roughly *L1/3* (from the continuum mechanics models). The most likely mechanism of energy dissipation is thermalization of phonons generated at the contact zone during sliding. New modes of energy dissipation, resulting from inelastic processes, may activate at higher stresses (Carpick & Salmeron, 1997). For example, evidence of tip-induced atomic-scale wear has been reported for alkali-halide materials (Carpick et al., 1998b). Pressure-activated modes of energy dissipation are reported in organic thin films due to progressive molecular deformation (Barrena et al., 1999). These examples represent stress-dependent increases in the number of energy dissipation channels and are therefore manifested in increases in the

Finally, we comment on the relative magnitude of the observed shear strength. The theoretical prediction for the shear strength of a crystalline material in the absence of dislocations is roughly given by *G*/30 (Cottrell, 1988), where *G* is the shear modulus. We can define an "effective" interfacial shear modulus 2 380 GPa *G GG G G eff WC diamond WC diamond* . This

system is thus far below the ideal material shear strength (Hurtado & Kim, 1998). Previous AFM results of Carpick *et al.* (Carpick et al., 1996a; 1998b) and Lantz *et al.* (Lantz et al., 1997a; 1997b) observed shear strengths near the ideal limit. An ideal shear strength in the range of *G*/30 suggests a "crystalline" or commensurate interface that is free of dislocations, where the commensurability may be brought about by atomic displacements induced by interfacial forces. Our measured shear strength indicates that there may be very little atomic commensurability for the diamond/tungsten-carbide interface, which is plausible considering the high stiffness of these materials. More importantly, the hydrogen passivation of the diamond surface strongly reduces the adhesive force, and also the friction force. In fact, removal of the hydrogen passivation would result in a value for the shear strength which is much larger than the ideal theoretical prediction of *G*/30 (van den Oetelaar & Flipse, 1997).

We have investigated the nanotribological properties of a tungsten carbide tip in contact with a clean Pt(111) single crystal surface under ultrahigh vacuum conditions using scanning probe techniques. Because of the conductive nature of the cantilever and tip, we could alternate between contact atomic force microscopy (AFM) and non-contact scanning tunneling microscopy (STM) using the same probe. Several types of interfaces were found depending on the chemical state of the surfaces. The first type is characterized by strong irreversible adhesion followed by material transfer between tip and sample. This resulted in substantial amounts of material being transferred from the tip to the sample upon contact. This material often covered areas far exceeding that of the contact region. Low adhesion and no material transfer characterize a second type of contacts, which is associated with the presence of passivating adsorbates in both (full passivation) or in one of the two contacting surfaces (half-passivation). Half-passivated contacts where the clean side is the Pt(111)

. The shear strength of this

shear strength compared with purely elastic, wearless friction.

gives, for the diamond/tungsten-carbide contact, 1600 *Geff*

**3. Wear and third bodies in nanocontacts** 

The sliding of materials in contact often involves the transfer of material from one surface to the other. This material, referred to sometimes as the third body, influences the transient behavior of the sliding contact and can completely dominate the steady-state sliding behavior of many interfaces, especially for low friction coatings (Singer, 1992; 1998). Studies of low-friction materials such as diamond-like carbon coatings, MoS2 coatings, and Tiimplanted steels indicate that chemically-modified transfer films are formed during initial sliding, and these films determine the long-term frictional behavior of the interface.

At small length scales third-bodies can also have a large impact on the contact properties. For example, hard disks and micro-electromechanical systems (MEMS) are critically limited by friction and adhesion-related failures due to the large surface-to-volume ratios of these devices (McFadden & Gellman, 1997). For such devices, an understanding of nano-scale third body behavior is important. Modeling work supports this notion. Robbins and co-workers have performed molecular dynamics simulations that indicate that molecular intermediate species in asperity contacts have a dramatic effect on friction (He et al., 1999; He & Robbins, 2001). They argue that contacts between crystalline or amorphous materials should, in general, exhibit very low friction due to the lack of interfacial lattice commensurability. The simulations show that molecules trapped at the interface, e.g. hydrocarbons, cause static friction that is consistent with observed macroscopic friction behavior.

The role of third bodies and transferred species at small scales is clearly worthy of further experimental study, specifically through the use of scanning force microscopy techniques. Already, fundamental insights into many aspects of friction have been obtained through the use of scanning force microscopy (Carpick & Salmeron, 1997). These studies have addressed several important topics such as atomic-scale stick-slip behavior, friction in the wearless (low-load) regime, friction in the presence of molecular lubricant films, the role of interfacial contact area, and wear initiation. However, there have been few studies of third body effects and transferred species. One example is the work by Qian et al. (Qian et al., 2000) who showed that in atomic force microscope (AFM) experiments, friction measurements exhibit transient behavior, where several tens of scans were required before friction behavior become reproducible. They proposed that the phenomenon is due to transfer of

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 113

Fig. 8. AES spectrum showing the clean Pt(111) surface, e.g., platinum peaks at 237 eV and 251 eV are clear visible . Inset: a LEED pattern displaying the a clear Pt(111)-(1x1) pattern.

**100 150 200 250 300**

**Pt(237)**

**Pt(251)**

**Energy [eV]**

Two cantilevers with different spring constants of 88 N/m and 3.1 N/m were used. The stiffer cantilever was used for conductance and friction measurements while the other was used for certain high resolution friction measurements only. The tips were cleaned in UHV immediately prior to the measurements, by applying short voltage pulses and/or by rubbing them against the surface. Normal cantilever force constants were taken from the manufacturer, and the normal/lateral force ratio was calculated using the method described in (Ogletree et al., 1996). The absolute accuracy of the forces measured is limited due to significant uncertainty in the material properties of the cantilever and approximations used in the force constant calculations. However, relative changes in friction could be accurately determined by using the same cantilever and tip during a series of measurements. A large dynamic range, two-stage I-V converter was built, which provided a large frequency range while achieving sufficient gain. For this work, lower I-V gains (*e.g.,* 104) were used to measure the current flowing through the tip-sample junction while in AFM-contact mode,

higher gains (*e.g.,* 109), were used for tunneling microscopy using the AFM tip.

As mentioned above, the Pt(111) sample was cleaned by sputtering and annealing, and its state checked by AES and LEED to verify the chemistry and structure of the surface. Because of the nanoscale dimensions of the tip apex, we could not assess spectroscopically its chemical state in the UHV chamber. However, we found that scanning at high loads on sacrificial areas of the sample consistently produced tips with highly adhesive properties and metallic conductance characteristics. Thus, rubbing a contaminated WC tip on a *clean metal* surface is an effective way to clean the tip. The adhesion force for tips prepared in this manner was large enough that even at the lowest load scanning was not possible without

**3.3.1 Irreversible adhesion between clean interfaces** 

**0**

**1**

**2**

**3**

**4**

**AES Signal [a.u.]**

**5**

**6**

**7**

**3.3 Results and discussions** 

material between the tip and sample, and observed that the mechanism of transfer depended on the relative humidity and applied load. Carpick *et al.* (Carpick et al., 1996b) observed that the frictional shear strength and interfacial adhesion energy of a Pt/muscovite mica interface in ultrahigh vacuum progressively decreased with each scan, but recovered if the tip was "cleaned" by blunting it to expose fresh Pt. The authors suggested that potassium adsorbates transferred from the mica surface to the tip could explain the strong, progressive reduction of adhesion and friction observed in the experiment. Using the surface forces apparatus (SFA) Drummond *et al*. (Drummond et al., 2001) performed experiments where WS2 nanoparticles were suspended in a tetradecane fluid, and then compressed and sheared between the two mica sheets of the SFA. They found that the nanoparticles formed a transfer film of nanometer-scale thickness on the mica that reduced friction appreciably.

In this work, we discuss the results of a study of a Pt (111)/tungsten carbide single asperity interface using a combination of ultrahigh vacuum (UHV) AFM and scanning tunneling microscopy (STM) techniques. Since both the Pt sample and the carbide tip are conductive, we were able to measure the local electrical conductance of the contact and the friction force simultaneously (Enachescu et al., 1998; 1999a). In addition, the conducting tip allows STM operation, whereby high-resolution non-contact images of the sample can be obtained before and after the contact experiments.

#### **3.2 Experimental**

The experiments were performed in a UHV chamber (base pressure 7 x 10-11 Torr), equipped with a home-built AFM (Dai et al., 1995), low-energy electron diffraction (LEED), Auger electron spectroscopy (AES), differentially pumped ion sputtering, and sample cooling and annealing capabilities. The Pt(111) single-crystal sample was cleaned by sputtering with Ar+ ions of 1 keV energy, both in hot conditions (600 0C) and at room temperature, for 10-20 min. After sputtering O2 was introduced in the chamber at 10-6 Torr for ~3 min while the sample temperature was kept at 600 0C. O2 exposure and heating were then stopped. The cycle was repeated two or three times. Finally, the sample was flashed to 950-1000 0C for ~1 min. and cooled down at ~2 0C/sec. The AES pattern taken after the cleaning procedure indicated a clean Pt(111) surface, with the carbon peak at 271 eV not visible above the noise level, while platinum peaks at 237 eV and 251 eV are clear visible (Fig. 8). A sharp (1x1) LEED pattern was also observed (inset, Fig. 8).

Commercially available triangular silicon cantilevers with integrated tips, coated with approximately 20 nm of tungsten carbide, were used for all measurements. The cantilevers were characterized by scanning electron microscopy (SEM) in order to determine tip and lever dimensions, and also by AES to determine the chemical composition of the lever and tip shaft. The measurements showed the presence of both tungsten oxide and carbide, which is not uncommon for such coatings. For convenience we will refer to these as "tungsten carbide" tips. The similar chemical composition of the lever and tip is quite normal, as the WC coating is covering not only the tip but also the cantilever. For *such conditions* one may suggest that tip cleanliness is similar to the cantilever cleanliness, i.e., both covered with tungsten oxide and carbide. However, for our AFM/STM measurements most of the time the tip cleanliness is *not* similar to the cantilever cleanliness, as we often clean the AFM-STM tip and thus, removing the tip contaminant.

Fig. 8. AES spectrum showing the clean Pt(111) surface, e.g., platinum peaks at 237 eV and 251 eV are clear visible . Inset: a LEED pattern displaying the a clear Pt(111)-(1x1) pattern.

Two cantilevers with different spring constants of 88 N/m and 3.1 N/m were used. The stiffer cantilever was used for conductance and friction measurements while the other was used for certain high resolution friction measurements only. The tips were cleaned in UHV immediately prior to the measurements, by applying short voltage pulses and/or by rubbing them against the surface. Normal cantilever force constants were taken from the manufacturer, and the normal/lateral force ratio was calculated using the method described in (Ogletree et al., 1996). The absolute accuracy of the forces measured is limited due to significant uncertainty in the material properties of the cantilever and approximations used in the force constant calculations. However, relative changes in friction could be accurately determined by using the same cantilever and tip during a series of measurements. A large dynamic range, two-stage I-V converter was built, which provided a large frequency range while achieving sufficient gain. For this work, lower I-V gains (*e.g.,* 104) were used to measure the current flowing through the tip-sample junction while in AFM-contact mode, higher gains (*e.g.,* 109), were used for tunneling microscopy using the AFM tip.

#### **3.3 Results and discussions**

112 Atomic Force Microscopy – Imaging, Measuring and Manipulating Surfaces at the Atomic Scale

material between the tip and sample, and observed that the mechanism of transfer depended on the relative humidity and applied load. Carpick *et al.* (Carpick et al., 1996b) observed that the frictional shear strength and interfacial adhesion energy of a Pt/muscovite mica interface in ultrahigh vacuum progressively decreased with each scan, but recovered if the tip was "cleaned" by blunting it to expose fresh Pt. The authors suggested that potassium adsorbates transferred from the mica surface to the tip could explain the strong, progressive reduction of adhesion and friction observed in the experiment. Using the surface forces apparatus (SFA) Drummond *et al*. (Drummond et al., 2001) performed experiments where WS2 nanoparticles were suspended in a tetradecane fluid, and then compressed and sheared between the two mica sheets of the SFA. They found that the nanoparticles formed a transfer film of nanometer-scale thickness on the mica that reduced

In this work, we discuss the results of a study of a Pt (111)/tungsten carbide single asperity interface using a combination of ultrahigh vacuum (UHV) AFM and scanning tunneling microscopy (STM) techniques. Since both the Pt sample and the carbide tip are conductive, we were able to measure the local electrical conductance of the contact and the friction force simultaneously (Enachescu et al., 1998; 1999a). In addition, the conducting tip allows STM operation, whereby high-resolution non-contact images of the sample can be obtained

The experiments were performed in a UHV chamber (base pressure 7 x 10-11 Torr), equipped with a home-built AFM (Dai et al., 1995), low-energy electron diffraction (LEED), Auger electron spectroscopy (AES), differentially pumped ion sputtering, and sample cooling and annealing capabilities. The Pt(111) single-crystal sample was cleaned by sputtering with Ar+ ions of 1 keV energy, both in hot conditions (600 0C) and at room temperature, for 10-20 min. After sputtering O2 was introduced in the chamber at 10-6 Torr for ~3 min while the sample temperature was kept at 600 0C. O2 exposure and heating were then stopped. The cycle was repeated two or three times. Finally, the sample was flashed to 950-1000 0C for ~1 min. and cooled down at ~2 0C/sec. The AES pattern taken after the cleaning procedure indicated a clean Pt(111) surface, with the carbon peak at 271 eV not visible above the noise level, while platinum peaks at 237 eV and 251 eV are clear visible (Fig. 8). A sharp (1x1) LEED pattern

Commercially available triangular silicon cantilevers with integrated tips, coated with approximately 20 nm of tungsten carbide, were used for all measurements. The cantilevers were characterized by scanning electron microscopy (SEM) in order to determine tip and lever dimensions, and also by AES to determine the chemical composition of the lever and tip shaft. The measurements showed the presence of both tungsten oxide and carbide, which is not uncommon for such coatings. For convenience we will refer to these as "tungsten carbide" tips. The similar chemical composition of the lever and tip is quite normal, as the WC coating is covering not only the tip but also the cantilever. For *such conditions* one may suggest that tip cleanliness is similar to the cantilever cleanliness, i.e., both covered with tungsten oxide and carbide. However, for our AFM/STM measurements most of the time the tip cleanliness is *not* similar to the cantilever cleanliness, as we often clean the AFM-STM

friction appreciably.

**3.2 Experimental** 

before and after the contact experiments.

was also observed (inset, Fig. 8).

tip and thus, removing the tip contaminant.

#### **3.3.1 Irreversible adhesion between clean interfaces**

As mentioned above, the Pt(111) sample was cleaned by sputtering and annealing, and its state checked by AES and LEED to verify the chemistry and structure of the surface. Because of the nanoscale dimensions of the tip apex, we could not assess spectroscopically its chemical state in the UHV chamber. However, we found that scanning at high loads on sacrificial areas of the sample consistently produced tips with highly adhesive properties and metallic conductance characteristics. Thus, rubbing a contaminated WC tip on a *clean metal* surface is an effective way to clean the tip. The adhesion force for tips prepared in this manner was large enough that even at the lowest load scanning was not possible without

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 115

Fig. 10. Schematic drawing illustrating the deformation of a cantilever subject to high friction forces. (a) The cantilever is shown at the upper left for the case of no load. The dashed line represents the trajectory of the laser for the zero load case, and is included in all subsequent sketches. (b) After snapping into contact, the *z*-displacement is increased. Normally this would cause an increase in the laser signal (after the initial decrease due to the snap-in). But large friction forces cause bending at the end of the lever in the opposing direction, and producing an apparent *decrease* in the load. Friction is preventing the tip from sliding relative to the surface. (c) As the *z*-displacement is further increased, the friction force eventually reaches its limiting value. The tip will begin to slide relative to the surface, and the upward bending induced by the *z*-displacement overtakes the tendency to bend in the opposite direction induced by the friction force, so now the apparent load begins to

increase. Eventually the *z*-displacement reaches a maximum value. (d) As the z-

the *z*-displacement is decreased. Stages (d) and (e) are clearly evident in Fig. 9.

displacement direction is reversed, the friction force will now resist motion in the opposite direction, and so the bending it induces causes an apparent increase in the load. During this phase, the tip is not sliding relative to the surface. (e) Eventually the friction force reaches a limiting value and once again the tip begins to slide. The apparent load will now decrease as

In addition, an unusual hysteresis feature in the force-displacement plots was observed at the largest z-displacements as seen in Fig. 9. The recorded data begins with the tip initially in contact after being pushed back by approximately 450 nm. As the cantilever is retracted, the apparent force on the cantilever *increases*, then eventually begins to decrease, as one would normally expect. This result can be explained by considering the effect of friction on the cantilever bending due to the tilted geometry with respect to the plane of the sample (22.5° in this case). Friction causes the cantilever to bend in addition to bending due to the normal force between the tip and sample. The direction of the bending will depend on the

severe damage. These contacts were characterized by means of force-displacement curves, as in the example shown in Fig. 9, where a pull-off force of *Lc* = 12.0 ± 1.2 µN was measured with the cantilever of 88 N/m normal spring constant. Assuming, for simplicity, an *elastic* adhesive contact, this force can be related to the work of adhesion of the interface. Within the extremes of the Johnson-Kendall-Roberts (JKR) model (Johnson et al., 1971), and the Derjaguin-Muller-Toporov (DMT) model (Derjaguin et al., 1975), we obtain an "effective" work of adhesion between 12 and 16 J/m2. For this calculation, we used a value of 160 ± 20 nm for the tip radius, which was measured experimentally by scanning over sharp edges of a faceted SrTiO3(305) sample (Carpick et al., 1996a).

Fig. 9. Force-displacement curve for a tungsten carbide tip in contact with a clean Pt(111) sample. The plot is shown for the retracting portion only. The load appears to increase when retraction starts, even thought the tip-sample separation is increasing. This is due to the strong adhesion of the tip to the surface, which prevents sliding. Consequently, the tip pivots about the contact point. After that, the load decreases down to the pull-off point, as is usual in force-displacement curves. The deformations of the cantilever giving rise to this behavior are illustrated in Fig. 10.

This adhesion energy is likely an overestimate because we have neglected the possibility of plastic failure of the junction between the tip and sample. It is extremely difficult to apply such a model to this contact without knowing more about the contact geometry or the species at the interface. In any event, this effective work of adhesion is three orders of magnitude higher than that found in previous UHV AFM measurements (Carpick & Salmeron, 1997), such as 0.02 J/m2 between silicon nitride AFM tips and the muscovite mica surface in UHV, or 0.4 J/m2 for a Pt tip on a mica surface in UHV (Carpick et al., 1996b). Note that the surface energy of most metals is in the range of 1 to 5 J/m2, i.e., 2 to 10 J/m2 are required to split an ideal crystal in half to create two new surfaces (Israelachvili, 1992). Our value of 16 J/m2 is beyond this range, consistent with the notion that we have likely overestimated the adhesion by assuming elastic contact. Nevertheless, our measurement indicates that we have observed extremely strong adhesion for this pair of materials. This indicates that strong bonds, at least several eV per atom suggestive of covalent bonds, are formed between the clean Pt(111) surface and the tungsten carbide tip.

severe damage. These contacts were characterized by means of force-displacement curves, as in the example shown in Fig. 9, where a pull-off force of *Lc* = 12.0 ± 1.2 µN was measured with the cantilever of 88 N/m normal spring constant. Assuming, for simplicity, an *elastic* adhesive contact, this force can be related to the work of adhesion of the interface. Within the extremes of the Johnson-Kendall-Roberts (JKR) model (Johnson et al., 1971), and the Derjaguin-Muller-Toporov (DMT) model (Derjaguin et al., 1975), we obtain an "effective" work of adhesion between 12 and 16 J/m2. For this calculation, we used a value of 160 ± 20 nm for the tip radius, which was measured experimentally by scanning over sharp edges of

**-500 -400 -300 -200 -100 0**

**A**

**Tip Height [nm]**

Fig. 9. Force-displacement curve for a tungsten carbide tip in contact with a clean Pt(111) sample. The plot is shown for the retracting portion only. The load appears to increase when retraction starts, even thought the tip-sample separation is increasing. This is due to the strong adhesion of the tip to the surface, which prevents sliding. Consequently, the tip pivots about the contact point. After that, the load decreases down to the pull-off point, as is usual in force-displacement curves. The deformations of the cantilever giving rise to this

This adhesion energy is likely an overestimate because we have neglected the possibility of plastic failure of the junction between the tip and sample. It is extremely difficult to apply such a model to this contact without knowing more about the contact geometry or the species at the interface. In any event, this effective work of adhesion is three orders of magnitude higher than that found in previous UHV AFM measurements (Carpick & Salmeron, 1997), such as 0.02 J/m2 between silicon nitride AFM tips and the muscovite mica surface in UHV, or 0.4 J/m2 for a Pt tip on a mica surface in UHV (Carpick et al., 1996b). Note that the surface energy of most metals is in the range of 1 to 5 J/m2, i.e., 2 to 10 J/m2 are required to split an ideal crystal in half to create two new surfaces (Israelachvili, 1992). Our value of 16 J/m2 is beyond this range, consistent with the notion that we have likely overestimated the adhesion by assuming elastic contact. Nevertheless, our measurement indicates that we have observed extremely strong adhesion for this pair of materials. This indicates that strong bonds, at least several eV per atom suggestive of covalent bonds, are formed between the clean Pt(111) surface

a faceted SrTiO3(305) sample (Carpick et al., 1996a).

**-15**

**-10**

**-5**

**Force [**

behavior are illustrated in Fig. 10.

and the tungsten carbide tip.

**0**

**5**

**10**

Fig. 10. Schematic drawing illustrating the deformation of a cantilever subject to high friction forces. (a) The cantilever is shown at the upper left for the case of no load. The dashed line represents the trajectory of the laser for the zero load case, and is included in all subsequent sketches. (b) After snapping into contact, the *z*-displacement is increased. Normally this would cause an increase in the laser signal (after the initial decrease due to the snap-in). But large friction forces cause bending at the end of the lever in the opposing direction, and producing an apparent *decrease* in the load. Friction is preventing the tip from sliding relative to the surface. (c) As the *z*-displacement is further increased, the friction force eventually reaches its limiting value. The tip will begin to slide relative to the surface, and the upward bending induced by the *z*-displacement overtakes the tendency to bend in the opposite direction induced by the friction force, so now the apparent load begins to increase. Eventually the *z*-displacement reaches a maximum value. (d) As the zdisplacement direction is reversed, the friction force will now resist motion in the opposite direction, and so the bending it induces causes an apparent increase in the load. During this phase, the tip is not sliding relative to the surface. (e) Eventually the friction force reaches a limiting value and once again the tip begins to slide. The apparent load will now decrease as the *z*-displacement is decreased. Stages (d) and (e) are clearly evident in Fig. 9.

In addition, an unusual hysteresis feature in the force-displacement plots was observed at the largest z-displacements as seen in Fig. 9. The recorded data begins with the tip initially in contact after being pushed back by approximately 450 nm. As the cantilever is retracted, the apparent force on the cantilever *increases*, then eventually begins to decrease, as one would normally expect. This result can be explained by considering the effect of friction on the cantilever bending due to the tilted geometry with respect to the plane of the sample (22.5° in this case). Friction causes the cantilever to bend in addition to bending due to the normal force between the tip and sample. The direction of the bending will depend on the

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 117

**-100 -80 -60 -40 -20 0 20**

**Tip Height [nm]**

Fig. 11. A force-displacement plot for the same lever as in Fig. 7, but taken over a passivated area of the Pt(111) sample. The adhesion force is much lower, and the unusual behavior in

Topography, friction, and point contact current (corresponding to contact conductance) were recorded simultaneously as the tip was scanned over the surface, (Enachescu et al., 1999b) as shown in Fig. 12(a,b). The friction and conductivity maps exhibit a strong correlation, with regions of high and low friction corresponding to regions of high and low electrical conductivity respectively. We propose that high friction and high local conductivity are associated with "cleaner" interfaces, while low friction and low local conductivity correspond to regions covered with more interfacial adsorbates. The spatial distribution of friction and conductance values remained consistent throughout several images. This indicates that the tip was not changing during the image acquisition, but

rather, different regions of the sample had different amounts of adsorbates present.

**3.3.3 Contacts with half-passivated interfaces: Atomic lattice resolution images** 

In contrast with the fully passivated interfaces, when the Pt surface is clean, we could frequently observe stick-slip behavior with the atomic-lattice periodicity of the Pt(111) substrate, as shown in the image of Fig. 13(a) and the trace of a friction line in Fig. 13(c). The Fourier transform of the image (Fig. 13(b)) shows more clearly the 3-fold symmetry with 0.27 nm periodicity, in agreement with the lattice constant of Pt(111). The image was acquired under zero externally applied load. The occurrence of stick-slip behavior was always

Typical I-V characteristics obtained with contacts in areas with different degree of passivation are shown in the graphs of Fig. 12(c). We acquired current-voltage (I-V) curves in a 16 x 16 grid, while for each I-V curve the voltage was scanned from -50 mV to +50 mV. Ohmic behavior curves (straight lines) were always observed in the regions exhibiting high friction, while semiconductor-like behavior (sigmoid shapes) were observed in low friction areas. This observation can be understood on the basis of the poor conductivity of the contamination layers present, which decreases as their quantity, and thus passivation capacity, increases.

**-0.8 -0.6 -0.4 -0.2**

the plot due to friction forces is not present.

**0 0.2 0.4 0.6 0.8**

**Force [**

direction of the friction force, which always acts to oppose any tendency for sliding. These two forces compete by changing the slope of the cantilever in opposite directions. This is illustrated in Fig. 10. Because of the strong bonding, the tip cannot slide over the surface and the cantilever is forced to adopt an S-shape like the one shown in Fig. 10(c). As the sample is retracted and the lever reverts to its normal bending shape (shown in Fig. 10(d)), it produces an apparent increase in the force initially. After passing through a maximum (point A in Fig. 9), the force decreases as expected. This effect is explained by the strong friction force on a cantilever fixed at one end and with a tilted geometry with respect to the plane of the sample (22.5° in this case). The slope of the force-displacement curve is inverted because static friction prevents the tip from sliding with respect to the surface. Instead the tip is pivoting about the contact point, and the slope of the end of the cantilever is increasing. Eventually the tip pivots enough that the maximum static friction force is reached and the tip can slide relative to the surface. Stages (d) and (e) sketched in Fig. 10 are evident in Fig. 9.

The fact that this hysteresis phenomenon is observed for a very stiff lever, *i.e.* a spring constant with two to three orders of magnitude higher than the typical contact AFM levers (88 N/m), indicates that strong friction forces are occurring in tandem with the strong adhesion forces. In addition, our experiments show a much stronger interaction between the AFM tip and the metallic surface compared to the results published by Bennewitz *et. al.* (Bennewitz et al., 1999; 2001). Those measurements involved silicon AFM levers having a spring constant of 0.024 N/m in contact with clean a Cu(111) surface under UHV conditions.

#### **3.3.2 Contacts between fully passivated interfaces: Friction and conductance measurements**

As we have seen with clean tips and clean surfaces, contact-mode measurements cannot be performed without severely disrupting the contact region. To perform contact experiments while avoiding strong modifications, the surfaces must be chemically passivated. This can be achieved intentionally or unintentionally by the presence of adsorbate layers. In our case we used the unintentionally passivation, provided without effort by the contaminant's presence, i.e., the contamination behaved like a "passivation" layer for the tip-sample interaction. An interesting question is whether these layers must be present on each or on only one of the surfaces for substantial passivation. The latter case implies that the layers are attached strongly to one of the surfaces and interact only weakly with the other, such that the contact can shear at this weak interface. As we will show in this and the next section, it is indeed possible to have both situations, which we shall call fully passivation when layers of material are present on each contacting surface, and half passivation, when one of the two surfaces remains clean, during and after friction scanning.

 On the Pt surface, the most common contaminant after annealing in UHV is carbon, as verified with AES. On the WC tip, in addition to oxygen present as a tungsten oxide, adventitious hydrocarbon or graphitic carbon can also accumulate. Ex-situ AES on the body of the cantilever and on the tip shaft did indeed reveal the presence of O, C and W as the only observable constituents. We will first examine results where the Pt surface is covered by a layer of C-contamination, the only impurity element detected in the Auger spectra. Force-displacement data obtained with such passivated surfaces show low adhesion values in the range of ~1 J/m2, depending on the spatial location of the tip over the surface, as shown in the example of Fig. 11.

direction of the friction force, which always acts to oppose any tendency for sliding. These two forces compete by changing the slope of the cantilever in opposite directions. This is illustrated in Fig. 10. Because of the strong bonding, the tip cannot slide over the surface and the cantilever is forced to adopt an S-shape like the one shown in Fig. 10(c). As the sample is retracted and the lever reverts to its normal bending shape (shown in Fig. 10(d)), it produces an apparent increase in the force initially. After passing through a maximum (point A in Fig. 9), the force decreases as expected. This effect is explained by the strong friction force on a cantilever fixed at one end and with a tilted geometry with respect to the plane of the sample (22.5° in this case). The slope of the force-displacement curve is inverted because static friction prevents the tip from sliding with respect to the surface. Instead the tip is pivoting about the contact point, and the slope of the end of the cantilever is increasing. Eventually the tip pivots enough that the maximum static friction force is reached and the tip can slide relative to the surface. Stages (d) and (e) sketched in Fig. 10 are evident in Fig. 9.

The fact that this hysteresis phenomenon is observed for a very stiff lever, *i.e.* a spring constant with two to three orders of magnitude higher than the typical contact AFM levers (88 N/m), indicates that strong friction forces are occurring in tandem with the strong adhesion forces. In addition, our experiments show a much stronger interaction between the AFM tip and the metallic surface compared to the results published by Bennewitz *et. al.* (Bennewitz et al., 1999; 2001). Those measurements involved silicon AFM levers having a spring constant of 0.024 N/m in contact with clean a Cu(111) surface under UHV conditions.

As we have seen with clean tips and clean surfaces, contact-mode measurements cannot be performed without severely disrupting the contact region. To perform contact experiments while avoiding strong modifications, the surfaces must be chemically passivated. This can be achieved intentionally or unintentionally by the presence of adsorbate layers. In our case we used the unintentionally passivation, provided without effort by the contaminant's presence, i.e., the contamination behaved like a "passivation" layer for the tip-sample interaction. An interesting question is whether these layers must be present on each or on only one of the surfaces for substantial passivation. The latter case implies that the layers are attached strongly to one of the surfaces and interact only weakly with the other, such that the contact can shear at this weak interface. As we will show in this and the next section, it is indeed possible to have both situations, which we shall call fully passivation when layers of material are present on each contacting surface, and half passivation, when one of the two

 On the Pt surface, the most common contaminant after annealing in UHV is carbon, as verified with AES. On the WC tip, in addition to oxygen present as a tungsten oxide, adventitious hydrocarbon or graphitic carbon can also accumulate. Ex-situ AES on the body of the cantilever and on the tip shaft did indeed reveal the presence of O, C and W as the only observable constituents. We will first examine results where the Pt surface is covered by a layer of C-contamination, the only impurity element detected in the Auger spectra. Force-displacement data obtained with such passivated surfaces show low adhesion values in the range of ~1 J/m2, depending on the spatial location of the tip over the surface, as

**3.3.2 Contacts between fully passivated interfaces: Friction and conductance** 

surfaces remains clean, during and after friction scanning.

shown in the example of Fig. 11.

**measurements** 

Fig. 11. A force-displacement plot for the same lever as in Fig. 7, but taken over a passivated area of the Pt(111) sample. The adhesion force is much lower, and the unusual behavior in the plot due to friction forces is not present.

Topography, friction, and point contact current (corresponding to contact conductance) were recorded simultaneously as the tip was scanned over the surface, (Enachescu et al., 1999b) as shown in Fig. 12(a,b). The friction and conductivity maps exhibit a strong correlation, with regions of high and low friction corresponding to regions of high and low electrical conductivity respectively. We propose that high friction and high local conductivity are associated with "cleaner" interfaces, while low friction and low local conductivity correspond to regions covered with more interfacial adsorbates. The spatial distribution of friction and conductance values remained consistent throughout several images. This indicates that the tip was not changing during the image acquisition, but rather, different regions of the sample had different amounts of adsorbates present.

Typical I-V characteristics obtained with contacts in areas with different degree of passivation are shown in the graphs of Fig. 12(c). We acquired current-voltage (I-V) curves in a 16 x 16 grid, while for each I-V curve the voltage was scanned from -50 mV to +50 mV. Ohmic behavior curves (straight lines) were always observed in the regions exhibiting high friction, while semiconductor-like behavior (sigmoid shapes) were observed in low friction areas. This observation can be understood on the basis of the poor conductivity of the contamination layers present, which decreases as their quantity, and thus passivation capacity, increases.

#### **3.3.3 Contacts with half-passivated interfaces: Atomic lattice resolution images**

In contrast with the fully passivated interfaces, when the Pt surface is clean, we could frequently observe stick-slip behavior with the atomic-lattice periodicity of the Pt(111) substrate, as shown in the image of Fig. 13(a) and the trace of a friction line in Fig. 13(c). The Fourier transform of the image (Fig. 13(b)) shows more clearly the 3-fold symmetry with 0.27 nm periodicity, in agreement with the lattice constant of Pt(111). The image was acquired under zero externally applied load. The occurrence of stick-slip behavior was always

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 119

Since the friction force is approximately 190 nN, the energy dissipated after a displacement of one unit cell (~3 Å) is about 350 eV. Given the tip radius of 160 nm, and a total load of ~0.8 N (see Fig. 11) a contact area of roughly 100 nm2 can be calculated containing approximately 1000 atoms. This corresponds to an average energy dissipated per atom of 0.3 eV. We would predict that this energy is not enough to break the strong chemical bonds of the Pt atoms, and indeed

(a) (b)

(c)

In most cases the adhesion between the AFM tip and the Pt(111) was so strong that we were not able to scan the tip over the sample and often the fracture of low spring-constant AFM cantilevers was observed. Reproducible scanning over the Pt(111) surface was only possible

when a small amount of contaminant was present between the tip and the sample.

Fig. 13. (a) 2.5 x 2.5 nm2 lateral force image obtained with a passivated tip and a clean Pt(111) sample. Atomic-lattice stick-slip friction is clearly observed and provides proof that sliding is taking place along a clean Pt surface; (b) Fast Fourier transform of the data in (a), showing the periodicity and symmetry of the Pt(111) surface; (c) line trace of the line

indicated in (a) showing the clear stick-slip behavior.

**3.3.4 Tunneling experiments using the AFM tip** 

that is what we observe.

Fig. 12. Simultaneous friction (a) and point contact current (b) images of a Pt(111) surface acquired with a conductive WC-coated Si cantilever of 88 N/m spring constant. Image size is 500 x 500 nm2. Regions with high and low friction are clearly correlated to regions of high and low local conductivity. (c) Corresponding *I-V* spectra acquired at the points of "clean" and "passivated" areas. The bias between tip and sample was varied from -50 mV to +50 mV. High friction regions are correlated with Ohmic conductance behavior, while the lower friction regions exhibit non-Ohmic conductance, indicative of an insulating or semiconducting interlayer.

associated with the presence of low adhesion, low friction, and low contact current, indicative of chemically inactive tip. These atomic stick-slip images were recorded by carefully choosing flat terrace locations. We did not scan areas that included steps, in order to avoid the increased reactivity of the step sites.

This result indicates that the passivating layers are on the tip side of the interface, where they are bound strongly enough to withstand the applied shear stresses without transference to the Pt surface. In contrast to these unintentional adsorbates used as passivation layer, other intentional, well-defined adsorbates may not be well bound as to the WC tip and thus, may not provide any additional understanding to the tip-metal interaction we described in this work. An analysis of the energy balance during friction is of interest here.

(a) (b)

**-0.6 -0.4 -0.2 0 0.2 0.4 0.6**

**Voltage [V]** (c) Fig. 12. Simultaneous friction (a) and point contact current (b) images of a Pt(111) surface acquired with a conductive WC-coated Si cantilever of 88 N/m spring constant. Image size is 500 x 500 nm2. Regions with high and low friction are clearly correlated to regions of high and low local conductivity. (c) Corresponding *I-V* spectra acquired at the points of "clean" and "passivated" areas. The bias between tip and sample was varied from -50 mV to +50 mV. High friction regions are correlated with Ohmic conductance behavior, while the lower

**-60 -40 -20**

friction regions exhibit non-Ohmic conductance, indicative of an insulating or

associated with the presence of low adhesion, low friction, and low contact current, indicative of chemically inactive tip. These atomic stick-slip images were recorded by carefully choosing flat terrace locations. We did not scan areas that included steps, in order

This result indicates that the passivating layers are on the tip side of the interface, where they are bound strongly enough to withstand the applied shear stresses without transference to the Pt surface. In contrast to these unintentional adsorbates used as passivation layer, other intentional, well-defined adsorbates may not be well bound as to the WC tip and thus, may not provide any additional understanding to the tip-metal interaction we described in this work. An analysis of the energy balance during friction is of interest here.

**Current [**

to avoid the increased reactivity of the step sites.

semiconducting interlayer.

Since the friction force is approximately 190 nN, the energy dissipated after a displacement of one unit cell (~3 Å) is about 350 eV. Given the tip radius of 160 nm, and a total load of ~0.8 N (see Fig. 11) a contact area of roughly 100 nm2 can be calculated containing approximately 1000 atoms. This corresponds to an average energy dissipated per atom of 0.3 eV. We would predict that this energy is not enough to break the strong chemical bonds of the Pt atoms, and indeed that is what we observe.

Fig. 13. (a) 2.5 x 2.5 nm2 lateral force image obtained with a passivated tip and a clean Pt(111) sample. Atomic-lattice stick-slip friction is clearly observed and provides proof that sliding is taking place along a clean Pt surface; (b) Fast Fourier transform of the data in (a), showing the periodicity and symmetry of the Pt(111) surface; (c) line trace of the line indicated in (a) showing the clear stick-slip behavior.

#### **3.3.4 Tunneling experiments using the AFM tip**

In most cases the adhesion between the AFM tip and the Pt(111) was so strong that we were not able to scan the tip over the sample and often the fracture of low spring-constant AFM cantilevers was observed. Reproducible scanning over the Pt(111) surface was only possible when a small amount of contaminant was present between the tip and the sample.

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 121

Fig. 15. STM image of the Pt(111) surface after AFM mechanical contact has been made between the tip and sample. The 200 x 200 nm2 STM image was acquired with a tunneling bias of -0.2 V and a tunneling current of 160 pA, while the I-V converter was operating at a

(a) (b)

(c) Fig. 16. (a) 100 x 100 nm2 STM image of a highly stepped region of the Pt(111) surface. The image was acquired under the following tunneling conditions: bias = -0.2 V; current = 200 pA; (b) 10x10 nm2 AFM friction force image acquired in contact at the center of (a); (c) Subsequent STM image, acquired under the same conditions as in (a), showing substantial

material deposition from the AFM tip during the previous contact

108 gain.

To further investigate the Pt(111) surface we performed non-contact scanning tunneling microscopy (STM) experiments using the conductive WC AFM tip. For these experiments the AFM cantilever must be sufficiently stiff, otherwise the cantilever will jump into contact (Carpick & Salmeron, 1997; Enachescu et al., 1998) before or during tunneling conditions at small physical gaps, *e.g.,* 0.7-1.0 nm. We used an 88 N/m stiff lever, which was enough to avoid the jump to contact.

Fig. 14 shows an STM image acquired with an AFM lever prior to any tip-sample contact. The 200 nm x 200 nm image was acquired with a bias of 0.1 V (sample negative) and a tunneling current of 160 pA. Several monoatomic steps, with a height of 0.22 ± 0.01 nm, and terraces are visible. A few isolated protrusions are also observed on the terraces and also attached to the step edges. They correspond probably to contamination. The round flat islands attached to the steps have a height of 0.1 nm and clearly pin the steps, which would have been flowing during the high temperature anneal.

Fig. 14. 200 x 200 nm2 STM image of the Pt(111) surface taken *before* any contact has been made between the tip and sample. A series of single atom steps, 0.22 ± 0.01 nm in height, are observed. Two of the steps are seen to be pinned by contamination, which is likely carbon that has diffused to the surface region from the bulk during processing of the crystal. The image was acquired for an I-V converter gain of 108 under the following tunneling condition: tunneling bias of 0.1 V (sample negative) and a tunneling current of 160 pA.

Fig. 15 shows a 200 x 200 nm2 STM image acquired with the AFM lever *after* mechanical contact has been made between the tip and sample. Specifically, the tip was brought into contact with the sample and then retracted. The tip was not scanned during this contact. The tip was then brought back to within tunneling range and used to acquire the STM image with a bias of -0.2 V and a tunneling current of 160 pA. Large, irregular features up to 13.7 nm in height are seen covering roughly 75% of the image. In the topmost 25% of the image Pt(111) steps can still be seen. This image demonstrates that the bonding of the tip to the surface was very strong, and that rupture of the contact occurred within the tip material itself. This material is then left over the Pt surface.

Another example of STM imaging followed by contact and AFM imaging and then again STM imaging is shown in Fig. 16. The 100 nm x 100 nm image in (a) was acquired at a bias of -0.2 V and tunneling current of 200 pA and shows a stepped region of the Pt(111) surface

To further investigate the Pt(111) surface we performed non-contact scanning tunneling microscopy (STM) experiments using the conductive WC AFM tip. For these experiments the AFM cantilever must be sufficiently stiff, otherwise the cantilever will jump into contact (Carpick & Salmeron, 1997; Enachescu et al., 1998) before or during tunneling conditions at small physical gaps, *e.g.,* 0.7-1.0 nm. We used an 88 N/m stiff lever, which was enough to

Fig. 14 shows an STM image acquired with an AFM lever prior to any tip-sample contact. The 200 nm x 200 nm image was acquired with a bias of 0.1 V (sample negative) and a tunneling current of 160 pA. Several monoatomic steps, with a height of 0.22 ± 0.01 nm, and terraces are visible. A few isolated protrusions are also observed on the terraces and also attached to the step edges. They correspond probably to contamination. The round flat islands attached to the steps have a height of 0.1 nm and clearly pin the steps, which would

Fig. 14. 200 x 200 nm2 STM image of the Pt(111) surface taken *before* any contact has been made between the tip and sample. A series of single atom steps, 0.22 ± 0.01 nm in height, are observed. Two of the steps are seen to be pinned by contamination, which is likely carbon that has diffused to the surface region from the bulk during processing of the crystal. The image was acquired for an I-V converter gain of 108 under the following tunneling condition: tunneling bias of 0.1 V (sample negative) and a tunneling current of 160 pA.

Fig. 15 shows a 200 x 200 nm2 STM image acquired with the AFM lever *after* mechanical contact has been made between the tip and sample. Specifically, the tip was brought into contact with the sample and then retracted. The tip was not scanned during this contact. The tip was then brought back to within tunneling range and used to acquire the STM image with a bias of -0.2 V and a tunneling current of 160 pA. Large, irregular features up to 13.7 nm in height are seen covering roughly 75% of the image. In the topmost 25% of the image Pt(111) steps can still be seen. This image demonstrates that the bonding of the tip to the surface was very strong, and that rupture of the contact occurred within the tip material

Another example of STM imaging followed by contact and AFM imaging and then again STM imaging is shown in Fig. 16. The 100 nm x 100 nm image in (a) was acquired at a bias of -0.2 V and tunneling current of 200 pA and shows a stepped region of the Pt(111) surface

avoid the jump to contact.

have been flowing during the high temperature anneal.

itself. This material is then left over the Pt surface.

Fig. 15. STM image of the Pt(111) surface after AFM mechanical contact has been made between the tip and sample. The 200 x 200 nm2 STM image was acquired with a tunneling bias of -0.2 V and a tunneling current of 160 pA, while the I-V converter was operating at a 108 gain.

Fig. 16. (a) 100 x 100 nm2 STM image of a highly stepped region of the Pt(111) surface. The image was acquired under the following tunneling conditions: bias = -0.2 V; current = 200 pA; (b) 10x10 nm2 AFM friction force image acquired in contact at the center of (a); (c) Subsequent STM image, acquired under the same conditions as in (a), showing substantial material deposition from the AFM tip during the previous contact

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 123

(Smith et al., 1986) performed an experiment imaging the HOPG surface by purposely placing the tip in contact with the surface. In contact, the situation is similar to that in pointcontact spectroscopy (Yanson et al., 1981). This mode of microscopy was called point-contact microscopy (PCM) (Smith et al., 1986), which differs from STM in that the tip is much closer to the sample in the region where the potential barrier is significantly reduced and tipsample forces are repulsive. In this mode, Smith *et al.* succeeded in imaging the HOPG lattice by measuring the current flowing through the contact. However, they were able to report lattice resolution only at low temperatures, *i.e.,* when the microscope was immersed in liquid helium. Since their contact area involved a large number of atoms, they explained the "atomic" resolution by considering the conduction to be due to a single atom on the tip, which we now consider unlikely. Other experimental evidence indicates that true tunneling through a vacuum gap might not occur in the case of graphite in normal circumstances, and

In AFM contact mode, lattice resolution can be obtained both in topography and friction channels. This is usually explained as the result of stick-slip phenomena (Marti et al., 1987). AFM measurements involving a conductive lever have been reported (Lantz et al., 1997a, 1997b; Enachescu et al., 1998) in conjunction with tip-sample contact area evaluation. To date, there have been no reports of AFM contact experiments on graphite deposited on

This work presents results on the simultaneous implementation of AFM and PCM techniques by using a conductive AFM lever. We demonstrate the possibility of obtaining lattice resolution concurrently in three channels: topography, friction, and contact current. This is achieved by using both HOPG and 1 ML of graphite deposited on a Pt(111) singlecrystal. We show that PCM is as capable of similar lateral resolution as contact AFM imaging. We also found that PCM is sensitive to local conductivity variations due to moiré superstructures resulting from the high order commensurability of the graphite and Pt lattices at different relative rotations. Moreover, we show that lattice resolution in PCM mode is achievable for currents of up to 0.9 mA and contact pressures estimated at 5 GPa.

All experiments were performed in a UHV chamber (base pressure 7 x 10-11 Torr) equipped with AFM, Auger Electron Spectroscopy (AES) and Low-energy Electron Diffraction (LEED) (Dai et al., 1995). Two different samples, HOPG and Pt(111), were used. The HOPG sample was cleaved along the (0001) plane in air and then immediately placed in the vacuum

The samples could be heated by means of electron bombardment from a hot dispenser cathode. The Pt sample was prepared using standard procedures of Argon ion bombardment, oxygen treatment, and annealing until a clean and ordered surface was produced, as verified by AES and LEED. The clean surface was then exposed to ethylene at room temperature by backfilling the chamber with ethylene. Exposures were typically greater than 10 Langmuir to ensure saturation of the Pt(111) surface. After exposure, the sample was heated to about 1250K, resulting in the decomposition of ethylene and formation of a single monolayer of graphite on the Pt(111) surface. When observed with LEED, we found that the graphite layer produced characteristic fragmented rings (Hu et al.,

that the tip is in contact with the surface (Salmeron et al., 1991).

metals.

**4.2 Experimental** 

chamber.

before contact. Fig. 16(b) shows a 10x10 nm2 AFM contact friction force image at zero externally applied load acquired in the center of Fig. 16(a). During imaging the bias was held at zero volts. High friction forces are observed in this image. The tip was then withdrawn from contact and another STM image acquired, shown in Fig. 16(c), under identical tunneling conditions as Fig. 16(a). It is obvious from Fig. 16(c) that substantial material deposition from the AFM tip during contact scanning process has taken place.

Using the friction force during slip of several N, the energy dissipated during imaging can be estimated to be of the order of 108 eV. The amount of material left on the surface (Fig. 16(c)) occupies an area of roughly 2,500 nm2. The energy dissipated during friction is therefore sufficient to break the junction, even if very strong (~5 eV) bonds need to be broken.

## **4. Local conductivity in nanocontacts: The integration of point-contact microscopy**

The electrical current through the point-contact junction of an AFM tip was used to image the surfaces of bulk graphite (HOPG) and the surface of a graphitized carbon monolayer on Pt(111) under ultra-high vacuum (UHV) conditions. Lattice-resolved images were obtained simultaneously in topography, lateral friction, and contact current channels. Lattice resolution in current maps persisted up to 0.9 mA and pressures of up to 5 GPa. In both bulk graphite and the case of graphitized carbon monolayer on Pt(111), the current images showed only one maximum per unit cell. In addition, the contact current images of the graphite monolayer revealed local conductivity variations. We observed local conductivity variations in the form of moiré superstructures resulting from high order commensurability with the Pt lattice.

#### **4.1 Background**

Since the invention of the scanning tunneling microscope (STM) (Binnig et al., 1983), graphite, specifically highly-oriented pyrolytic graphite (HOPG), has become a popular substrate due to its flat cleavage surface and its inert nature, which makes it possible to obtain images in air with "atomic resolution" (Hansma, 1985) (Binnig et al., 1986b). However, the literature reports a number of well-known puzzling features, such as uncharacteristically large corrugation amplitudes (Binnig et al., 1986; Selloni et al., 1985; 1986; Tersoff, 1986; Batra & Ciraci, 1988; Batra et al., 1987; Soler et al., 1986), enhanced lateral resolution (Binnig et al., 1986; Selloni et al., 1985; 1986; Tersoff, 1986; Batra & Ciraci, 1988; Batra et al., 1987; Soler et al., 1986; Park & Quate, 1986), a weak dependence of the tunneling current on the position of the tip in the direction perpendicular to the surface (Salmeron et al., 1991), and anomalously large superperiodicities (Kuwabara et al., 1990). These features generated a debate about the imaging mechanism. In most STM images, one observes only one maximum per unit cell, indicating that the carbon atoms are not imaged as individual units. In a favored explanation, the lattice periodicity is due to the tip imaging a single electron state of the graphite layer (Tersoff, 1986). STM images taken on one monolayer of graphite deposited on metals also show only the lattice periodicity, and not single atomic positions (Land et al., 1992a; 1992b).

In the debate concerning the imaging mechanism of HOPG in STM, it was suggested that the STM tip could be in contact with the HOPG. In order to clarify this issue, Smith *et al.*  (Smith et al., 1986) performed an experiment imaging the HOPG surface by purposely placing the tip in contact with the surface. In contact, the situation is similar to that in pointcontact spectroscopy (Yanson et al., 1981). This mode of microscopy was called point-contact microscopy (PCM) (Smith et al., 1986), which differs from STM in that the tip is much closer to the sample in the region where the potential barrier is significantly reduced and tipsample forces are repulsive. In this mode, Smith *et al.* succeeded in imaging the HOPG lattice by measuring the current flowing through the contact. However, they were able to report lattice resolution only at low temperatures, *i.e.,* when the microscope was immersed in liquid helium. Since their contact area involved a large number of atoms, they explained the "atomic" resolution by considering the conduction to be due to a single atom on the tip, which we now consider unlikely. Other experimental evidence indicates that true tunneling through a vacuum gap might not occur in the case of graphite in normal circumstances, and that the tip is in contact with the surface (Salmeron et al., 1991).

In AFM contact mode, lattice resolution can be obtained both in topography and friction channels. This is usually explained as the result of stick-slip phenomena (Marti et al., 1987). AFM measurements involving a conductive lever have been reported (Lantz et al., 1997a, 1997b; Enachescu et al., 1998) in conjunction with tip-sample contact area evaluation. To date, there have been no reports of AFM contact experiments on graphite deposited on metals.

This work presents results on the simultaneous implementation of AFM and PCM techniques by using a conductive AFM lever. We demonstrate the possibility of obtaining lattice resolution concurrently in three channels: topography, friction, and contact current. This is achieved by using both HOPG and 1 ML of graphite deposited on a Pt(111) singlecrystal. We show that PCM is as capable of similar lateral resolution as contact AFM imaging. We also found that PCM is sensitive to local conductivity variations due to moiré superstructures resulting from the high order commensurability of the graphite and Pt lattices at different relative rotations. Moreover, we show that lattice resolution in PCM mode is achievable for currents of up to 0.9 mA and contact pressures estimated at 5 GPa.

#### **4.2 Experimental**

122 Atomic Force Microscopy – Imaging, Measuring and Manipulating Surfaces at the Atomic Scale

before contact. Fig. 16(b) shows a 10x10 nm2 AFM contact friction force image at zero externally applied load acquired in the center of Fig. 16(a). During imaging the bias was held at zero volts. High friction forces are observed in this image. The tip was then withdrawn from contact and another STM image acquired, shown in Fig. 16(c), under identical tunneling conditions as Fig. 16(a). It is obvious from Fig. 16(c) that substantial material deposition from the AFM tip during contact scanning process has taken place.

Using the friction force during slip of several N, the energy dissipated during imaging can be estimated to be of the order of 108 eV. The amount of material left on the surface (Fig. 16(c)) occupies an area of roughly 2,500 nm2. The energy dissipated during friction is therefore

The electrical current through the point-contact junction of an AFM tip was used to image the surfaces of bulk graphite (HOPG) and the surface of a graphitized carbon monolayer on Pt(111) under ultra-high vacuum (UHV) conditions. Lattice-resolved images were obtained simultaneously in topography, lateral friction, and contact current channels. Lattice resolution in current maps persisted up to 0.9 mA and pressures of up to 5 GPa. In both bulk graphite and the case of graphitized carbon monolayer on Pt(111), the current images showed only one maximum per unit cell. In addition, the contact current images of the graphite monolayer revealed local conductivity variations. We observed local conductivity variations in the form of moiré superstructures resulting from high order commensurability

Since the invention of the scanning tunneling microscope (STM) (Binnig et al., 1983), graphite, specifically highly-oriented pyrolytic graphite (HOPG), has become a popular substrate due to its flat cleavage surface and its inert nature, which makes it possible to obtain images in air with "atomic resolution" (Hansma, 1985) (Binnig et al., 1986b). However, the literature reports a number of well-known puzzling features, such as uncharacteristically large corrugation amplitudes (Binnig et al., 1986; Selloni et al., 1985; 1986; Tersoff, 1986; Batra & Ciraci, 1988; Batra et al., 1987; Soler et al., 1986), enhanced lateral resolution (Binnig et al., 1986; Selloni et al., 1985; 1986; Tersoff, 1986; Batra & Ciraci, 1988; Batra et al., 1987; Soler et al., 1986; Park & Quate, 1986), a weak dependence of the tunneling current on the position of the tip in the direction perpendicular to the surface (Salmeron et al., 1991), and anomalously large superperiodicities (Kuwabara et al., 1990). These features generated a debate about the imaging mechanism. In most STM images, one observes only one maximum per unit cell, indicating that the carbon atoms are not imaged as individual units. In a favored explanation, the lattice periodicity is due to the tip imaging a single electron state of the graphite layer (Tersoff, 1986). STM images taken on one monolayer of graphite deposited on metals also show only the lattice periodicity, and not single atomic

In the debate concerning the imaging mechanism of HOPG in STM, it was suggested that the STM tip could be in contact with the HOPG. In order to clarify this issue, Smith *et al.* 

sufficient to break the junction, even if very strong (~5 eV) bonds need to be broken.

**4. Local conductivity in nanocontacts: The integration of point-contact** 

**microscopy** 

with the Pt lattice.

**4.1 Background** 

positions (Land et al., 1992a; 1992b).

All experiments were performed in a UHV chamber (base pressure 7 x 10-11 Torr) equipped with AFM, Auger Electron Spectroscopy (AES) and Low-energy Electron Diffraction (LEED) (Dai et al., 1995). Two different samples, HOPG and Pt(111), were used. The HOPG sample was cleaved along the (0001) plane in air and then immediately placed in the vacuum chamber.

The samples could be heated by means of electron bombardment from a hot dispenser cathode. The Pt sample was prepared using standard procedures of Argon ion bombardment, oxygen treatment, and annealing until a clean and ordered surface was produced, as verified by AES and LEED. The clean surface was then exposed to ethylene at room temperature by backfilling the chamber with ethylene. Exposures were typically greater than 10 Langmuir to ensure saturation of the Pt(111) surface. After exposure, the sample was heated to about 1250K, resulting in the decomposition of ethylene and formation of a single monolayer of graphite on the Pt(111) surface. When observed with LEED, we found that the graphite layer produced characteristic fragmented rings (Hu et al.,

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 125

(a) (b)

(c) Fig. 17. Simultaneously acquired lattice resolution images of HOPG under UHV conditions: (a) normal lever deflection (with topographical and buckling effects), with a corrugation of 117 pN, corresponding to a height of 33.6 pm; (b) lateral friction image, average force of 0.5 nN and corrugation of 20 nN; (c) PCM image, average current of 945 nA and peak-to-peak corrugation of 160 nA. Image was taken with an applied load of 100 nN without feedback.

the diameter of the area of contact was similarly estimated to be 5.78 nm, which contains approximately 4000 atoms. Here the 0.246 nm graphite lattice periodicity is also clearly revealed. We were able to obtain lattice resolution at currents up to 0.9 mA and high load. The average pressure at high load was approximately 5 GPa, which is less than the theoretical yield stress of Pt (~17 GPa). At pressures higher than 5 GPa and/or currents higher than 0.9 mA, the images were unstable, although the graphite lattice was still visible. As a side note, we found that we were able to obtain lattice resolution almost all the time and immediately in current mode, while lattice resolution was not as readily visible in the topography and/or friction channels. In many cases, the friction was so low that there was no stick-slip present, *i.e.,* the tip moved continuously over the graphite layer. Because of these reasons, we can rule out the atomic stick-slip mechanism as a reason for the lattice

Image size is 2.5 nm x 2.5 nm.

resolution observed in PCM mode.

1987), with several dominant bright segments. Some of the ring segments were in-line with the Pt spots, indicating alignment or near-alignment of the graphite and Pt lattices. Others were at an angle relative to the Pt spots, indicating that the graphite lattice was rotated with respect to the Pt lattice.

We measured derivative Auger spectra of the surface with an RFA-type electron analyzer, using a normally incident electron beam with an energy of 2500 eV and retarding field oscillation amplitude of 7 eV peak-to-peak. The ratio of the peak-to-peak heights of the carbon (275 eV) and platinum (237 eV) AES transitions in the derivative spectrum was found to be about 3.8, independent of the amount and method of ethylene exposure. We attribute this to saturation of the surface once a graphite monolayer is formed, such that no further decomposition of ethylene can take place. Because of this, we concluded that there was 1 ML of graphite on the Pt(111) surface. This conclusion is supported by Land *et al.*  (Land et al., 1992a, 1992b)*,* who determined by STM, under similar preparation conditions, that the deposited graphite layer was 1 ML thick.

All experiments were performed with a silicon cantilever with a spring constant of 3.5 N/m, and coated with a ~20 nm thick conductive tungsten-carbide layer (bulk resistivity ~30 cm). The tips were characterized by Scanning Electron Microscopy (SEM) and AES. Previous UHV-AFM measurements on a Pt(111) sample showed that similar tungsten-carbide coated tips are wear resistant and conductive (Enachescu et al., 1998, 1999b). The tips were treated in UHV immediately prior to the measurements by applying short voltage pulses while in contact and/or by rubbing them at high loads on the surface. We designed and built a flexible I-V converter that allowed us to measure high contact currents by taking measurements spanning the range from pA to mA.

#### **4.3 Results and discussions**

#### **4.3.1 Topography, friction, and current imaging**

Fig. 17 shows three 2.5 nm x 2.5 nm images of the HOPG surface, which were acquired simultaneously. The feedback control was turned off in order to avoid the convolution of topography and friction, and to minimize noise. The images correspond to: (a) normal lever deflection, (b) lateral force or friction, and (c) contact current. A positive bias of 1.0 V was applied to the sample, and the external load during imaging was 100 nN. The average current in (c) was 0.94 A, with a modulation of about 17%. In all three images, the 0.246 nm graphite lattice periodicity is clearly observed. Using the DMT contact mechanics model (Müller et al., 1983; Derjaguin et al., 1975), with a measured pull-off force of 115 nN, we estimate that our contact radius is 4.1 nm, and therefore contains about 2000 atoms. The contact radius calculated here is only approximate, since the Tabor parameter (Greenwood & Johnson, 1998; Tabor, 1977) for this system is 0.67, which indicates that the DMT model is not entirely appropriate. Moreover, none of the analytical contact mechanics models are directly applicable to a non-isotropic material such as graphite (Sridhar et al., 1997).

Similar lattice-resolved images were obtained on 1 ML of graphite on Pt(111) (Gr/Pt(111)). An example of this is shown in Fig. 18. As in the previous case, the 2.5 nm x 2.5 nm images correspond to: (a) normal lever deflection (under feedback-off conditions), (b) lateral force, and (c) contact current. The external load in this case was 300 nN, and the sample bias was 0.5 V. The average current was 52.7 A, with a current modulation of about 2%. In this case,

1987), with several dominant bright segments. Some of the ring segments were in-line with the Pt spots, indicating alignment or near-alignment of the graphite and Pt lattices. Others were at an angle relative to the Pt spots, indicating that the graphite lattice was rotated with

We measured derivative Auger spectra of the surface with an RFA-type electron analyzer, using a normally incident electron beam with an energy of 2500 eV and retarding field oscillation amplitude of 7 eV peak-to-peak. The ratio of the peak-to-peak heights of the carbon (275 eV) and platinum (237 eV) AES transitions in the derivative spectrum was found to be about 3.8, independent of the amount and method of ethylene exposure. We attribute this to saturation of the surface once a graphite monolayer is formed, such that no further decomposition of ethylene can take place. Because of this, we concluded that there was 1 ML of graphite on the Pt(111) surface. This conclusion is supported by Land *et al.*  (Land et al., 1992a, 1992b)*,* who determined by STM, under similar preparation conditions,

All experiments were performed with a silicon cantilever with a spring constant of 3.5 N/m, and coated with a ~20 nm thick conductive tungsten-carbide layer (bulk resistivity ~30 cm). The tips were characterized by Scanning Electron Microscopy (SEM) and AES. Previous UHV-AFM measurements on a Pt(111) sample showed that similar tungsten-carbide coated tips are wear resistant and conductive (Enachescu et al., 1998, 1999b). The tips were treated in UHV immediately prior to the measurements by applying short voltage pulses while in contact and/or by rubbing them at high loads on the surface. We designed and built a flexible I-V converter that allowed us to measure high contact currents by taking

Fig. 17 shows three 2.5 nm x 2.5 nm images of the HOPG surface, which were acquired simultaneously. The feedback control was turned off in order to avoid the convolution of topography and friction, and to minimize noise. The images correspond to: (a) normal lever deflection, (b) lateral force or friction, and (c) contact current. A positive bias of 1.0 V was applied to the sample, and the external load during imaging was 100 nN. The average current in (c) was 0.94 A, with a modulation of about 17%. In all three images, the 0.246 nm graphite lattice periodicity is clearly observed. Using the DMT contact mechanics model (Müller et al., 1983; Derjaguin et al., 1975), with a measured pull-off force of 115 nN, we estimate that our contact radius is 4.1 nm, and therefore contains about 2000 atoms. The contact radius calculated here is only approximate, since the Tabor parameter (Greenwood & Johnson, 1998; Tabor, 1977) for this system is 0.67, which indicates that the DMT model is not entirely appropriate. Moreover, none of the analytical contact mechanics models are

directly applicable to a non-isotropic material such as graphite (Sridhar et al., 1997).

Similar lattice-resolved images were obtained on 1 ML of graphite on Pt(111) (Gr/Pt(111)). An example of this is shown in Fig. 18. As in the previous case, the 2.5 nm x 2.5 nm images correspond to: (a) normal lever deflection (under feedback-off conditions), (b) lateral force, and (c) contact current. The external load in this case was 300 nN, and the sample bias was 0.5 V. The average current was 52.7 A, with a current modulation of about 2%. In this case,

respect to the Pt lattice.

that the deposited graphite layer was 1 ML thick.

measurements spanning the range from pA to mA.

**4.3.1 Topography, friction, and current imaging** 

**4.3 Results and discussions** 

(c)

Fig. 17. Simultaneously acquired lattice resolution images of HOPG under UHV conditions: (a) normal lever deflection (with topographical and buckling effects), with a corrugation of 117 pN, corresponding to a height of 33.6 pm; (b) lateral friction image, average force of 0.5 nN and corrugation of 20 nN; (c) PCM image, average current of 945 nA and peak-to-peak corrugation of 160 nA. Image was taken with an applied load of 100 nN without feedback. Image size is 2.5 nm x 2.5 nm.

the diameter of the area of contact was similarly estimated to be 5.78 nm, which contains approximately 4000 atoms. Here the 0.246 nm graphite lattice periodicity is also clearly revealed. We were able to obtain lattice resolution at currents up to 0.9 mA and high load. The average pressure at high load was approximately 5 GPa, which is less than the theoretical yield stress of Pt (~17 GPa). At pressures higher than 5 GPa and/or currents higher than 0.9 mA, the images were unstable, although the graphite lattice was still visible.

As a side note, we found that we were able to obtain lattice resolution almost all the time and immediately in current mode, while lattice resolution was not as readily visible in the topography and/or friction channels. In many cases, the friction was so low that there was no stick-slip present, *i.e.,* the tip moved continuously over the graphite layer. Because of these reasons, we can rule out the atomic stick-slip mechanism as a reason for the lattice resolution observed in PCM mode.

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 127

Fig. 19. PCM image showing two moiré superstructure domain on graphite/Pt(111). The upper left has a periodicity of ~2.0 nm, and the superstructure lattice was determined to be ( 63 x 63 )R19 with respect to the graphite lattice. Image size is 60 nm x 60 nm. The other domain in this image is (5 x 5) with respect to the graphite lattice, although it is not resolved

Fig. 20. Close-up image of the ( 63 x 63 )R19 domain in Fig. 19, showing the graphite lattice, as well as the moiré superstructure. Image size is 10 nm x 10 nm. Average current is

structure in Fig. 20 corresponds to a superstructure with a ( 63 x 63 )R19 unit cell with respect to the graphite lattice. Contrary to standard usage, we report on the moiré structures with respect to the overlayer instead of the substrate, since we can directly count the number of graphite unit cells in the moiré superstructure. Using the known lattice constants of graphite and Pt and the measured angles, we can calculate that the moiré periodicity is almost exactly 7 Pt lattice spacings, and the moiré cell is rotated by 22° with respect to the Pt lattice. Indeed, one can create the 2.0 nm periodic superstructure by rotating the Pt [1-10] direction with the graphite [1010] direction by 2.68°, as shown in Fig. 21. There is a small lattice misfit of 0.60 % associated with the coincidence of the graphite

90 µA and corrugation is 5 µA at a bias of 0.8 V.

at this scale.

(c)

Fig. 18. Simultaneously acquired lattice resolution images of 1 ML of graphite deposited on Pt(111) in UHV: (a) normal lever, corrugation of 164 pN; (b) lateral friction image, average force of 0.4 nN and corrugation of 17 nN; (c) PCM image, average current of 53 µA and peak-to-peak corrugation of 1.1 µA at a bias of 0.53 V. Image was taken with an applied load of 300 nN without feedback. Image size is 2.5 nm x 2.5 nm.

#### **4.3.2 Moire´ structures**

It is known that, for similar preparation conditions (Land et al., 1992a; Hu et al., 1987), graphite forms several orientational domains on a Pt(111) sample. Depending on the preparation conditions and annealing temperature, different sizes and orientations of domains can be prepared. In Fig. 19, we show a 60 nm x 60 nm image of two graphite domains on Pt(111).

The hexagonal periodicity observed in the upper left domain in this image is about 2.0 nm. The large unit cell arises from the superposition of the incommensurate lattices of graphite and Pt(111) at a particular angle. In higher resolution images of this domain, such as the one shown in Fig. 20, the graphite lattice of 0.246 nm, together with the larger 2.0 nm cell, is revealed. Using the real space image and its 2-D Fourier transform, we find that the

(a) (b)

(c) Fig. 18. Simultaneously acquired lattice resolution images of 1 ML of graphite deposited on Pt(111) in UHV: (a) normal lever, corrugation of 164 pN; (b) lateral friction image, average force of 0.4 nN and corrugation of 17 nN; (c) PCM image, average current of 53 µA and peak-to-peak corrugation of 1.1 µA at a bias of 0.53 V. Image was taken with an applied load

It is known that, for similar preparation conditions (Land et al., 1992a; Hu et al., 1987), graphite forms several orientational domains on a Pt(111) sample. Depending on the preparation conditions and annealing temperature, different sizes and orientations of domains can be prepared. In Fig. 19, we show a 60 nm x 60 nm image of two graphite

The hexagonal periodicity observed in the upper left domain in this image is about 2.0 nm. The large unit cell arises from the superposition of the incommensurate lattices of graphite and Pt(111) at a particular angle. In higher resolution images of this domain, such as the one shown in Fig. 20, the graphite lattice of 0.246 nm, together with the larger 2.0 nm cell, is revealed. Using the real space image and its 2-D Fourier transform, we find that the

of 300 nN without feedback. Image size is 2.5 nm x 2.5 nm.

**4.3.2 Moire´ structures** 

domains on Pt(111).

Fig. 19. PCM image showing two moiré superstructure domain on graphite/Pt(111). The upper left has a periodicity of ~2.0 nm, and the superstructure lattice was determined to be ( 63 x 63 )R19 with respect to the graphite lattice. Image size is 60 nm x 60 nm. The other domain in this image is (5 x 5) with respect to the graphite lattice, although it is not resolved at this scale.

Fig. 20. Close-up image of the ( 63 x 63 )R19 domain in Fig. 19, showing the graphite lattice, as well as the moiré superstructure. Image size is 10 nm x 10 nm. Average current is 90 µA and corrugation is 5 µA at a bias of 0.8 V.

structure in Fig. 20 corresponds to a superstructure with a ( 63 x 63 )R19 unit cell with respect to the graphite lattice. Contrary to standard usage, we report on the moiré structures with respect to the overlayer instead of the substrate, since we can directly count the number of graphite unit cells in the moiré superstructure. Using the known lattice constants of graphite and Pt and the measured angles, we can calculate that the moiré periodicity is almost exactly 7 Pt lattice spacings, and the moiré cell is rotated by 22° with respect to the Pt lattice. Indeed, one can create the 2.0 nm periodic superstructure by rotating the Pt [1-10] direction with the graphite [1010] direction by 2.68°, as shown in Fig. 21. There is a small lattice misfit of 0.60 % associated with the coincidence of the graphite

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 129

At higher magnification, different periodic superstructures on each domain can be seen. The image in Fig. 23, which was obtained from the left domain of Fig. 22, shows a (3 x 3) modulation of the graphite lattice. Its 2-D Fourier transform is shown in Fig. 24. The larger hexagonal pattern, marked by six circles, corresponds to the 0.246 nm graphite lattice, while the smaller hexagon, marked by squares, represents the 0.738 nm superstructure lattice. The calculated angle of the graphite lattice with respect to Pt(111) lattice is 19.1°, which is in

Fig. 23. PCM image of a (3 x 3) moiré superstructure, showing the graphite lattice. Image size is 5 nm x 5 nm. Average current is 79 µA and modulation amplitude is 0.93 µA at a bias

Fig. 24. Fourier transform of image in Fig. 23, showing the graphite lattice periodicity of 0.246 nm marked by circles, and the moiré superstructure periodicity of 0.738 nm marked

Fig. 22 was identified as a ( 31 x 31 )R9 structure.

The domain in the middle of Fig. 22 has a (5 x 5) modulation of the graphite periodicity, as shown in the 5 nm x 5 nm image of Fig. 25. In this moiré structure, the angle between the graphite and the Pt(111) lattices is calculated to be 23.4°. The domain in the lower right of

agreement with the measured angle.

of 0.7 V.

by squares.

Fig. 21. Schematic of the ( 63 x 63 )R19 (with respect to graphite) moiré domain superstructure. With respect to the Pt(111) substrate, the moiré domain is (7 x 7)R22. The Pt atoms are shown as circles with a scaled diameter equal to the lattice constant of Pt (0.278 nm). The graphite lattice is shown as hexagons in which carbon atoms are located at the vertices with C-C distance of 0.142 nm and lattice constant of 0.246 nm. Vectors are drawn to indicate the orientation of the two lattices and the moiré domain. Image size is 5.5 nm x 5.5 nm.

lattice at 1.954 nm and the Pt lattice at 1.942 nm, which can be accounted for by a corresponding relaxation of the graphite layer or the platinum substrate.

Other graphite domains having different orientations and moiré superstructures have been observed, frequently adjacent to each other. The image in Fig. 22 shows three contiguous graphite domains, each having different orientations. It is interesting to note that the average current in these domains is different, even if all other conditions (bias, load, tip structure) are the same. The average current can also vary appreciably inside a single domain, such as at a platinum step, as we discuss below.

Fig. 22. PCM image showing different moiré domains. Image size is 100 nm x 100 nm. Note that the average current is different on each domain. Average currents are 86, 100 and 97 µA for the left, center and right domains, at a bias of 0.8 V.

Fig. 21. Schematic of the ( 63 x 63 )R19 (with respect to graphite) moiré domain

corresponding relaxation of the graphite layer or the platinum substrate.

domain, such as at a platinum step, as we discuss below.

for the left, center and right domains, at a bias of 0.8 V.

superstructure. With respect to the Pt(111) substrate, the moiré domain is (7 x 7)R22. The Pt atoms are shown as circles with a scaled diameter equal to the lattice constant of Pt (0.278 nm). The graphite lattice is shown as hexagons in which carbon atoms are located at the vertices with C-C distance of 0.142 nm and lattice constant of 0.246 nm. Vectors are drawn to indicate the orientation of the two lattices and the moiré domain. Image size is 5.5 nm x 5.5 nm.

lattice at 1.954 nm and the Pt lattice at 1.942 nm, which can be accounted for by a

Other graphite domains having different orientations and moiré superstructures have been observed, frequently adjacent to each other. The image in Fig. 22 shows three contiguous graphite domains, each having different orientations. It is interesting to note that the average current in these domains is different, even if all other conditions (bias, load, tip structure) are the same. The average current can also vary appreciably inside a single

Fig. 22. PCM image showing different moiré domains. Image size is 100 nm x 100 nm. Note that the average current is different on each domain. Average currents are 86, 100 and 97 µA At higher magnification, different periodic superstructures on each domain can be seen. The image in Fig. 23, which was obtained from the left domain of Fig. 22, shows a (3 x 3) modulation of the graphite lattice. Its 2-D Fourier transform is shown in Fig. 24. The larger hexagonal pattern, marked by six circles, corresponds to the 0.246 nm graphite lattice, while the smaller hexagon, marked by squares, represents the 0.738 nm superstructure lattice. The calculated angle of the graphite lattice with respect to Pt(111) lattice is 19.1°, which is in agreement with the measured angle.

Fig. 23. PCM image of a (3 x 3) moiré superstructure, showing the graphite lattice. Image size is 5 nm x 5 nm. Average current is 79 µA and modulation amplitude is 0.93 µA at a bias of 0.7 V.

Fig. 24. Fourier transform of image in Fig. 23, showing the graphite lattice periodicity of 0.246 nm marked by circles, and the moiré superstructure periodicity of 0.738 nm marked by squares.

The domain in the middle of Fig. 22 has a (5 x 5) modulation of the graphite periodicity, as shown in the 5 nm x 5 nm image of Fig. 25. In this moiré structure, the angle between the graphite and the Pt(111) lattices is calculated to be 23.4°. The domain in the lower right of Fig. 22 was identified as a ( 31 x 31 )R9 structure.

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 131

Fig. 26. PCM image of a Pt step covered by a continuous layer of graphite. The topography image (not shown) is completely flat, and does not reveal the presence of a step in the graphite layer at this scale. Image size is 10 nm x 10 nm. The average current is 39 µA on the

although tilted with respect to the Pt substrate. On larger scale images of regions containing wide Pt terraces, it is possible to measure a height difference between the terraces. The same (5 x 5) moiré superstructure was detected on both sides of the Pt step, which indicates that a

In these experiments, we noticed that the tip-sample contact is not always conductive, unlike in previous experiments with similar cantilevers (Enachescu et al., 1998), possibly because of contamination as a result of gases used during sample preparation. In particular, when such contamination is observed, current vs. load curves indicate that the current is often not proportional to contact area, with a weak dependence on load, much less than would be expected from contact area variations. The step observed by PCM in Fig. 26 is about 1.5 nm wide, denoting a lateral resolution in PCM mode of this value. Using the DMT contact mechanics model as we did earlier, we estimate that the diameter of the contact area is approximately 8 nm, which indicates a contact AFM lateral resolution of no less than 8 nm. We can use the Sharvin model for point-contact resistance (Sharvin, 1965; Wexler, 1966) and the measured point-contact resistivity to estimate the area through which current flows in our contact. The diameter of this area is estimated to be 0.9 nm, which is consistent with

One explanation for our observation of the different resolutions in AFM and PCM is that the tip is covered with a poorly conducting layer, which is partially broken when the tip is cleaned by applying voltage pulses. This phenomenon may be limited to the tungstencarbide coating of the tip used in this experiment. Another explanation may be that only the highest-pressure region of the contact area contributes to the point-contact current. However, a graph of current vs. load strongly favors the former explanation. We note that the weak dependence of current on load indicated by these graphs resembles the similarly weak I-Z dependence observed in the past in STM experiments on graphite (Salmeron et al., 1991). This supports the idea that, in most cases, STM on graphite is actually point-contact imaging. The change in local conductivity over the Pt step is likely due to the increased

upper terrace and 28 µA on the lower terrace at a bias of 1.0 V.

continuous sheet of graphite is covering the step.

the observed lateral resolution in the PCM image.

Fig. 25. Close-up image of a (5 x 5) moiré superstructure, showing the graphite lattice. The arrows indicate the moiré lattice. Image size is 5 nm x 5 nm.

Table 1 is a list of the experimentally observed moiré structures. Using the ratio of the lattice constants of graphite and Pt, we were able to calculate near-coincidences of the graphite and Pt lattices at different angles, and thus predict the existence of all of the structures.


Table 1. Moiré superstructures experimentally observed in PCM mode. Structures with respect to the Pt lattice are deduced from the measured structures on graphite. In certain cases, the angle of the Pt lattice was known, which aided in the extrapolation.

#### **4.3.3 Measuring local conductivity using PCM**

To determine the lateral resolution of PCM, we acquired images of regions containing platinum steps. We observed that the graphite layer covers the Pt steps continuously from the upper terrace to the lower adjacent one, as shown in Fig. 26. It is important to mention that the image in Fig. 26 is a contact current image. At distances far from the step in this image, the average current is the same on both terraces. However, close to the step, on what we have identified as the lower terrace, the contact current decreases by approximately 30%. At this scale, the topography image shows no contrast, since the graphite layer is almost flat,

Fig. 25. Close-up image of a (5 x 5) moiré superstructure, showing the graphite lattice. The

Table 1 is a list of the experimentally observed moiré structures. Using the ratio of the lattice constants of graphite and Pt, we were able to calculate near-coincidences of the graphite and

> Angle between graphite and Pt lattices [0]

Moiré periodicity [nm]

Coincidence misfit [%]

Pt lattices at different angles, and thus predict the existence of all of the structures.

(3x3) (7x7) *R*19 19.1 0.738 0.60 (19x19) *R*23 (4x4) 23.4 1.07 3.4 (5x5) (19x19) *R*23 23.4 1.23 1.7 (31x31) *R*9 (5x5) 8.9 1.37 1.2 (52x) *R*14 (43x43) *R*8 21.5, 6.3 1.77 2.5 (1x1) *R*26 Unknown 1.92 < 4.0 (x) *R*19 (7x7) *R*22 2.7 1.95 0.60 (x) *R*6 Unknown 2.10 < 4.9

Table 1. Moiré superstructures experimentally observed in PCM mode. Structures with respect to the Pt lattice are deduced from the measured structures on graphite. In certain

To determine the lateral resolution of PCM, we acquired images of regions containing platinum steps. We observed that the graphite layer covers the Pt steps continuously from the upper terrace to the lower adjacent one, as shown in Fig. 26. It is important to mention that the image in Fig. 26 is a contact current image. At distances far from the step in this image, the average current is the same on both terraces. However, close to the step, on what we have identified as the lower terrace, the contact current decreases by approximately 30%. At this scale, the topography image shows no contrast, since the graphite layer is almost flat,

cases, the angle of the Pt lattice was known, which aided in the extrapolation.

**4.3.3 Measuring local conductivity using PCM** 

arrows indicate the moiré lattice. Image size is 5 nm x 5 nm.

Moiré periodicity relative to Pt(111)

Moiré periodicity relative to graphite

Fig. 26. PCM image of a Pt step covered by a continuous layer of graphite. The topography image (not shown) is completely flat, and does not reveal the presence of a step in the graphite layer at this scale. Image size is 10 nm x 10 nm. The average current is 39 µA on the upper terrace and 28 µA on the lower terrace at a bias of 1.0 V.

although tilted with respect to the Pt substrate. On larger scale images of regions containing wide Pt terraces, it is possible to measure a height difference between the terraces. The same (5 x 5) moiré superstructure was detected on both sides of the Pt step, which indicates that a continuous sheet of graphite is covering the step.

In these experiments, we noticed that the tip-sample contact is not always conductive, unlike in previous experiments with similar cantilevers (Enachescu et al., 1998), possibly because of contamination as a result of gases used during sample preparation. In particular, when such contamination is observed, current vs. load curves indicate that the current is often not proportional to contact area, with a weak dependence on load, much less than would be expected from contact area variations. The step observed by PCM in Fig. 26 is about 1.5 nm wide, denoting a lateral resolution in PCM mode of this value. Using the DMT contact mechanics model as we did earlier, we estimate that the diameter of the contact area is approximately 8 nm, which indicates a contact AFM lateral resolution of no less than 8 nm. We can use the Sharvin model for point-contact resistance (Sharvin, 1965; Wexler, 1966) and the measured point-contact resistivity to estimate the area through which current flows in our contact. The diameter of this area is estimated to be 0.9 nm, which is consistent with the observed lateral resolution in the PCM image.

One explanation for our observation of the different resolutions in AFM and PCM is that the tip is covered with a poorly conducting layer, which is partially broken when the tip is cleaned by applying voltage pulses. This phenomenon may be limited to the tungstencarbide coating of the tip used in this experiment. Another explanation may be that only the highest-pressure region of the contact area contributes to the point-contact current. However, a graph of current vs. load strongly favors the former explanation. We note that the weak dependence of current on load indicated by these graphs resembles the similarly weak I-Z dependence observed in the past in STM experiments on graphite (Salmeron et al., 1991). This supports the idea that, in most cases, STM on graphite is actually point-contact imaging. The change in local conductivity over the Pt step is likely due to the increased

Nanoscale Effects of Friction, Adhesion and Electrical Conduction in AFM Experiments 133

effect alone is insufficient to fully describe the electronic structure of steps. It is therefore important that the presence and the magnitude of local dipole moments at steps be

Scanning probe microscopy can be used to investigate the electronic structure of steps. Marchon *et al*. observed a reduction in the tunneling barrier at surface steps on sulfur-covered Re(0001) (Marchon et al., 1988) using scanning tunneling microscopy (STM). Later Jia *et al.* used this effect to calculate the step dipole for Au(111) and Cu(111) (Jia et al., 1998a; 1998b). Arai and Tomitori investigated step contrast as a function of tip bias on Si(111) (7x7) using dynamic atomic force microscopy (D-AFM) (Arai & Tomitori, 2000) and suggested that step dipoles could explain their observations. In contrast Guggisberg *et al.* investigated the same system using STM feedback combined with D-AFM force detection and concluded that the step dipole moments in Si(111)-(7x7) were negligible (Guggisberg et al., 2000). They attributed the D-AFM contrast effects to changes in the van der Waals and electrostatic polarization forces, which are reduced above and increased below the step edges relative to the flat terrace. In this work we report measurements of the strength of the fields produced by the step dipoles through direct measurement of the electrostatic force they produce on biased tips. We use a combined STM-AFM system (Enachescu et al., 1998; Park et al., 2005b) with cantilevers that are made conductive by a ~30 nm coating of W2C. Relatively stiff cantilevers of 48 or 88 N/m were used to avoid jump-to-contact instabilities close to the surface. Attractive forces cause the cantilevers to bend toward the surface during imaging, as illustrated in Fig 27(a). Scanning is done at constant current as in standard STM mode, while forces are measured simultaneously from the cantilever deflection (Park et al., 2005c).

The force acting on the tip is the sum of van der Waals and electrostatic contributions. The former is independent of the applied bias. The electrostatic contributions are additive and

where D is the tip-surface distance, R the tip radius and *f*, *g* and *h* are functions of the tip and sample geometry. P is the dipole moment, and *V* is the electrostatic potential difference between tip and sample. The first term in (1) represents the attractive force from polarization (i.e. image charges) induced by the applied voltage. The second term is due to surface dipoles *P* interacting with the biased tip, and is proportional to the bias. The last term is the force between the dipole *P* and its image on the tip. Of these contributions only the second term is linear with applied voltage, and provides an easy way to determine the

The measurements were carried out in ultra high vacuum with an optical deflection AFM. Several samples were used, including Pt(111), Au(111) and the two-fold surface of a Al74Ni10Co16 decagonal quasicrystal prepared by cutting the crystal parallel to the ten-fold axis. The growth and characterization of the Al-Ni-Co quasicrystal are outlined in detail elsewhere (Fisher et al., 1999). Due to the aperiodic nature of the atomic layering in the latter sample, steps of various heights were readily obtained on a single surface. The Pt single crystal and the quasicrystal (Park et al., 2004) samples were sputtered and annealed in UHV.

2 2 *F f* (/) (/) (/) *D RV g D R PV h D R P* (1)

measured experimentally.

can be written as (Jackson, 1975):

net effect of the dipole field.

**5.2 Experimental** 

distance between the graphite layer and the underlying Pt substrate. The increased distance acts much like a tunneling barrier. In our measurements, we are able to measure current independently of topography, since the tip-sample contact is affected only by the mechanics of the system. The STM technique uses feedback on current to measure topography, so, for example, in the case of the blanketed Pt step, the STM tip would see the decrease in current and move closer to the sample to compensate. Thus, an STM image of a blanketed step would show a topographic step in the graphite layer with a width of 0.2 nm (i.e., typical STM resolution), while contact AFM indicates that the step width is many tens of nanometers. This width is the distance from the platinum step where the graphite layer begins to separate from the platinum substrate. Since the PCM technique is capable of separating mechanical and electrical measurements, it can offer additional insight into the electronic and tribological properties of surfaces.

The STM images of Land et al. (Land et al., 1992a; 1992b) indicate that there is local conductivity modulation at both the lattice and the moiré periodicities. If we imagine the atoms in our AFM contact contributing to the contact current as a collection of STM tips, one for each atom, the total contact current would be the sum of the contribution of these tips. We would still expect to see both the lattice and the moiré periodicities in the resulting PCM image, although the magnitude of modulation relative to the average current would decrease. The modulation would sum to zero only in special, destructively interfering cases. This will be discussed in more detail in a future work.
