Atomic Force Microscope Images of Nanobubbles on a Hydrophobic Surface and Corresponding Force-Separation Data

Transcription

1 160 Langmuir 2002, 18, Atomic Force Microscope Images of Nanobubbles on a Hydrophobic Surface and Corresponding Force-Separation Data J. W. G. Tyrrell and P. Attard* Ian Wark Research Institute, University of South Australia, Mawson Lakes, SA 5095, Australia Received July 30, In Final Form: October 17, 2001 Domains, apparently nanobubbles, have been observed on a hydrophobic glass surface in water using an atomic force microscope operated in tapping mode. Phase images show the domains to be softer than the underlying substrate. Complementary force curves between a silica colloid probe and the glass surface display features characteristic of the hydrophobic interaction, including a jump-in distance that is comparable to the height of the imaged domains. Images and force curves have been acquired over a range of ph conditions to probe the nature of the overall interaction and gain some insight into the conformation of nanobubbles on the sample surface. The bubbles appear to regrow following their removal by the application of high loads through either contact mode imaging or tapping mode imaging at high drive amplitudes. They are not present in a solvaphilic fluid (ethanol) but regrow following the subsequent reintroduction of H 2O. The correlation between image and force data, supported by existing results in the literature, provides strong evidence to favor nanobubbles as the origin of the hydrophobic force. Introduction The long-range attractive force exhibited by hydrophobic surfaces in aqueous solutions is termed the hydrophobic interaction. It has been observed experimentally by many workers, with references 1 identifying the long-range attraction dating back to the early 1970s (a recent review 2 is given by Christenson and Claesson). The origin and mechanism of this interaction have been a source of debate. Initially, most thoughts focused on a long-range electrostatic coupling between surfaces, with a number of theoretical mechanisms and subsequent modifications being proposed and documented in the literature. 3-8 However, a number of experimental measurements 9-13 showed the long-range attractive force to persist even at molar electrolyte concentrations. At such concentrations, any electrostatic interaction between surfaces would be heavily screened and hence very short range. This indicates the hydrophobic interaction to be nonelectrostatic in origin. There have been a number of nonelectrostatic mechanisms put forward, but none of these have been widely accepted. (1) Blake, T. D.; Kitchener, J. A. J. Chem. Soc., Faraday Trans , 68, (2) Christenson, H. K.; Claesson, P. M. Adv. Colloid Interface Sci. 2001, 91, 391. (3) Attard, P. J. Phys. Chem. 1989, 93, (4) Podgornik, R. J. Chem. Phys. 1989, 91, (5) Tsao, Y. H.; Evans, D. F.; Wennerström, H. Langmuir 1993, 9, 779. (6) Miklavic, S. J.; Chan, D. Y. C.; White, L. R.; Healy, T. W. J. Phys. Chem. 1994, 98, (7) Spalla, O.; Belloni, L. Phys. Rev. Lett. 1995, 74, (8) Miklavic, S. J. J. Chem Phys. 1995, 103, (9) Christenson, H. K.; Fang, J.; Ninham, B. W.; Parker, J. L. J. Phys. Chem. 1990, 94, (10) Christenson, H. K.; Claesson, P. M.; Parker, J. L. J. Phys. Chem. 1992, 96, (11) Parker, J. L.; Claesson, P. M.; Attard, P. J. Phys. Chem. 1994, 98, (12) Craig, V. S. J.; Ninham, B. W.; Pashley, R. M. Langmuir 1998, 14, (13) Carambassis, A.; Jonker, L. C.; Attard, P.; Rutland, M. Phys. Rev. Lett. 1998, 80, (14) Eriksson, J. C.; Ljunggren, S.; Claesson, P. M. J. Chem. Soc., Faraday Trans , 85, 163. The impetus for drawing the aforementioned debate to a close is twofold. First, hydrophobic interactions underpin many industrial processes, for example, froth flotation, colloid destabilization, emulsion flocculation, and macromolecular assembly, where the ability to characterize and understand the fundamental mechanisms impacts directly on the quality of products ranging from cosmetics to explosives. And second, measurements of a long-range attractive force superficially appear to contradict the longstanding Derjaguin-Landau-Verwey-Overbeek (DLVO) theory for electric double layer and van der Waals interactions. It is therefore important to understand the physical basis of the hydrophobic interaction and to place it in the context of these classic theories forming the foundation of colloid and surface science. It has been proposed that the attraction is due to the bridging of nanobubbles that pre-exist on hydrophobic surfaces. A body of direct force measurements 11,13,18-21 has been assembled in support of the idea, including a number of experiments 18,19,22-25 that show the hydrophobic attraction is diminished in deaerated water. The nanobubble theory is arguably the most attractive, the main objection being the theoretically short lifespan of bubbles tens of nanometers in diameter due to their high internal pressure. 26,27 Until now, no convincing visual evidence of (15) Ruckenstein, E.; Churaev, N. J. Colloid Interface Sci. 1991, 147, 535. (16) Bérard, D. R.; Attard, P.; Patey, G. N. J. Chem. Phys. 1993, 98, (17) Lum, K.; Chandler, D.; Weeks, J. D. J. Phys. Chem. B 1999, 103, (18) Considine, R. F.; Hayes, R. A.; Horn, R. G. Langmuir 1999, 15, (19) Considine, R. F.; Drummond, C. J. Langmuir 2000, 16, 631. (20) Yakubov, G. E.; Butt, H.-J.; Vinogradova, O. I. J. Phys. Chem. B 2000, 104, (21) Vinogradova, O. I.; Yakubov, G. E.; Butt, H.-J. J. Chem. Phys. 2001, 114, (22) Wood, J.; Sharma, R. Langmuir 1995, 11 (1), (23) Meagher, L.; Craig, V. S. J. Langmuir 1994, 10, (24) Ishida, N.; Sakamoto, M.; Miyahara, M.; Higashitani, K. Langmuir 2000, 16, (25) Mahnke, J.; Stearnes, J.; Hayes, R. A.; Fornasiero, D.; Ralston, J. Phys. Chem. Chem. Phys. 1999, 1, /la CCC: $ American Chemical Society Published on Web 01/02/2002

2 Nanobubbles on a Hydrophobic Surface Langmuir, Vol. 18, No. 1, nanobubbles has been provided (see the discussion of the work of Ishida et al. 28 below), with their presence being inferred solely from measurements of the interaction force as a function of separation. In this paper, we present striking tapping mode atomic force microscope (AFM) images, obtained for a range of ph conditions, showing the presence of domains on a hydrophobic glass surface immersed in solution. Phase images show the domains to be softer than the underlying substrate. The image data are complemented with force curves showing the interaction between a colloid probe and the glass surface. These force curves display features characteristic of the hydrophobic interaction and, when coupled with the images, provide strong evidence to favor nanobubbles as the origin of the hydrophobic force. A surprising finding is that the coverage of the surface by nanobubbles is near 100%. In addition, supplementary experiments involving the introduction of a solvaphilic fluid (ethanol) into the tapping mode AFM fluid cell resulted in the removal of the domains in both the height and phase images and the loss of the long-range attraction. Given this evidence, it is likely that the introduction of a wetting fluid initiates the removal of bubbles from the sample surface. Images display phase and height contrast when the cell is refilled with H 2 O. A preliminary account 29 of this work has been published recently. Experimental Section Preparation of Surfaces. Glass surfaces were rinsed thoroughly with deionized water (Elga UHQ purification system), bathed in a warm concentrated KOH solution for >2 min, rinsed again with deionized water, and then left to dry in a clean room environment. Surfaces were rendered hydrophobic through exposure to dichlorodimethylsilane vapor 25,30 for 3 min and then rinsed with deionized water. Following silanation, water was observed to bead readily on the glass surface. An advancing water contact angle of 101 (receding angle ) 80 ) was measured on the glass surface with the sessile-drop method. Surfaces were found to have a root mean square (rms) roughness of <0.5 nm, when imaged in air using an AFM. For the force measurements, a silica sphere (Geltech Inc., Alachua, FL, radius 7.5 µm) was micromanipulated onto the underside of a Si 3N 4 cantilever and fixed in position with a high melting point wax (Shell Epikote 1004). The sphere and cantilever were exposed to silane vapor 30 for a period of 30 s. A longer exposure was not deemed prudent because of the delicate nature of the sphere-cantilever assembly. Preparation of Solutions. Water was provided by an Elga UHQ purification system. Water exposed to the atmosphere at room temperature gave a ph reading of 5.6, due to the presence of dissolved atmospheric CO 2. Drops of either KOH or HNO 3 were added to achieve the desired ph. Measurements. Experimental data were acquired using a Nanoscope IIIa AFM (Digital Instruments Inc.) with an extender module to provide tapping mode capability. The use of a fluid cell enabled measurements to be performed in liquid (force measurements were performed in the same solution as was used for imaging). A syringe was placed on the outlet line from the fluid cell, and ph solutions were drawn through the cell. This technique was found to be beneficial in eliminating entrained macroscopic bubbles and preserving the integrity of the O-ring seal against the sample surface. The resonant frequency of the cantilever was recorded before and after mounting of the particle to enable calibration of the (26) Attard, P. Langmuir 1996, 12, (27) Ljunggren, S.; Eriksson, J. C. Colloids Surf., A 1997, , 151. (28) Ishida, N.; Inoue, T.; Miyahara, M.; Higashitani, K. Langmuir 2000, 16, (29) Tyrrell, J. W. G.; Attard, P. Phys. Rev. Lett. 2001, 87, (30) Butt, H.-J. J. Colloid Interface Sci. 1994, 166, 109. spring constant as prescribed by the Cleveland method. 31 This confirmed the manufacturer s value of 0.58 N m -1 to within confidence limits based upon the calculated mass of the colloid probe ((20%). Tapping mode images were obtained using a Si 3N 4 cantilever (k 0.38 N m -1 ), with no probe attached, in accordance with technical support notes provided by the manufacturer. 32 The cantilever was modulated at frequencies in the range khz and driven with an amplitude of mv to give an rms reading of 0.5 V prior to contact. The free amplitude (peakto-peak) of the cantilever oscillation was determined from amplitude versus distance plots to be approximately 2.5 µm. Image quality was severely degraded when changes of (10 mv were made to the drive amplitude. At higher drive amplitudes, the cantilever was thought to possess energy sufficient to penetrate the bubble layer, whereas at lower drive amplitudes the cantilever was thought to be unable to break free from the bubble layer. Proportional and integral gains were set to a value of 1.0. A 1 µm 1 µm area was imaged at a scan rate of 2 Hz. This corresponded to a scan speed of 4 µm/s along the fast axis. Image analysis (SPM off-line analysis 4.43r5, Digital Instruments Inc.) was conducted after first applying a low pass filter to reduce any high frequency noise associated with the data. A threshold height was set to enable the software to automatically identify features present within the image for analysis (e.g., domain size). Force-separation curves have been extracted from original force-displacement data in accordance with guidelines 33 outlined by Ducker et al.. The linear slope of the constant compliance region and its consistency in value across all force-displacement curves of common beam geometry suggest that hard contact was achieved between surfaces. Zero force was taken as an average of data points acquired at maximum separation, typically 500 nm, to minimize the contribution of the surface-probe interaction. Zero separation was taken as an average of data points in the region of closest approach. Noise around the point of zero separation is thought to be due to either asperities or stick slip friction during the small, but inevitable, lateral motion of the cantilever across the sample surface. This occurs in the direction of the cantilever s long axis as the sample advances vertically. Forces were measured with and without the attached colloid probe. It was necessary to correct for the presence of interference fringes when processing the data in order to extract the decay length. This was achieved by subtracting a fitted sine wave from the original data. In all cases, a good correlation was seen between the fitted wavelength and that of the laser source, confirming the piezo calibration. Failure to correct for interference effects was found to distort decay lengths. Results Given the strong body of evidence presented here and elsewhere, 11,13,18-25 domains shown in the tapping mode AFM images and inferred from the nature of forceseparation data will be called nanobubbles throughout. Although this gives the most coherent interpretation of the data, it is recognized that other identifications of the domains and force features cannot be ruled out. Figure 1 shows force-separation data corresponding to the approach and subsequent jump into contact of a hydrophobic glass surface and a silica sphere (radius 7.5 µm) mounted on a Si 3 N 4 microfabricated cantilever. Particle and surface exhibit a repulsion at long range. The decay length of the interaction appears to be independent of approach velocity, given the common curve traced by all data sets. The decay length of 25 nm is less than the Debye length of 192 nm, which one would expect (31) Cleveland, J. P.; Manne, S.; Bocek, D.; Hansma, P. K. Rev. Sci. Instrum. 1993, 64, 403. (32) (33) Ducker, W. A.; Sendon, T. J.; Pashley, R. M. Langmuir 1992, 8, 1831.

3 162 Langmuir, Vol. 18, No. 1, 2002 Tyrrell and Attard Figure 1. Force-separation curves for a hydrophobic glass surface approaching a colloid probe (silica sphere, radius 7.5 µm) at ph 5.6. The experiment was performed at three approach velocities: 0.2, 1.0, and 6.9 µms -1 (from left to right during the jump-in, the velocities are 6.9, 6.9, 6.9, 1.0, 0.2, 0.2, and 0.2 µm s -1, respectively). The inset shows the region of the jump-in and the soft compliance in detail. for an electric double layer repulsion as predicted by DLVO theory for ph 5.6 conditions with no added electrolyte. When the sample approach velocity is increased from 0.2 to 6.9 µm/s, the separation at which particle and surface jump into contact (jump-in distance) decreases from 21 to 15 nm. This trend, namely, a reduction in the range of the hydrophobic interaction with increasing velocity of surface approach, has been observed elsewhere. 13,37,38 Given the repulsive interaction upon approach, the decrease in jumpin distance with increasing velocity translates as an increase in the magnitude of the maximum repulsion between particle and surface. In each case, irrespective of jump-in distance, the particle jumps to a separation of around 5 nm. As the particle and surface jump into contact, they pass through a soft compliance region until hard-wall contact is made. Hard-wall contact occurs at large applied loads and, in the absence of any deformation between probe and sample, is signified by the displacement of the sample being the same as the displacement of the cantilever. The soft compliance region can be seen as a hook in the force curve and appears to be more pronounced when the sample is advanced at higher velocities. Physically, the soft compliance region is thought to reflect the lateral spreading of bubbles within the contact zone and compression of entrapped gas. 13,39 The aforementioned features, a repulsion at large separations prior to the onset of a long-range attraction and jump toward contact as well as a soft compliance regime prior to hard-wall contact, have been observed in previous AFM measurements 13,18 of the interaction between hydrophobic surfaces. Figure 2 introduces three approach curves from the same series as Figure 1. These curves are anomalous with regard to those shown in Figure 1 but noteworthy because (34) Israelachvili, J. N. Intermolecular and Surface Forces, 2nd ed.; Academic Press: London, (35) Derjaguin, B. V.; Landau, L. Acta Physicochim. 1941, 14, 633. (36) Verwey, E. J.; Overbeek, J. Th. G. Theory of Stability of lyophobic colloids; Elsevier: Amsterdam, (37) Craig, V. S. J.; Ninham, B. W.; Pashley, R. M. Langmuir 1999, 15, (38) Zhang, X.; Zhu, Y.; Granick, S. J. Am. Chem. Soc. 2001, 123, (39) Attard, P. Langmuir 2000, 16, Figure 2. Anomalous force-separation curves (from the same series as in Figure 1). Velocities are 0.2, 1.0, and 0.2 µm s -1 from left to right during the jump. The data are magnified in the inset. Figure 3. Force-separation curves showing a hydrophobic glass surface withdrawing from the colloid probe at ph 5.6. The experiment was performed at three retract velocities: 0.2, 1.0, and 6.9 µms -1 (from left to right during the jump-out, the velocities are 0.2, 0.2, 1.0, 0.2, 0.2, 6.9, 6.9, and 6.9 µm s -1, respectively) and under ph 5.6 conditions. The horizontal portions indicate photodiode saturation due to strong adhesion between particle and surface. their unusual features were observed on a number of occasions. It is interesting to note the presence of an initial attractive feature followed by a repulsive barrier at a separation of 25 nm, a value within the height range of the domains shown later. The origin of the repulsive barrier is possibly an electrostatic repulsive interaction between adjacent bubbles on the sample surface that temporarily inhibits the lateral spread of the bridging bubble. Two of the curves shown in Figure 2 display a shallower minimum and much more extended soft compliance than data presented in Figure 1. Withdrawal curves, also known as retract or pull-off curves, corresponding to the approach curves shown in Figures 1 and 2, are given in Figure 3. The experimental setup has been optimized to capture the long-range repulsion seen upon approach in Figures 1 and 2. As a direct consequence, any strong adhesion will saturate the photodetector and result in some loss of information. However, it is possible to interpret the jump-off distance as a measure of adhesion, that is, extrapolating the jump out of contact to the point of intersection with the y-axis. Accepting this protocol, increasing the separation velocity from 0.2 to 6.9 µm/s gives an increase in adhesion from 64 to 102 nn.

4 Nanobubbles on a Hydrophobic Surface Langmuir, Vol. 18, No. 1, It is interesting to compare this adhesion force to the value of 1300 nn as predicted by the Laplace equation, F ) 4πRγ cos θ (1) which describes the force due to capillary condensation between a sphere and an identical flat surface (in this case the condensate is a vapor bridge). Here, F is the capillary force, R is the radius of the particle (7.5 µm), γ is the interfacial surface tension ( Nm -1 ), and θ is the solid-liquid contact angle (101 ). The discrepancy between measured and predicted adhesion values is too large to be attributed to any uncertainty in the cantilever spring constant ( 20%). However, eq 1 assumes that the contact angle made by the liquid against the glass surface is identical to that made by the liquid against the particle surface. A decrease in the contact angle made by the liquid against the particle surface would reduce the curvature of the capillary bridge and consequently decrease the predicted adhesion. An alternative expression for capillary adhesion has been derived by minimization of the constrained thermodynamic potential. 39 The so-called bridging cylinder approximation (BCA) has been shown to be more accurate than eq 1 for colloidal sized particles. 39 The BCA is readily generalized to the present case of asymmetric surfaces. The pull-off force is F )-πp 0 rj 2-2πγrj 2 (2) where the equilibrium radius of the capillary vapor bridge is rj ) -3γ/R + 9γ 2 /R 2-8( γ 1 + γ 2 )(P 0 /R + γ 1 /R γ 2 /R 2 2 ) 2(P 0 /R + γ 1 /R γ 2 /R 2 2 ) (40) Fundamentals of Adhesion and Interfaces; Rimai, D. S., DeMejo, L. P., Mittal, L. P., Eds.; VSP: Utrecht, (3) Figure 4. Force-separation curves showing a hydrophobic glass surface approaching the colloid probe and subsequent jump-in characteristics at ph 9.4. The experiment was performed at three approach velocities: 0.4, 2.0, and 13.8 µm s -1 (from left to right during the jump-in, the velocities are 2.0, 2.0, 13.8, 13.8, 13.8, 0.4, 0.4, and 0.4 µms -1, respectively). The data are magnified in the inset. Here, P 0 ) 10 5 Nm -2 is atmospheric pressure, R i is the radius of curvature of each body (R ) R 1 + R 2 ), γ i ) γ cos θ i is the difference in surface energy, and θ i is the contact angle measured through the liquid phase. In the present case, θ 1 ) 101, R 1 ), and R 2 ) 7.5 µm. Taking θ 1 ) gives a capillary adhesion of nn, comparable to the range of measured values. The sphere is expected to be less hydrophobic than the substrate because it was exposed to the silane vapor for a shorter period of time (30 s as opposed to 3 min for the substrate). We have confirmed that the contact angle for glass substrates is dependent on silanation time for exposure periods less than 3 min. Alternatively, the general observation 34,40 that AFM measurements tend to underestimate quantities derived from surface thermodynamics is worth noting, and this may also contribute to the discrepancy seen between measured and predicted values of adhesion. Finally, steps are observed in the withdrawal curve at large separations. It is proposed that these steps manifest as the snapping of multiple bridging strands, formed as the bridging bubbles collapse to submicroscopic radii at large separations. Theoretically, it is known that at large separations submicroscopic bridging bubbles are stable with respect to microscopic bridging bubbles. 39 It is plausible that the collapse of the microscopic bridging bubble is quite fast and relatively violent and that it should result in several submicroscopic bridging strands. Christenson and Claesson 41 visually observe multiple bridging bubbles following the collapse of a macroscopic bridging bubble. The snapping of each strand results in a step in the measured force. An analogy would be the stepwise characteristics of withdrawal curves seen in polymeric systems, 34,42 where each step corresponds to the rupture of a polymer strand bridging two surfaces. In the present case, of course, no polymers are present (see discussion below). The effect on particle-surface interactions of increasing ph conditions from ph 5.6 to ph 9.4 can be seen by comparing approach curves shown in Figure 1 and Figure 4. The strength of the repulsion, as indicated by the maximum in the approach curve prior to jump-in, increases as the ph is increased from ph 5.6 to ph 9.4. The surface charge is expected to become more negative on both the bubble and the probe as the ph is increased. In both cases, ph 9.4 and ph 5.6, particle and surface jump into contact at separation values around 15 to 20 nm. Again, the exact value exhibits a dependence on approach velocity; namely, as the sample approach velocity is increased, the separation at which particle and surface jump into contact (jumpin distance) decreases. The data in Figure 5 are almost linear on the semilog plot, which is indicative of exponential decay. When artifacts due to laser interference are removed, the repulsion is found to be almost independent of ph and driving velocity and to have a decay length of 25 nm. This value is much shorter than the expected DLVO decay lengths of κ -1 ) 192 nm (ph 5.6) and 61 nm (ph 9.4). The origin of the faster than expected decay at short separations (less than a Debye length from bubble-surface contact) is unclear. We have ruled out hydrodynamic drainage effects because the curves are identical for different approach speeds. We can also rule out bubble deformation effects, since in general these soften the force, that is, increase the decay length. The magnitude of the repulsion shows a slight increase with ph, which does suggest some electrostatic contribution. However, the computed electric double layer force between the probe and bubble of radius 100 nm is much less than the resolution of the apparatus and could not be measured (41) Christenson, H. K.; Claesson, P. M. Science 1988, 239, 390. (42) Senden, T. J.; di Meglio, J.-M.; Auroy, P. Eur. Phys. J. B 1998, 3, 211.

5 164 Langmuir, Vol. 18, No. 1, 2002 Tyrrell and Attard Figure 5. Force on a logarithmic scale showing repulsion prior to the jump into contact. The characters are ph 9.4 and the open symbols are ph 5.6, for three different driving velocities in each case ( µms -1 ). The fitted line is F ) A exp -h/λ, with A ) 2.5 nn and λ ) 25 nm. The inset shows jump-in distance plotted as a function of approach velocity for ph 5.6 (open circles) and ph 9.4 (closed triangles) conditions. Figure 6. Force-separation curves showing a hydrophobic glass surface approaching the colloid probe at ph 3.0. The experiment was performed at three approach velocities: 0.2, 1.0, and 6.9 µm s -1. here. The calculated hydrodynamic drainage repulsion between a 100 nm bubble and the probe is also much less than the present resolution. This discrepancy between predicted and observed decay lengths, seen also by other workers, 13 indicates the presence of additional contributions. These contributions, increasing the rate of decay of the force at small surface separations, may arise from image charge effects, which decay as κ -1 /2, due to the dielectric discontinuity at the air-water interface. Alternatively, the roughness and heterogeneity arising from the presence of multiple bubbles on the substrate (see images below) may contribute the observed deviation from DLVO behavior. Approach curves measured under ph 3.0 conditions are given in Figure 6. No repulsion is evident between probe and surface at ph 3.0. This contrasts with the repulsive behavior observed at both ph 5.6 and ph 9.4 but is nevertheless consistent with an electric double layer contribution to the prejump repulsion. One expects both the bubble surface and the silica probe to have a less negative surface charge at lower ph. Indeed, the bubble surface may be positively charged. Jump-in distances ranging from 22 to 38 nm are greater than those of between 13 and 21 nm, observed at higher ph. Such an increase in jump distance would be expected for an attractive double layer interaction between the probe and the bubble. The low interfacial spring constant of a bubble (0.072 N/m) 43 implies that attractive interactions are manifest by a jump of the interface toward the probe rather than by a jump of the cantilever. Interestingly, there appears to be no correlation between jump-in distance and approach velocity for the data obtained at ph 3.0. The decrease in jump distance with increased velocity is observed only when a repulsive component exists between particle and surface (see Figures 1 and 4). Figure 7 shows tapping mode AFM images of a hydrophobic glass surface immersed in water at ph 5.6 (phase and height), 4.4, 3.0, and 9.4. The following observations reflect the order of the experiments, conceived to examine the effect of a decrease followed by an increase in ph upon the system. As the ph is reduced from 5.6 to 3.0, the irregular bubble network evident at ph 5.6 and ph 4.4 coalesces to form larger, discrete bubbles of regular shape. Image analysis data, as provided in Table 1, corroborate such visual observations. The irregular bubble network is supported by high coefficient of variation (cv) values, which decrease from 70% to 34% as the bubbles adopt regular, discrete conformations at low ph. Bubbles can be seen to increase in size as the ph is decreased from 5.6 to 3.0, confirmed by a corresponding increase in mean area from 4108 to 6366 nm 2. As the ph is increased from 3.0 to 9.4, mean area decreases from 6366 to 3910 nm 2. The discrete nature of the domains is enhanced, arguably as a result of the increase in surface charge and corresponding repulsion between neighboring sites at high ph. Obviously, these quantitative values depend on the threshold, but for any threshold the trends with ph are qualitatively the same. It is possible that the poor image quality seen in Figure 7d can be attributed to the absence of a repulsive component in the tip-surface interaction. In general, the morphology of the nanobubbles appeared to be independent of scan rate over the practical range of values adopted for imaging (0.5-2 Hz). Figure 7a shows the phase signal corresponding to the height image given in Figure 7b. The strong correlation between height and phase images is striking. Domains present in the height image are clearly associated with a shift in phase, indicative of a change in material property with respect to the substrate, for example, the presence of a bubble on the sample surface. In general, soft materials absorb energy from the oscillating cantilever and cause large phase shifts. It can be concluded that the domains are composed of a much softer material than the underlying substrate, which is what you would expect for bubbles. Phase shifts of up to 40 have been observed in these experiments (not shown). The correlation between mean bubble height (shown in Table 1) and jump-in distance is good for ph 4.4, 5.6, and 9.4 conditions. For the case of ph 3.0, the jump-in distance is seen to vary between 23 and 38 nm (shown in Figure 6), the latter being in excess of the maximum bubble height (26 nm). This discrepancy could be rationalized by bubbles being present on the probe surface at ph 3.0 or by an attractive double layer interaction and subsequent jump of the interface toward the probe as discussed above. Figure 8 presents force-separation data for a tip interacting with the surface, that is, no attached colloid probe. Notice the presence of features identified in the previous probe-surface data: a long-range repulsion upon approach at high ph, the absence of repulsion at low ph, an anomalous postjump repulsion at ph 5.6, and a marked soft compliance regime. Given the smaller contact area of (43) Attard, P.; Miklavic, S. J. Langmuir, submitted.

6 Nanobubbles on a Hydrophobic Surface Langmuir, Vol. 18, No. 1, Figure 7. A series of AFM tapping mode images revealing nanobubbles on a hydrophobic glass surface. From (A) to (E), the images correspond to ph 5.6 (phase), ph 5.6 (height), ph 4.4 (height), ph 3.0 (height), and ph 9.4 (height) conditions. The sequence of height images reflects the order in which data were acquired. Phase and height information was acquired simultaneously. Cross sections, taken from left to right across the center, are shown adjacent to the height images. Table 1. Image Analysis Data Extracted from Tapping Mode Images to Reveal Mean Bubble Dimensions under ph 3.0, 4.4, 5.6, and 9.4 Conditions ph mean bubble size (std dev) [nm 2 ] coefficient of variation [%] mean bubble height (std dev) [nm] (2138) (3) (4824) (2) (2279) (2) (1641) (1) the tip (radius nm) in comparison with that of the colloid probe (radius 7500 nm), it is possible that contact is made with only a single bubble. As expected from the reduction in contact area, the maximum adhesion measured between tip and surface (1-10 nn) is less than that of the probe-surface system ( nn). During the adhesion between tip and surface, the cantilever is displaced over a smaller distance and subsequently no information was lost due to photodiode saturation. Longrange adhesive behavior is observed, characteristic of a bridging bubble between two surfaces. 30,44 Again, some steps can be seen in the withdrawal curves at large separations, characteristic of the rupture of submicroscopic bridging bubbles. (44) Preuss, M.; Butt, H.-J. Langmuir 1998, 14, Figure 8. Force-separation curves showing a hydrophobic glass surface withdrawing from a Si 3N 4 microfabricated cantilever tip. The inset shows the corresponding approach curves. The sample velocity is 0.4 µm s -1 for all curves. Taking the minimum in each curve as a reference point, from top to bottom, curves correspond to ph 9.4, 3.0, and 5.6, respectively. Discussion The values for mean bubble height are in broad agreement with image cross sections presented by Ishida

7 166 Langmuir, Vol. 18, No. 1, 2002 Tyrrell and Attard Figure 9. AFM tapping mode image at ph 5.6 immediately following ( min) a featureless contact mode image also in water ona1µm square in the right-hand corner of the picture, centered at about (2 µm, 0.8 µm). The features on either side of the 1 µm square probably correspond to the accumulation of tip debris arising from the large applied load exerted by the tip during contact imaging, since they cause little phase change. et al. 28 However, the lateral dimensions, shape, and distribution over the surface are quite different. In the present study, bubbles are seen to be distributed over the whole scan area, whereas the circular domains shown by Ishida et al. (diameter > 500 nm) adopt an almost linear configuration and occupy only a small portion of their treated sample. Given that Ishida et al. hydrophobize the whole area of their sample, the low surface coverage of bubbles is surprising. If the bubbles seen by Ishida et al. are a true reflection of reality, then their alignment might suggest nucleation along an imperfection in the sample, such as a scratch. However, there is no mention of any sample imperfections and the contact mode AFM image appears to reveal a smooth surface. 28 Furthermore, the large feature imaged in Figure 1A of ref 28 has a height of 40 nm, yet the corresponding force curve (Figure 2B of ref 28) shows a jump and deformation of only 20 nm, followed by rigid body contact. This suggests that the feature comprises a rigid 20 nm core surrounded by a 20 nm deformable corona. The effectiveness of the hydrophobization process is likely to affect the distribution of features over the sample surface. As discussed in greater detail below, vapor-phase silanation is associated with the generation of a robust monolayer, as opposed to treatments performed in the liquid phase. Certainly, the high advancing contact angle (101 ) and relatively low hysteresis (receding contact angle of 80 ) observed in the current study tend to support this notion and may help to explain differences between images presented here and elsewhere. 28 The domains shown in Figure 7 were observed only when the hydrophobized silica sample was immersed in water. No such features were observed when the sample was imaged in air (contact mode). When operated in contact mode in water, the cantilever appeared to disrupt or displace the bubbles as contact mode images failed to reveal features present in tapping mode images. In addition, during tapping mode the loss in image quality with an increase in drive amplitude, as reported earlier, is consistent with the removal or disruption of bubbles on the sample surface. Modulation of the cantilever at higher amplitude produced data reminiscent of the bare glass surface, and a subsequent reduction in amplitude saw the domains re-emerge, suggesting the regrowth of nanobubbles in the absence of high loads applied by the tip. The ability of the nanobubbles to regrow following disruption was further highlighted in a separate experiment (Figure 9). Tapping mode image data reveal nanobubbles to be present over the entire 3 3 µm scan area containing a 1 1 µm region previously imaged, both up and down the slow scan axis, in contact mode (the contact mode image was featureless). The cross-section trace in Figure 9 shows this region to be essentially indistinguishable from the whole scan. The time between scans was min, which indicates the time scale for regrowth. Given the intermittent contact between the tip and surface in tapping mode, it is thought that bubbles are unlikely to be induced by the probe and almost certainly pre-exist on the sample surface. In addition, all images were obtained with an untreated silicon nitride tip, which is generally regarded as hydrophilic and as such is unlikely to induce bubbles. One might question whether the silanation process itself gives rise to the domains seen in the images. Biggs and Grieser 45 observe no evidence of large polysiloxane structures on silica plates when hydrophobization is conducted in the vapor phase, as was the case in the present study (in liquid-phase hydrophobization, structures up to 60 nm in height were seen 45 ). They quote a layer thickness of 2 nm as measured using an elipsometric technique. This value is an order of magnitude less than the mean height of bubbles observed in the present study. Trip and Hair, 46 when performing silanation using trichloromethylsilane (TCMS) in the vapor phase, found no evidence of polymeric material on the surface of their silica sample. This conclusion was drawn from the absence of bands associated with Si-O-Si bonds in corresponding spectra. Parker et al. 11 used the monomer, dimer, and trimer to prepare silanated surfaces and showed the consequent hydrophobic attraction to be present in all three cases. Since the monomer is unable to polymerize, it follows that the (45) Biggs, S.; Grieser, F. J. Colloid Interface Sci. 1994, 164, 425. (46) Trip, C. P.; Hair, M. L. Langmuir 1992, 8, 1961.

8 Nanobubbles on a Hydrophobic Surface Langmuir, Vol. 18, No. 1, hydrophobic attraction is not due to polymer bridging. The removal of the features by contact mode imaging, and their subsequent re-emergence, is also inconsistent with polymer mounds, as is the reduction in hydrophobic attraction observed 18,19,22-25 in deaerated water. And even if, contrary to the preceding evidence, polymeric material were present on the sample surface, it would be hydrophobic in nature 2 and hence unlikely to extend into the fluid (H 2 O) stimulating a long-range attraction between surfaces. To further address the possibility of a polymer layer on the sample surface, a supplementary experiment was performed by the authors where a solvaphilic fluid (ethanol, contact angle ) 29 ) is introduced into the tapping mode AFM fluid cell. This resulted in the loss of phase contrast and the removal of associated features in the corresponding height image. The corresponding force curves between tip and surface in pure ethanol show no evidence of any long-range attractive force. This agrees with the trend observed by Parker et al., 11 who saw a reduction in the magnitude and range of the attraction upon approach in a 50% ethanol-water mix. Images display phase and height contrast when the cell is refilled with H 2 O. These observations are consistent with the nanobubble hypothesis. Summary and Conclusions The key results of this study can be summarized as follows. Increasing the ph increases the prejump repulsive component of the interaction, which suggests that it is electrostatic in origin, and appears to make the bubbles on the substrate smaller and more uniform. In addition, the decay length of the repulsion displays no dependence on approach velocity, ruling out any direct hydrodynamic contributions. The jump-in distance generally increases with decreasing approach velocity, but only when a prejump repulsion exists between bubble and probe. This distance is close to the height of the nanobubbles seen in the images. The extent of the soft compliance region increases with increasing velocity, which can be rationalized by invoking the compression of gas within a bubble, or bubbles, bridging the probe and surface. Analysis of the maximum adhesion force between probe and sample can be reconciled with capillary adhesion by assuming a contact angle of 80 for the probe. Steps in the retract curve at large separations may be attributed to the rupture of multiple submicroscopic bridging bubbles. The conformation adopted by bubbles on the sample surface appears to vary with ph. This is possibly due to changes in the magnitude of surface charge present on the bubbles. To conclude, the coupling of images with force curves displaying the hydrophobic interaction signature is powerful evidence in support of nanobubbles as the origin of the hydrophobic interaction. The strong correlation between height and phase images is striking. Phase images reveal that the domains are composed of a much softer material than the underlying substrate, as would be expected for bubbles. This conclusion is reinforced by the similar value of the jump-in distance and measured height of bubbles on the glass surface and by the observation that features regrow after their removal by contact scraping, hard tapping, or flushing with ethanol. Acknowledgment. Funding for the work was kindly provided by the Australian Research Council through the Special Research Centre for Particle and Material Interfaces at the Ian Wark Research Institute. We acknowledge the expertise of Anthony Quinn and Rick Fabretto in performing contact angle measurements on the glass surface and thank Kristen Bremmell for suggesting the experiment described in Figure 9. LA