Langmuir 2006, 22, 6951-6955
6951
Direct Visualization of Dewetting of Molecularly Thin Liquid Films on Solid Surfaces Kenji Fukuzawa* Department of Micro/Nano Systems Engineering, Nagoya UniVersity, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan, and JST-PRESTO, 4-1-8 Hon-cho, Kawaguchi 332-0012, Japan
Taichi Shimuta, Tomohiko Yoshida, and Yasunaga Mitsuya Department of Micro/Nano Systems Engineering, Nagoya UniVersity, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan
Hedong Zhang Department of Complex Systems Science, Nagoya UniVersity, Furo-cho, Chikusa-ku, Nagoya 464-8603 Japan ReceiVed March 20, 2006. In Final Form: May 30, 2006 The effect of the surface energy γ, disjoining pressure, Π, and roughness on the dewetting of molecularly thin liquid lubricant films on magnetic disks, which have sub-nanometer surface topography, has been investigated by visualizing the dewetting process directly using ellipsometric microscopy. The dewetting process of thin liquids on the rough surface is determined not only by the well-known instability of films, which is determined by the sign of dΠ/dh, but also by the sign of Π and the surface topography of the substrate even if its roughness is of the sub-nanometer order. The dewetting film formed small droplets, which were not along the surface topography of the substrate, when Π < 0. On the other hand, it formed grooves along the surface topography with a sub-nanometer roughness when Π > 0. Moreover, the sub-nanometer roughness initiated the dewetting of the metastable liquid thin films.
1. Introduction Control of liquid thin films on solid surfaces is very important in various fields such as colloid and interface science, tribology, microfluidics, and soft lithography.1-3 In computer hard disk drives (HDDs), the control of liquid films is necessary to improve the recording density. High-density HDDs require a very low flying height, less than 10 nm. To meet this demand, the current thickness of liquid lubricant films, which are necessary for good durability and reliability of HDDs, is molecularly thin and of the order of 1 nm. Therefore, distribution control of molecularly thin lubricant films is indispensable to high-density recording.4 Polar liquid is frequently used to improve the adhesion of liquid thin films onto solid surfaces. In recent HDDs, polar liquid lubricants have been widely used because the polar lubricant molecules can be tightly adsorbed to the magnetic disk surface by the polar intermolecular interaction between the polar end group and the overcoat carbon of the disk.5,6 However, the polar lubricants tend to form an autophobic surface, where the molecules of the first absorbed layer form an inert surface and those of the upper layer cannot adsorb to the disk surface, and they tend to dewet and cohere to each other. This means that a film with a * Corresponding author. Telephone: +81-52-789-2747. Fax: +81-52789-2747. E-mail:
[email protected]. (1) Kim, E.; Xia, Y.; Whitesides, M. Nature 1995, 376, 581-584. (2) Gleiche, M.; Chi, L. F.; Fuchs, H. Nature 2000, 403, 173-175. (3) Chou, S. Y.; Krauss, P. R.; Renstrom, P. J. Appl. Phys. Lett. 1995, 67, 3114-3116. (4) Mate, C. M.; Toney, M. F.; Leach, K. A. IEEE Trans. Magn. 2001, 37, 1821-1823. (5) Ma, X.; Gui, J.; Smoliar, L.; Grannen, K.; Marchon, B.; Jhon, M. S.; Bauer, C. L. J. Chem. Phys. 1999, 110, 3129-3137. (6) Izumizawa, S.; Jhon, M. S. J. Appl. Phys. 2002, 91, 7583-7585.
thickness of more than a monolayer may exhibit dewetting due to autophobicity.7-9 Dewetting has been attracting the interest of many researchers, and intensive research has clarified that the total surface energy profile γ(h) of the air/liquid/substrate interface determines the stability.7-14 Here, h is the thickness of the liquid film. The liquid film is unstable when γ′′ ()d2γ/dh2) < 0, whereas it is stable when γ′′ > 0. Therefore, when γ′′ < 0, dewetting occurs at every part of the lubricant, which is called spinodal dewetting. If we express the stability condition in terms of disjoining pressure Π ()-dγ/dh), which is the pressure between the liquid film and the substrate surface,15-17 the film is stable when Π′ < 0, whereas it is unstable when Π′ > 0. Most of the previous studies focused on thin liquids on flat substrates. In addition, most research on the relationship between the dewetting and topography of the solid surface aimed at control of the dewetting patterns, and there have been few reports on systematic analysis of the (7) Waltman, R. J.; Khurshudov, A.; Tyndall, G. W. Tribol. Lett. 2002, 12, 163-169. (8) Xu, L.; Ogletree, D. F.; Salmeron, M.; Tang, H.; Gui, J.; Marchon, B. J. Chem. Phys. 2000, 112, 2952-2957. (9) Guo, Q.; Li, L.; Hsia, Y.-T.; Jhon, M. S. J. Appl. Phys. 2005, 97, 10P3021-10P302-3. (10) Kim, H. I.; Mate, C. M.; Hannibal, K. A.; Perry, S. S. Phys. ReV. Lett. 1999, 82, 3496-3499. (11) de Gennes, P. G. ReV. Mod. Phys. 1985, 57, 827-863. (12) Reiter, G. Phys. ReV. Lett. 1992, 68, 75-78. (13) Redon, C.; Brochard-Wyart, F.; Rondelez, F. Phys. ReV. Lett. 1991, 66, 715-718. (14) Teletzke, G. F.; Davis, H. T.; Scriven, L. E. Chem. Eng. Commun. 1987, 55, 41-81. (15) Israelachivili, J. N. Intermolecular and Surface Forces; Academic Press: San Diego CA, 1992. (16) Derjaguin, B. V.; Chouraev, N. V. J. Colloid Interface Sci. 1978, 66, 389-398. (17) Mate. C. M. J. Appl. Phys. 1992, 72, 3084-3090.
10.1021/la060741z CCC: $33.50 © 2006 American Chemical Society Published on Web 07/01/2006
6952 Langmuir, Vol. 22, No. 16, 2006
relationship.18-21 Mougin and Haidara investigated the relationship between the surface energy profile and the chemical surface energy heterogeneities in order to clarify the effect of the chemical nanostructure on the dewetting thin liquid films.22 In our study, the effect of physical heterogeneities on the dewetting was focused on. The sub-nanometer- or nanometer-high topography may affect the dewetting because the thickness where dewetting occurs due to the intermolecular interaction is on the nanometer scale. Therefore, the effect of the surface topography on dewetting is a very important issue to be clarified for the stability of the molecularly thin liquid. In this paper, substrates that were flat but had sub-nanometer topography were selected, and the growth of the dewetting pattern of molecularly thin liquid films was directly visualized by ellipsometric microscopy. We report our investigation of the effect of the surface energy profile γ(h) and the surface topography on dewetting.
2. Materials and Methods Magnetic disks with a hydrogenated carbon overcoat whose diameter was about 88.9 mm (3.5 in.) were used as the sample substrate and a perfluoropolyether (PFPE) lubricant (FomblinZDOL4000, Solvay Solexis) was used as a sample liquid, because the disk surface was flat but had well-defined sub-nanometer topography and the interaction between the liquid lubricant and the disk has been intensively clarified. In the commercial production of magnetic disks, a sub-nanometer-high structure, which is called nanotexture, is fabricated by abrasive matching of rotating magnetic disks to improve the recording properties of magnetic disks. The surface topography of current magnetic disks should be industrially controlled with a sub-nanometer precision all over the disk surface, because the gap between the head and disk in HDDs is required to be of the order of 1 nm, as mentioned above. That was why magnetic disks were selected as a solid surface with a sub-nanometer topography. The molecular structure of the lubricant is given by
HO-CH2-CF2-[(O-CF2-CF2)p-(O-CF2)q]-O-CF2CH2-OH (1) where the ratio p/q is typically between 2/3 and 1. The lubricant, which is widely used in magnetic disk drives, is a linear chain polymer having a molecular weight of about 4000 g/mol, the estimated molecular diameter of the gyration is about 3 nm for the bulk, and the diameter of the linear chain is about 0.7 nm. The length of the molecule is about 14 nm. It has a hydroxyl group at each end and is considered to be adsorbed onto the carbon overcoat surface due to the interaction between the end groups and the carbon surface. The surface topography of the disk was obtained with a three-dimensional optical profiler (NewView 6000, Zygo). The lubricant was applied to sample disks by dip coating. In this method, which is used for the commercial production of HDDs, the substrate was immersed and then pulled out of the lubricant solution at a constant speed. The thickness of the liquid film on the substrate could be changed with sub-nanometer resolution by adjusting the removal speed and the lubricant solution density. The thickness of lubricant films was measured with a commercially available scanning ellipsometer (MARY-102, Five Lab). The solvent was hydrof(18) Rockford, L.; Liu, Y.; Mansky, P.; Russell, T. P. Phys. ReV. Lett. 1999, 82, 2602-2065. (19) Sharma, A.; Khanna, R. J. Chem. Phys. 1999, 110, 4929-4936. (20) Brusch, L.; Kuhne, H.; Thiele, U.; Bar, M. Phys. ReV. E 2002, 66, 0116021-011602-5. (21) Higgins, A. M.; Jones, R. A. L. Nature 2000, 404, 476-478. (22) Mougin, N.; Haidara, H. Europhys. Lett. 2003, 61, 660-666.
Fukuzawa et al.
luoroether (C4F9OC2H5, HFE-7200, 3M). The solvent is believed to evaporate quickly from the dipped lubricant films, and the obvious change in the lubricant thickness due to the solvent evaporation is not observed. However, the detailed study on the residual solvent in the dipped lubricant films has not been done yet, and the difference in the tribological performance due to the kind of the solvent was reported.23 In general, residual solvents are possible to affect the dewetting process of thin liquid films. In this study, the dewetting of the lubricant films was considered to be less affected by the solvent than by the other factors, because the dewetting patterns were determined by the surface energy profile and the topography of the substrate as mentioned below. The dispersive and polar surface energies of the liquid films on the disk were obtained by measuring the contact angle.15,24,25 Here, the polar surface energy is the nondispersive surface energy, which includes the hydrogen and dipole-dipole bondings. In this experiment, n-hexadecane and water were used as apolar and polar test liquids, respectively. Because some liquid films were unstable, the measurements were conducted as soon as possible after the liquid film application and angles were measured at 30 s after the test liquids were dropped onto the sample surfaces in order to reduce the influence of the growth of the film roughness. It should be noted that the measured contact angles were not the angles in equilibrium but the specific or relative angles measured at a certain time after the test liquids were dropped. The dispersive surface energy γLd was obtained from the contact angle θd using the Girifalco-Good-Fowkes-Young equation.
cos θd ) -1 +
x
2xγLdγTd ) -1 + 2 γT
γLd γT
(2)
where γL, γT, and γLT are the surface energies of the interfaces air/lubricant, air/test liquid, and the interface lubricant/test liquid, and subscript d denotes the dispersive components. Note that γTd ) γT for n-hexadecane because it does not have polar surface energy. Because the surface energy γT for n-hexadecane is known, the dispersive surface energy γLd can be obtained from the measured contact angle θd. Next, the polar surface energy of the liquid films γLp was obtained from the measured contact angle θp, the dispersive surface energy γLd, and the following Owens and Wendt equation26 and the Young Dupre equation:
γLT ) γL + γT - 2xγLdγTd - 2xγLpγTp
(3)
γLT ) γL - γT cos θp
(4)
The total surface energy γ was obtained as the sum of the dispersive and polar surface energies (γ ) γLd + γLp). It has been recently reported that this approach is inapplicable to some cases based on experimental studies27 and new semiempirical approaches have been presented.28-30 However, the method has not been fully established yet. In this study, we used the Owens and Wendt method in order to obtain the film thickness dependence of the surface energy because they reported that the (23) Waltman, R. J.; Tyndall, G. W.; Wang, G. J.; Deng, H. Tribol. Lett. 2004, 16, 215-230. (24) Adamson, W.; Gast. A. Physical Chemistry of Surfaces; Wiley: 1997. (25) Tyndall, G. W.; Waltman, R. J.; Pocker, D. J. Langmuir 1998, 14, 75277536. (26) Owens, D. K. and Wendt, R. C. J. Appl. Polym. Sci. 1969, 13, 17411747. (27) Cantin, S.; Bouteau, M.; Benhabib, F.; Perrot, F. Colloids Surf. 2006, 276, 107-115. (28) Li, D.; Neumann, A. W. J. Colloid Interface Sci. 1990, 137, 304-307. (29) Li, D.; Neumann, A. W. J. Colloid Interface Sci. 1992, 148, 190-200. (30) van Oss, C. J.; Chaudhury, M. K.; Good, R. J. AdV. Colloid Interface Sci. 1987, 18, 35-64.
Dewetting of Molecularly Thin Liquid Films
Langmuir, Vol. 22, No. 16, 2006 6953
Figure 1. Schematic of ellipsometric microscope for observation of nanometer-thick lubricant film.
results for the perfluoropolymer monolayers were in fairly good agreements with the results obtained by Zisman’s method.31 It should be noted that the absolute value of the surface energy is not discussed in this study, but, rather, the thickness dependence of the surface energy. The dewetting process was observed by ellipsometric microscopy.32,33 The schematic setup of the ellipsometric microscope, which was made by us, is shown in Figure 1. Ellipsometric microscopy can directly visualize the two-dimensional thickness distribution of molecularly thin liquids on solid surfaces. Like usual ellipsometry, it uses an optical phenomenon where the reflectivity of p-polarized light, which is parallel to the plane of incidence, is different from that of s-polarized light, which is perpendicular to the plane, when the light reflects on a surface covered with a liquid film. The difference in reflectivities for pand s-polarized lights depends on the thickness of the film. An imaging device such as a CCD camera is used as an optical detector in the ellipsometric microscope, and the change in polarization corresponding to the thickness distribution of the thin film is converted into an image with dark-bright contrast by an analyzer. In this study, a white light (500-W Xe lamp) was used as the light source instead of a laser to reduce the optical interference noise.33 The wavelength width was adjusted with an interference filter whose central wavelength was 632.8 nm and the wavelength width was about 10 nm. The light from a white light source is difficult to collimate because the source is a lamp and its size is large. In this setup, after the light from the light source was focused onto a pinhole to make the light beam small, the sample was illuminated with the collimated light. A highly sensitive chilled CCD camera (C5985, Hamamatsu) was used as the imaging device. The numerical aperture of the objective lens used was 0.1. The incident angle was set at 70°. The ellipsometric microscope can provide real-time imaging, where the frame rate is limited by that of the imaging device. In this experiment, the sample was mounted on a stepping-motor-driven translation stage and the sample position was shifted to enlarge the field of view. The observation was started about 6 min after film application. In addition, some samples were observed by atomic force microscopy (AFM), using a commercially available AFM (NanoScope IV, Veeco). A commercially available silicon probe was used, and we attached a carbon nanotube with a diameter of about 15 nm onto the probe tip. The spring constant of the probe was around 1-2 N/m, and the tapping mode was selected as the imaging mode. (31) Zisman, W. A. Ind. Eng. Chem. 1963, 55, 18-38. (32) Fukuzawa, K.; Noda, T.; Mitsuya, Y. IEEE Trans. Magn. 2003, 39, 898902. (33) Fukuzawa, K.; Nakada, A.; Mitsuya, Y.; Zhang, H. Trans. ASME J. Tribol. 2004, 126, 755-760.
Figure 2. Measured surface energy γ(h) (a) and disjoining pressure Π(h) ()-dγ(h)/dh) obtained from surface energy (b).
3. Results and Discussion A. Surface Energy Measurement. The measured surface energy is shown in Figure 2a. The surface energy γ(h) had a wavy form due to the polar surface energy although the dispersive surface energy decreased monotonically. It should be noted that γ(h) is gradually approaching the value for the lubricant with a macrothickness (22 mJ/m2), which is obtained by the manufacturer, as the thickness h increases. This supports the validity of our surface energy measurement. In addition, γ(h) had minima at a thickness of about 3 and 9 nm and a maximum at about 5 nm. It is considered that the wavy form corresponds to the layer structure formed by the lubricant molecules.5 Results that support this model, which were obtained by molecular dynamics simulation, have been recently reported.6 The film with a thickness of about 3 nm forms a monolayer, and the polar end group adsorbs to the disk surface, which makes the lubricant surface inert. The film with a thickness of about 5 nm has an active surface because many of the polar end groups do not adsorb to the disk. The film with a thickness of about 9 nm forms a triple layer and makes the surface inert because the polar end groups of the upper layer are adsorbed to the polar end groups of the middle layer. Figure 2b shows the disjoining pressure Π(h), which was obtained from the fitting curve of γ(h) shown in Figure 2a. The liquid film is stable when Π′ < 0 (γ′′ > 0) and unstable Π′ > 0 (γ′′ < 0), as described above. According to γ and Π shown in Figure 2a,b, films with a thickness of h ) 4-7 nm should be unstable, whereas films with a thickness of h ) 2-4 nm and 7-9 nm should be stable. B. Observation of the Dewetting Process. The surface topography of the sample disk is shown in Figure 3a, and ellipsometric microscope images of the liquid films with various thicknesses are shown in Figure 3b. The images of the liquid films were taken about 20 min after film application. Figure 3a shows that the disk had small grooves and ridges in the circumferential direction of the disk. The roughness of the
6954 Langmuir, Vol. 22, No. 16, 2006
Fukuzawa et al.
Figure 5. Relationship between the time constant of the growth of dewetting and the film thickness.
intensity of the liquid film image. σ(t) was approximated as
σ ) A0(1 - exp(- t/τ)) + σ0 Figure 3. Images of disk topography taken with an optical surface profiler (a) and lubricant with the ellipsometric microscope (b). Ellipsometric microscope images were taken about 20 min after film application.
Figure 4. Images of 5.8 nm thick lubricant taken with the ellipsometric microscope at different times. Arrows indicate the dewetted groove regions.
nanotexture was about 0.3 nm in root-mean-square roughness. The bright and dark regions are thick and thin regions of the liquid films in Figure 3b. As predicted above, the film with a thickness of h ) 2.6 nm was stable, and the films with a thickness of h ) 4.3 and 5.8 nm were unstable and dewetted. The films with a thickness of more than 5 nm, where γ(h) was maximum as shown in Figure 2a, formed a dewetting pattern shaped like a stripe along the nanotexture in the circumferential direction. It is interesting that the dewetting pattern along the nanotexture formed although the roughness of the nanotexture was much smaller than the diameter of the liquid molecule (about 3 nm). The change in the film with a thickness of h ) 5.8 nm with the passage of time is shown in Figure 4. The stripe-shaped dewetting pattern consisted of a groove with ridges on both sides. The growth rate in the circumferential direction was much faster than that in the radial one. The groove grew as a crack did. Here, it should be noted that the results shown in Figures 3 and 4 differ in two aspects from the prediction according to Figure 2. First, the dewetting occurred in the film with a thickness of h ) 8.6 nm, which should be stable because γ′′ > 0. Second, the dewetting patterns were quite different from those of the films with a thickness of h ) 4.3 and 5.8 nm, which should be unstable and have no difference in the sign of γ′′ ( 7 nm had a larger time constant than those with a thickness of h ) 5-7 nm. This result suggests that the dewetting mechanisms of the films of these two regions are different. In general, even if γ′′ > 0, dewetting occurs when the curve of γ(h) has a local minimum point (h ) 3 nm in Figure 2a) and the lubricant tends to separate into wet and dewetted regions.13 This state is called metastable, and dewetting does not occur homogeneously but with stochastic generation of dewetted regions starting from defects (heterogeneous dewetting). The films with a thickness of h ) 5-7 nm had time constants that were almost independent of the thickness, whereas the time constant for the films with a thickness of h > 7 nm increased as the thickness increased. This result supports the dewetting of films with a thickness of h ) 5-7 nm occurring homogeneously and that of films with a thickness of h > 7 nm occurring heterogeneously, where the dewetting starts only when the films stochastically get the energy fluctuation necessary to overcome the energy barrier at h ) 5 nm and reach the local minimum of h ) 3 nm. The energy barrier (γ(5nm) - γ(h)) increased as the film thickness increased for the films with a thickness of h > 7 nm from γ(h) in Figure 2a. In addition, the dewetting started from the nanotexture because it occurred along the nanotexture. This suggests that the nanotexture initiated the separation of dewetted and thick film regions. Now, we discuss the second point. The dewetted films with a thickness of h < 5 nm formed droplets whereas those with a thickness of h > 5 nm formed grooves along the nanotexture, as shown in Figures 3 and 4. An ellipsometric microscope image of a film that had a partially thick region (h ) 5.0-6.2 nm) is shown in Figure 6a, and its film thickness distribution measured with a scanning ellipsometer is shown in Figure 6b. The ellipsometric microscope image was taken about 23 min after film application. This indicates that the dewetting pattern is determined by its thickness and is quite different from those of the films with a thickness of h > 5 nm and h < 5 nm. Considering the relationship between Π and h shown in Figure 2b, Π < 0 when h < 5 nm, whereas Π > 0 when h > 5 nm. Moreover, AFM images of liquid films with a thickness of h ) 4.4 and 5.6 nm are shown in Figure 7. The images were taken about 2.5 h after film application. It should be noted that the observation was done for samples left longer than those shown in Figures 3, 4,
Dewetting of Molecularly Thin Liquid Films
Langmuir, Vol. 22, No. 16, 2006 6955
Figure 8. Model of dewetting for films with Π < 0 (a) and Π > 0 (b).
Figure 6. Coexistence of different types of dewetting. (a) Ellipsometric microscope image of 4.3 nm thick lubricant with a thick part with a thickness of 5.0-6.2 nm. The image was taken about 23 min after film application. (b) The thickness distribution around the coexistence region measured with a scanning ellipsometer.
Figure 7. AFM images of lubricant with a thickness of h ) 4.4 nm (a) and h ) 5.6 nm (b). Images were taken about 2.5 h after film application. The scan size was 98.7 µm × 98.7 µm2.
and 6. Droplets grew in the film with a thickness of h ) 4.4 nm (Π < 0), and a stripe-shaped pattern grew in the film with a thickness of h ) 5.6 nm (Π > 0). This result coincides with those obtained by the ellipsometric microscopy. These results suggest that the difference in the dewetting pattern is determined by the sign of Π. When the interaction between the liquid molecules is more attractive than that between the molecule and substrate, Π < 0. Therefore, the molecules tend to cohere to each other and form droplets. In contrast, when Π > 0, the molecules do not cohere because the interaction between the molecule and substrate is more attractive than that between the molecules.7 This may cause the separation of thinner and thicker parts of the film. These results can be explained using the model shown in Figure 8. Here, it should be noted that liquid molecules flow from a low disjoining pressure region to a high-pressure region. We assume that a protruding region with Π > 0 (h > 5 nm) is formed by thermal fluctuation in the liquid film with Π < 0 (h < 5 nm) (Figure 8a). Once a protruding region with Π > 0 has formed, the liquid molecules flow from the surrounding flat film to the protrusion, because Π > 0 for the films with h > 5 nm and Π < 0 in the surrounding flat film (Figure 2b). Thus, the
dewetting starts from the protrusion, and the liquid molecules cohere to each other and form droplets. Moreover, when we examined Figure 7a in detail, we observed thin ridges along the nanotexture. This suggests that the nanotexture may initiate the thickness fluctuation. The protrusion shown in Figure 8a may form a thin and short ridge along the nanotexture and may quickly rupture into droplets due to the Plateau-Rayleigh instability.34,35 On the other hand, we assume that a groove region with Π < 0 (h < 5 nm) is formed by thermal fluctuation in the liquid film with Π > 0 (h > 5 nm) (Figure 8b). In the same manner as that for the films with Π < 0, once a groove region with Π < 0 has formed, the liquid flows from the groove to the surrounding film, because Π < 0 for the films with h ) 3-5 nm and Π > 0 in the surrounding film. The liquid continues to flow until the thickness decreases to about 3 nm. In addition, the nanotexture initiates the thickness fluctuation, because the groove was formed along the nanotexture in the experiment. As mentioned when discussing Figure 4, the growth rate in the circumferential direction, which was along the nanotexture, was much faster than that in the radial one. This result supports the idea that dewetting starts from the nanotexture. In films with Π > 0 (h > 5 nm), the liquid molecules are apt to take the form of the film and not to cohere, as mentioned above. Therefore, the dewetting causes the separation of the groove and flat film. In addition, when we examined Figures 4 and 7b in detail, we observed droplets in the groove (groove A in Figure 4). This suggests that the dewetting film in the groove forms droplets as explained in discussing Figure 8a when its thickness becomes less than 5 nm, where Π < 0.
4. Conclusions From observations of the dewetting patterns of molecularly thin liquid lubricant films on magnetic disks, we found that the growth of the dewetting is determined not only by the wellknown instability of films for liquids on flat substrates, which is determined by the sign of dΠ/dh, but also by the sign of Π and the surface topography of a disk even if its roughness is of the sub-nanometer order. The dewetting formed small droplets, which were not along the surface topography of the substrate, when Π < 0. On the other hand, it formed grooves along the nanotexutre when Π > 0. In addition, the nanotexture initiated the dewetting in the metastable liquid films. Acknowledgment. This work was supported in part by the Japanese Ministry of Education, Culture, Sports, Science and Technology under Grant No. 17206014 and by the Storage Research Consortium. LA060741Z (34) Brochard, F.; Redon, C. Langmuir 1992, 8, 2324-2329. (35) Reiter, G. Langmuir 1993, 9, 1344-1351.
