Langmuir 1989,5, 567-570
567
Influence of Surface Melting Characteristics on the Wetting Behavior of Solid Adsorbed Films? R. Chiarello and J. Krim* Department of Physics, Northeastern University, Boston, Massachusetts 02115 Received October 27, 1988 Surface melting is in principle a widespread phenomenon and has recently been observed for physisorbed films on graphite and magnesium oxide substrates. Neutron diffraction studies of oxygen films adsorbed on graphite have revealed a logarithmic temperature dependence for the quantity of fluid present in the surface melting regime and have confirmed a close link between film melting and wetting properties. These studies also suggest that multiple transitions from incomplete to complete wetting may be present, each occurring at a bulk triple point. Preliminary investigations of bulk oxygen indicate that its surface melting behavior may be quite distinct from that of the film phase. I. Introduction The wetting behavior of a film adsorbed on a solid substrate is an important issue in surface science concerning a wide range of fundamental and applied problems. Wetting behavior falls into two general categories.' In systems which exhibit complete wetting, the film thickens monotonically as the vapor pressure surrounding the substrate increases. Incomplete wetting refers to a material that, as the pressure increases, forms only a thin film, which coexists with bulk droplets or crystallites. Continuous transitions from incomplete to complete wetting at bulk triple points Tt are extremely common.2 Arguments for the occurrence of continuous triple point wetting suggest that a metastable extension of the bulk liquidvapor coexistence line persists into the bulk solid regime at temperatures below Tb3 This would require that the film (or film surface) melt at temperatures lower than Tt and would imply that surface melting is present whenever a triple point wetting transition occurs. Premelting on any surface, be it that of a film, 3D crystal, or otherwise, is itself based on broad thermodynamic grounds and is predicted to occur whenever the solid phase is wetted by its own liquid phase.4 Although for many years this behavior was believed to be widespread, it has only recently been confirmed by experiments on three-dimensional (3D) lead crystals5 and physisorbed films on graphitee8 and magnesium oxidee substrates. The temperature dependence of the thickness of the surface-melted layer is of great interest. This information allows understanding of the mechanisms which control both surface melting and 3D melting. Theories which assume short-range interactions predict a logarithmic dependence, while long-range interactions produce a power law dependence.l0 Not all experiments on adsorbed films have revealed the same temperature dependence. Oxygen films on graphite which are 2-4 layers thick' and thin methane films on magnesium oxideg show a logarithmic temperature dependence, while studies of 2G40-layer-thick argon films indicate a power law dependence for the melted layer thickness? One suggested reason for this discrepancy is that the argon films are dominated by long-range interactions while the relative thinness of the oxygen and methane films induces a logarithmic temperature depend e n ~ e .If~ so, a crossover from logarithmic to power law behavior is expected as the adsorbed film thickens. Such behavior has recently been observed both experimentally6 Presented a t the symposium on "Adsorption on Solid Surfaces",
62nd Colloid and Surface Science Symposium, Pennsylvania State University, State College, PA, June 19-22, 1988; W. A. Steele, Chairman.
and theoretically." We therefore anticipate that a study of surface melting in thick oxygen films or bulk oxygen should reveal a power law dependence. We describe here such a study of the surface melting of bulk oxygen adsorbed on graphite, beginning with a summary of the results obtained for oxygen f i l m ~ . ~ ~ ~ 11. Surface Melting of Adsorbed Oxygen Films A. Experimental Details. The neutron-scattering measurements were performed at Grenoble's Institut Laue-Langevin (ILL) on the D l b beam line operating in the elastic mode. The wavelength of the incident neutrons was 2.517 A. Elastically scattered neutrons were collected within the range 0.87-2.86 A-1. The overlayer diffraction patterns were obtained by subtracting off the background due to the bare graphite. The substrate consisted of 1.36 g of Papyex exfoliated graphite, with a density of 0.1 g/cm3 and having a surface area estimated to be 36 m2/g. Surface coverage was calibrated with respect to that of a d 3 X d 3 registered monolayer phase (0.064 mol/A2) of methane. The oxygen was condensed at 55 K and then slowly cooled to the temperature of interest. Annealing after cooling produced no change in the observed diffraction scans. The measurements were carried out between 34.5 and 54.5 K, and for coverages ranging from 0.24 to 0.9 mol/A2, the equivalent of 2-8 solid layers (density 0.11 mol/A2) or 3-12 liquid layers (density 0.075 mol/A2). B. Analysis and Results. Figure 1 shows the scans recorded at a surface coverage of 0.32 mol/A2. The scan at 49.5 K has been recorded well below the bulk triple (1) Pandit, R.; Schick, M.; Wortis, M. Phys. Reu. B 1982, 26, 5112. (2) Bienfait, M. Surf. Sci. 1985, 162, 411. (3) Krim, J.; Dash, J. G.; Suzanne, J. Phys. Reu. Lett. 1984,52, 640. (4) Kristensen, J. K.; Cotterill, R. M. J. Philos. Mag. 1987, 36, 437. Recent reviews include: Nenow, D.; Trayanov, A. J. Cryst. Growth 1986, 79, 801. van der Veen, J. W. M.; Phis, B.; Denier van der Gon, A. W. In Chemistry and Physics of Solid Surfaces; Springer-Verb New York, 1988; Vol. VII. Dash, J. G. In Proc. Soluay Conference on Surface Science; Springer-Verlag: New York, 1988. (5) Frenken, J. W. M.; van der Veen, J. F. Phys. Rev. Lett. 1986,54, 134. Frenken, J. W. M.; Maree, P. M.; van der Veen, J. F. Phys. Reo. B 1986,34,7506. Frenken, J. W. M.; Toennes, J. P.; Woll, Ch. Phys. Reu. Lett. 1988, 60, 1727. Phis, B. et al. Phys. Reu. Lett. 1987, 59, 2678. (6) Zhu, D. M.; Dash, J. G. Phys. Rev. Lett. 1986,57, 2959. Zhu, D. M.; Dash, J. G. Phys. Reo. Lett. 1988,60,432. Zhu, D . M.; Pengra, D.; Dash, J. G. Phys. Reu. B 1988,37, 5586. (7) Krim, J.; Coulomb, J. P.; Bouzidi, J. Phys. Rev. Lett. 1987,58,583. (8) Chiarello, R.; Coulomb, J. P.; Krim, J.; Wang, C. L. Phys. Rev. E 1988, 38, 8967. (9) Bienfait, M. Europhys. Lett. 1987,4, 79. Gay, J. M.; Suzanne, J.,
private communication. (10) Lipowsky, R. Phys. Reu. Lett. 1982,49,1575. Lipowsky, R. Phys. Reu. B 1985, 32, 1731. (11) Trayanov, A,; Tosatti, E. Phys. Reu. Lett. 1987, 59, 2207.
0 1989 American Chemical Society
568 Langmuir, Vol. 5, No. 3, 1989
I
Chiarello and Krim
1
2000
g
2000
1
I-
u aJ
1
0
2
2.2
2.4
2.6
c, 0
2
2.2
2.4
2.6
Q (k’) Figure 1. Diffraction scans recorded at a surface coverage of 0.32 mol/A2. Bulk y peaks are observed at T = 46.5 K. Too little of the bulk y phase is present at 49.5 K to be observed in the diffraction scan. point, yet it reveals a fluid-phase film. The scans between 41.5 and 46.5 K have been recorded in the surface-melted-film regime. All of the surface-melted-film scans show evidence of a compound film phase, indicated by the superposition of solidlike peaks with a fluid film background. The scattering intensity at Q = 2.0 A-1 is due only to the fluid film phase and is well away from any film or 3D solid peak. We employ the intensity at this point to estimate the quantity of fluid film present in each of the surfacemelted-film scans and then subtract the fluid film portion so as to reveal the line shape of the solid component. The resulting scans (for total surface coverages of both 0.24 and 0.32 mol/A2, excepting 0.32 mol/A2 at 46.5 K), including our best fit, are shown in Figure 2. For reference, we also include a scan recorded in the 2D solid regime at 0.24 mol/A2 and 34.5 K (Figure 2b). To determine the structure of the underlying solid, the diffraction data were fit by following a method described previously in great detai1.12J3 Multilayer line shapes were obtained by assuming the surface monolayer line shape is modified by a structure factor, which describes multilayer interference effects. For each possible molecular stacking arrangement (AA, AB, ABA, or ABC),the multilayer line shape has a well-defined characteristic functional form. It is the ordered arrangement of molecules in a given stacking sequence, along with the in-plane structure, that provides the characteristic shape of the multilayer diffraction pattern. The best fit to the data corresponded to O2 molecules oriented perpendicularly to the substrate and arranged in equilateral triangles within the planes. The solid component of the film grew in an ABC stacking arrangement with lattice parameters quite close to that of the 3D 0
Figure 2. Diffraction scans of the solid component of the surface-melted film along with the beat fit to the data. (a)0.32mol A2 at 41.5 K (liquid subtracted);a, = 2.3 layers; (b) 0.24 mol/A at 34.5 K ({ solid regime), n, = 2.2 layers; (c) 0.24 mol A2at 39.5 K (liquid subtracted), n, = 2.1 layers; (d) 0.32 mollda at 44.5 K (liquid subtracted),n, = 1.6 layers; (e) 0.24 mol/A2 at 44.5 K (liquid subtracted),n, = 0.7 layers.
c
Table I. Film Characteristics in the Surface Melting
Raaime
n,mol/A2 q,mol/A2
n,, mol/A2
nl, layers n,, layers
34.5 0.24 0 0.24 0 2.19
39.5 0.24 0.015 0.225 0.2 2.11
44.5 0.24 0.165 0.075 2.2 0.72
41.5 0.32 0.062 0.258 0.83 2.28
44.5
46.5 ~~~~
0.32 0.145 0.175 1.93 1.63
0.32 0.24 0.08 3.2 0.73
phase of oxygen. This rhombohedral phase is stable between 24 and 43.8 K and consists of an ABC stacking arrangement with 3.76 A between the individual planes of molecule^.'^ The results are summarized in Table I and are in excellent agreement with recent computer simulations of oxygen on graphite.16 The complete details of the fit are published elswherea8As the temperature increases from 40 to 48 K, the quantity of liquid increases. Assuming the liquid density to be 0.075 mol/A2 per layer, the surface-melted liquid layer thickness, nl, is well described by nl = 12.9 - 4.70 In (54.4 - 2“)
(1)
Figure 3a shows the number of liquid layers for coverages of 0.24 mol/A2 (diamonds) and 0.32 mol/A2 (squares) along with our best fit to the data, eq 1. The cross at 48 f 1 K and four liquid layers corresponds to the “feature” reported in heat capacity studies18J7at 47 K and microbalance studies’* at 49 K. We associate this feature with the high-temperature limit of the surface melting regime. The cross at 39 f 1K and 0 liquid layers corresponds to the low-temperature limit of the “fluid 11” regime reported (14)De Fotis, G.C. Phys. Reu. B 1981,23,4714.
(12)Kjems, J. K.; Passell, L.; Taub, H.; Dash, J. G.; Novaco, A. D. Phys. Reu. E 1976,13,1446. (13)Larese, J. Z.;Harada, M.; Passell, L.; Krim, J.; Satija, S. Phys. Rev. B 1988,37,4735.
(15)Bhethanabotla, V.; Steele, W. Con. J. Chem. 1988,66,866. (16)Awschalom, D. D.; Lewis, G. N.; Gregory, S. Phys. Rev. Lett. 1983,
51, 586.
(17)Stoltenberg, J.; Vilohes, 0. E. Phys. Rev. B 1980,22,2920. (18)Bartosch, C. E.; Gregory, S. Phys. Reu. Lett. 1985,54, 2513.
Influence of Surface Melting on Wetting
Langmuir, Vol. 5,No. 3, 1989 569 A
Y
Y
2 N I
2
2.25
2.5
2.75
Ln(T,-T) Figure 3. (a) Number of liquid layers as a function of In (2' 2") for coverages of 0.24 mol/A2 (diamonds) and 0.32 mo1/k2 (squares)layers, respectively. the solid line corresponds to the best logarithmic fit to the data, eq 1. Crosses denote the beginning and end of the surface-melted regime, as determined in previous studies (see text).1s1s (b) In nl as a function of In (Tt - 7') for coverages of 0.24 mol/A2 (diamonds)and 0.32 mol/A2 (squares) layers, respectively. The dashed line corresponds to the best logarithmic fit to the data. The dotted line corresponds to a power law with exponent -1/3.
in X-ray diffraction studies.lg Our study does not allow us to determine whether the surface melting is occurring in a continuous or layer-by-layer manner. Throughout the surface melting regime nl is within experimental error of an integer number of layers, suggestive of layer-by-layer melting. The evidence is, however, far from conclusive. Figure 3b shows In nl plotted as a function of In (Tt 7'). The dashed curve shows the best fit to the data and is not linear. The dotted line shows the expected behavior if the data were to follow the power law nl a (Tt - 7')-lI3. The logarithmic temperature dependence is preferred over the power law behavior. No film phase occurs, which is similar to 3D y oxygen, the bulk phase which is stable between 43.8 and 54.4 K. We conclude that both the /3 phase and the liquid phase wet graphite more strongly than y oxygen. Complete wetting by the y phase is therefore prevented. This is apparently due to the attractive nature of the substrate, and a preference for the higher density /3 phase results. The results are quite suggestive that multiple wetting transitions may be present. If the /3 phase indeed wets graphite, then a transition to incomplete wetting would occur at the /3-r triple point (43.8 K). A transition from incomplete to complete wetting is known to occur at the y-liquid triple point (54.4 K). The 46.5 K scan in Figure 1 is suggestive of this possibility. At this temperature, bulk y peaks are present and show coexistence with a surfacemelted-film phase. This behavior is indicative of incomplete wetting. No bulk y peaks are observed at either higher or lower temperatures. This shows that thicker film phases are able to form at both higher and lower temperatures, since the surface coverage is a constant throughout Figure 1. Incomplete wetting by the oxygen films near the melting point prevents the study of surface melting in thick adsorbed films. I t does not, however, prevent study of the (19) Heiney, P. A.; Stephens, P. W.; Mochrie, S. G. J.; Akimitsu, J.; Birgeneau, R. J.; Horn, P. M. Surf. Sci. 1983,125, 539.
I
1.9
4
N I
b
--
-T 1.75
-
n
2
2.1 2.2
2.3 2.4
Figure 4. Bulk diffraction scans recorded at an equivalent film coverage of 20 4 3 x 4 3 registered monolayers. The solid lines are the best fits to the data. Arrows indicate the 3D y oxygen
(210) and (211) diffraction peak postions, respectively. (a) 50.3 K, nl = 0.93 mol A2,n, = 0.35 mol/A2;(b) 52.0 K, nl = 0.99 mol/A2, n, = 0.29 mol/L2; (c) 53.6 K, nl = 1.11mol/A2 n, = 0.17 rnol/A2; (d) 54.3 K nl = 1.26 mol/A2,n, = 0.02 mol/A2; (e) 55.0 K, nl = 1.28 mol/A2.
surface melting behavior of the bulk y crystallites, which form when oxygen is introduced to the cell at the saturated bulk vapor pressure. Such a study is described in the next section. 111. Surface Melting of Bulk Oxygen A. Experimental Details. Neutron diffraction measurements on bulk O2were performed a t the high flux beam reactor of Brookhaven National Laboratory (BNL) on beam line H5 operating in the elastic mode. The instrumental resolution was 0.0123 A-' half peak width at half maximum intensity (hwhm), and the incident neutron wavelength was 1.651 A. Data were collected over the range 1.94-2.40 A-l. The vermicular graphite (Union Carbide GTA grade) substrate was loaded into a thinwalled aluminum sample cell mounted to a closed cycle helium refrigerator. The O2 was condensed at approximately 60 K and then cooled to the desired temperature. Diffraction scans were recorded at coverages ranging from 3 to about 20 4 3 x 4 3 registered monolayers. Surface coverage was calibrated by matching the liquid-phase normalized integrated diffraction intensity of the ILL data to that of the BNL data in the region of overlap and then extrapolating to higher coverage, assuming the integrated diffraction intensity to be proportional to surface coverage. B. Analysis and Results. Figure 4 shows the diffraction scans recorded in the range 50.3-55 K at an equivalent film coverage of 20 4 3 x 4 3 registered monolayers (1.28 mol/An). The scans were recorded above 50 K in order to avoid complications due to 2D solid peaks. (The 2D solid has completely melted at or below 49 K.) The scans taken throughout this temperature range are indicative of a 3D solid (sharp, narrow diffraction peaks) coexisting with a fluid phase (broad Lorentzian back-
570 Langmuir, Vol. 5,No. 3, 1989
Chiarello and Krim 1
Table 11. Bulk Surface Melting Characteristics
I
,
,
,
I
,
,
I
,
,
,
, , ,
,
T,K n, mol/A2 nl, mol/A2 (total) nl, mol/A2 nl, mol/A2 (bulk) n,, mol/A2 dzu, A 4109
L,A
A
2.41, A - ' b
rl,A-1
50.3
52.0
53.6
54.3
55.0
1.28 0.93 0.29 0.64 0.35 3.04 2.77 525 2.1 0.36
1.28 0.99 0.37 0.62 0.29 3.04 2.77 531 2.09 0.34
1.28 1.11 0.43 0.68 0.17 3.04 2.78 465 2.08 0.36
1.28 1.26 0.51 0.75 0.02 3.04 2.78 512 2.08 0.34
1.28 1.28
-
0
2.08 0.36
t
4
t
-0.3
7
k
Quantity of fluid on the portion of the substrate not covered by oxygen crystallites. The oxygen crystallites cover a maximum of 3% of the available surface area for adsorption. *pl and rl are the liquid peak position and full width at half maximum, respectively.
ground). The scan at 55 K is taken above the bulk triple point (54.4 K). Our best fits to the data are also shown by the solid lines in Figure 4. The scans were fit with a composite function consisting of the usual 3 D solid Bragg diffraction intensity functionm plus a Lorentzian, the functional form typically employed to describe scattering from a liquid. The least-squares fit to the 3 D peaks assumed the coherence length ( L ) ,the lattice parameters (d210and d211),and the normalized intensity to be adjustable parameters. The fit parameters of the Lorentzian function included the position of the fluid peak (pJ, the full width at half maximum (I'l), and the normalized intensity. The scan recorded above the bulk triple point (54.4 K)at 55 K was best fit by just a Lorentzian. The best-fit parameters are listed in Table 11. The peaks in Figure 4 at 2.07 and 2.27 A-' correspond to the (210)and (211)diffraction peaks of bulk y 02,respectively.21*22The coherance length L is determined by the width of the 3D solid diffraction peak. The peaks in Figure 4 are instrumental resolution limited, so the crystallites must have a coherance length of at least 510 A. If we assume the crystallites are cubic with edge length 510 A, we estimate the maximum area occupied by crystallites to be 3% of the total surface area available for adsorption. (Larger crystallites would result in less occupied area.) No peak broadening is observed as the temperature increases. I t can be seen from Table I1 and Figure 4, however, that as temperature increases from 50.3 to 54.3 K,the intensity of the (210)and (211)solid peaks decreases, while the intensity of the fluid background increases. This cannot alone be attributed to the increasing thickness of the liquid film adsorbed on the uncovered portions of the graphite substrate (i.e., the triple point wetting transition which is occurring in the film phase): we attribute the fluid excess to surface melting of the bulk crystallites. At 55 K the O2 crystallites have completely melted, as indicated by the scattering scan composed only of a Lorentzian. The integrated intensity from the liquid scattering portion of each scan is indicative of the total amount of oxygen in a fluid phase at any given temperature. Since the limiting thickness of the adsorbed liquid film at bulk condensation is known from ellipsometry,23we can subtract to estimate the quantity of fluid present on account of surface melting of the bulk crystallites. The values for the various temperatures are listed in Table 11. (20) See, for example: Warren, B. E. X-ray Diffraction; AddisonWesley: Reading, 1969. (21) Barett, C. S.; Meyer, L. Phys. Reu. 1967, 160, 694. (22) Barett, C. S.;Meyer, L.; Wasseman, J. Phys. Rev. 1967,163,851. (23) Drir, M.; Hess, G. B. Phys. Rev. B 1986, 33, 4758.
-0.5
1 -2
I
I
I
I
I
I
I
4
8
-1
1 , 0
0 I
I
I
I
I
I
1
Ln(T,-T)
Figure 5. (a) Quantity of liquid as a function of in (Tt- 7') for bulk oxygen (20 4 3 x 4 3 registered monolayers). Boxes correspond to nl (melted) given in Table 11. The solid line is the best logarithmic fit to the data. (b) In nl as a function of In (Tt - 7') for bulk oxygen. The solid line is the best power law fit to the data, which provides an exponent of -1/18.
We employ the quantity of liquid attributable to bulk surface melting as a rough gauge of the thickness of the surface-melted layer. In contrast to the results for oxygen films, there is no clear preference between the logarithmic (Figure 5a) or power law (Figure 5b) behaviors. Without information on the size and shape of the crystallites, we are unable to carry out a more definitive analysis.
IV. Discussion Our studies have confirmed a close link between the triple point wetting transition and surface melting. Indeed, it stands to reason that if the surface free energy of the substrate-fh combination can be lowered by the presence of an adsorbed liquid film below Tt (triple point wetting), then the solid phase of the adsorbate might also be wetted by its own liquid (surface melting). The converse does not appear to be the case. Complete wetting by a solid adsorbed film precludes the occurrence of a triple point wetting transition yet does not appear to preclude surface melting, as evidenced by the Arlgraphite systeme6 The 02/graphite triple point wetting transition has facilitated our studies of surface melting in the film phase. Unfortunately, it has greatly complicated our studies of surface melting in bulk oxygen. We have obtained evidence that surface melting does indeed occur, and it appears to be quite distinct from that observed for the film phase. Nonetheless, the presence of the graphite substrate and the film-phase wetting transition and the lack of precise knowledge concerning crystallite size and geometry prevent us from distinguishing between a logarithmic or power law temperature dependence. Such information could ideally be obtained by employing a substrate solid oxygen was known to wet or by studying bulk oxygen in the absence of any substrate. Acknowledgment. We are greatly indebted to J. P. Coulomb, our collaborator for the film experiments at ILL, and to J. Z. Larese and L. Passel1 for facilitating the experimental run at BNL. This research has been supported by the Low Temperature Program of the National Science Foundation, Grant DMR-8657211. Registry No. 02,7782-44-7; graphite, 7782-42-5.
