pubs.acs.org/Langmuir © 2010 American Chemical Society
Adsorption of Gemini Surfactants with Dodecyl Side Chains and Different Spacers, Including Partially Fluorinated Spacers, on Different Surfaces: Neutron Reflectometry Results Pei Xun Li,* Chu Chuan Dong, and Robert K. Thomas Physical and Theoretical Chemistry Laboratory, South Parks Road, Oxford OX1 3QZ, U.K. Received October 27, 2010. Revised Manuscript Received December 8, 2010 Neutron reflectometry has been used to study the adsorption of two symmetrical cationic (dimethyl ammonium bromide) gemini surfactants with two C12H25 chains and different partially fluorinated spacers at three different surfaces: air/water, hydrophilic silica/water, and hydrophobic (octadecyltricholorosilane (OTS))/water. In addition, the adsorption of purely hydrocarbon geminis with the same side chains and spacers of different lengths has been studied at the same two solid surfaces. The limiting close-packed areas for the two fluorocarbon geminis, C12-C3fC6C3-C12 and C12-C4fC4C4-C12, are 92 and 72 ( 4 at the hydrophilic silica surface, 81 and 89 ( 4 at OTS, and 137 and 106 ( 4 A˚2 at the air/water interface with decreases of 38 and 24% from air/water to the average solid value, respectively. These changes suggest that the packing at the air/water interface is inefficient, and this allows the extra hydrophobicity of the chain environment at the two solid surfaces to promote much more efficient packing. At the air/water interface, the fluorocarbon spacers are on average the fragments furthest away from the underlying water, further out than in the nearest comparable hydrocarbon gemini, C12-C12-C12. This is the probable explanation of the much lower value of the area per molecule at the air/water interface of C12-C4fC4C4-C12 compared to that of C12-C12-C12. It is also the probable cause of the inefficient packing of the hydrocarbon side chains. At the more hydrophobic OTS surface the situation is reversed and the fluorocarbon spacers are now the furthest from the hydrophobic surface, further out than the spacer in C12-C12-C12. This is an unusually large structural change that must be associated with the greatly improved packing at the OTS surface. The efficiency of the packing is also high for the hydrophilic surface, no doubt because the hydrocarbon chains can interact favorably in the adsorbed bilayer core. The values of the area per molecule obtained for the series of hydrocarbon geminis at the air/water, OTS/water and silica/water interfaces are respectively 139, 104, and 98 ( 4 A˚2 for C12-C12-C12, 114, 106, and 94 ( 4 A˚2 for C12-C10-C12, 104, 84, and 85 ( 4 A˚2 for C12-C6-C12, and 78, 66, and 70 ( 3 A˚2 for C12-C3-C12. The area per molecule is also about 20% less on average at the two solid surfaces than at the air/water interface. This can also be attributed to more efficient packing caused by the more favorable hydrophobic interactions possible at these two surfaces than at the air/water interface, again showing that the packing at the air/water interface is inefficient and probably resulting from the competition between spacer and chains, which will be most pronounced for the C12 spacer.
Introduction In a previous paper, we examined the effects of partial fluorination of the two side chains of a series of cationic gemini surfactants on their adsorption and layer structure at air/liquid and hydrophobic and hydrophilic solid/aqueous interfaces.1 Fluorinated carbon surfactants are interesting because of their superior surface activity with respect to hydrocarbon surfactants. In partially fluorinated surfactants, two additional factors contribute to the surface activity and to the adsorbed layer structure. These are (i) the unfavorable interaction between fluorocarbon and hydrocarbon fragments, which causes these segments to try to avoid each other, and (ii) the extra flexibility of hydrocarbon fragments over fluorocarbon fragments, which may enable the optimization of the segregation driven by (i). In the gemini cationic surfactants, the situation is further complicated by the presence of both a hydrophobic spacer, which is at least partially constrained to be close to the water, and hydrophobic chains. The competition between these factors was found to lead to significant differences in the packing and conformation of the chain-fluorinated geminis at the three different interfaces. In this article, we extend this work to two surfactants with fluorination only in the *Corresponding author. E-mail:
[email protected]. (1) Li, P. X.; Dong, C. C.; Thomas, R. K.; Wang, Y. L. Langmuir, in press, 2010.
1844 DOI: 10.1021/la104304y
spacer. This is of interest because the spacer should gradually be less immersed in the aqueous subphase (at the air/water interface) as its hydrophobicity increases. So far, the hydrophobicity of the spacer has been changed only by changing its length. Longer spacers can adopt conformations that are not all trans and that may bend up from the aqueous subphase, allowing some hydrophobic interaction between chains and spacer. The purpose of using partially fluorinated spacers is that the hydrophobicity of the spacer is increased and the flexibility of the hydrocarbon fragments is retained, and it may therefore be possible to drive the spacer more strongly into bent conformations. The tendency of the spacer to occupy the more hydrophobic regions of the interface will be quite different at the solid surface; therefore, we have also studied the geminis with partially fluorinated spacers at hydrophobic and hydrophilic solid/aqueous interfaces. To complete the study, we have extended earlier measurements of purely hydrocarbon geminis with different spacers at the air/water surface to the two solid interfaces. This is partially for comparison purposes and partially because it should help to establish the correct values of the limiting areas per molecule of these surfactants at the air/water interface, about which there is some doubt in that adsorption at the hydrophobic solid/aqueous surface is generally similar or somewhat stronger than at the hydrophobic air/water interface. The situation is more complicated at the
Published on Web 12/30/2010
Langmuir 2011, 27(5), 1844–1852
Li et al.
Article
hydrophilic solid/aqueous interface, at which the hydrocarbon geminis have already been studied by Atkin et al.2 using optical interferometry. Optical interferometry does not have sufficient resolution to study the structural arrangement of the layer. The gemini surfactants studied were R,ω-bis[dimethyl-dodecyl) ammonium] alkane dibromides with hydrocarbon spacers of different lengths and two where the alkane spacer was partially fluorinated. The partially fluorinated spacers used were both 12 carbons long, one with a central C4F8 group (C4H8C4F8C4H8) and the other with a central C6F12 group (C3H6C6F12C3H6). The spacers are linked to the side chains by dimethylammonium groups, giving a gemini with two positive quaternary ammonium groups (e.g., (C12H25-N(CH3)2-C3H6C6F12C3H6-N(CH3)2C12H25)Br2). This will be referred to in the abbreviated form, C12-C3(fC4)C3-C12. All of the materials have been previously characterized by surface tension and microcalorimetry.3
Experimental Methods The partially fluorinated gemini surfactants were prepared by a procedure similar to that used for the alkyl chain gemini surfactants and by the synthesis of the partially fluorinated spacers following the method described in the previous paper1 and by Jackson et al.4 The only difference is that the preparation of the spacer required the addition of two unsaturated alcohols to the diiodo fluorocarbon, diiodo-octafluorobutane and diiodo-hexafluorbutane, both obtained from Fluorochem. The purity of the samples was assessed by surface tension measurements (Kruss K10), with the absence of a minimum being the main criterion of purity. Neutron reflectivity measurements were made on the SURF reflectometer at the ISIS neutron source (U.K.)5 Measurements were made using a single detector at a fixed angle of θ = 1.5° using neutron wavelengths in the range of 0.5-6.8 A˚-1 to provide a momentum transfer range of 0.048-0.5 A˚-1, with the momentum transfer Q being defined as Q = 4π sin θ/λ. Reflectivity profiles were fitted using Java programs based on the kinematic approximation6,7 for the air/water interface and the optical matrix method8 for the two solid/liquid interfaces following the procedures described in the previous paper.1 The solid/liquid interface experiments used silicon single crystals with dimensions of 125 50 25 mm3 that were polished on the (111) face. The experimental solution was held in a Teflon container with a volume of approximately 25 mL, which was clamped against the silicon block between temperature-controlled aluminum and magnetic stirring plates. Sample changes were made through inlet and outlet ports located on opposite sides of the container, which could be connected to plastic tubes for injection by syringe. The silicon blocks were cleaned by immersion in piranha solution (a 5:4:1 mixture of H2O/H2SO4/H2O2) at 70 °C and by UV/ozone treatment. They were used for the hydrophilic/ aqueous experiments after rinsing and soaking in clean water (Elgastat Ultrapure). Following this treatment, the silicon (111) surface typically has a layer of oxide of about 15 A˚ in depth. Hydrophobic surfaces were generated from surfaces that had been freshly treated with UV/ozone. The clean silicon blocks were immersed in a solution of 2 10-3 M perdeuterated octadecyl (2) Atkin, R.; Craig, V. S. J.; Wanless, E. J.; Biggs, S. J. Phys. Chem. B 2003, 107, 2978. (3) Li, Y. J.; Li, P. X.; Dong, C. C.; Wang, X. Y.; Wang, Y. L.; Yan, H. K.; Thomas, R. K. Langmuir 2006, 22, 42. (4) Jackson, A. J.; Li, P. X.; Dong, C. C.; Thomas, R. K.; Penfold, J. Langmuir 2009, 25, 3957. (5) Penfold, J.; et al. J. Chem. Soc., Faraday Trans. 1997, 93, 3899. (6) Crowley, T. L.; Lee, E. M.; Simister, E. A.; Thomas, R. K.; Penfold, J.; Rennie, A. R. Colloids Surf. 1991, 52, 85. (7) Crowley, T. L.; Lee, E. M.; Simister, E. A.; Thomas, R. K. Physica B 1991, 173, 143. (8) Born, M.; Wolf, E. Principles of Optics, 7th ed.; Cambridge University Press: New York, 2003.
Langmuir 2011, 27(5), 1844–1852
Table 1. Values of the cmc and the Surface Tension at the cmc of Fluorinated Cationic Surfactants from Microcalorimetry and Surface Tension10 gemini
cmc/mM
γ/mN m-1
C12-C3fC6C3-C12 C12-C4fC4C4-C12 C12-C12-C12 C12-C6-C12 C12-C3-C12
0.16 0.27 0.28 1.03 1.10
35.1 34.8 42.8 36.8 30.8
trichlorosilane (d-OTS) in n-hexadecane (þ99% Sigma, used without further purification) at 20 °C overnight. Upon removal of the blocks, excess d-OTS was removed by wiping the surface with Kodak lens paper soaked in dry dichloromethane, followed by wiping with ethanol to remove the dichloromethane, and the blocks were finally rinsed with UHQ water. Following several rinses in the order dichloromethane, ethanol, and UHQ water until the residual patches of bulk polymerized OTS on the surface were removed, the hydrophobic layer was robust enough to withstand rubbing without introducing scratches or other defects detrimental to neutron reflection experiments. The d-OTS (isotopic purity 92%) was synthesized from perdeuterated octadecyl bromide as described elsewhere.9
Results and Discussion Air/Water Interface. The cmc values at 298.15 K of the compounds studied are listed in Table 1 for reference. (The concentrations selected for the neutron measurements are mainly defined relative to the cmc.) The main purpose of the neutron reflection measurements was to determine the composition and structure of the layers. Measurements of the reflectivity of the partially fluorinated surfactants in null reflecting water (NRW) give the surface coverages of the surfactants. To study the structure of the layers, it is necessary to measure the reflectivity of a number of different isotopic compositions (contrasts), as has been described elsewhere.11 The isotopic compositions used here were the protonated surfactant and that with the two side chains deuterated, each run in NRW and D2O. As in the previous paper, the surfactant layer was modeled using four structural components but the fragments were different because of the different labeling. They were the hydrocarbon chain, the headgroup, and the hydrocarbon part of the spacer taken together, the fluorocarbon part of the spacer, and water, W. The surfactant components were represented by Gaussian distributions, and the water was represented by a distribution that is space-filling up to a cutoff value and then decays as a halfGaussian. The number of independent parameters needed to fit a particular set of data is often then taken to be the thicknesses of the three fragments (σC, σH, and σfS), the area per molecule (A), and the three fragment separations (δC-H, δfS-H, and δW-H), where subscripts C, H and fS refer to chains, heads, and spacers, respectively. The scattering lengths and volumes of the fragments are also required to calculate the reflectivities, and these have been taken from Table 2 of the previous paper.1 In addition to those values, the volumes of C4F8 and C6F12 were taken to be 150 and 225 A˚3, respectively. A Java program in which the parameters could be adjusted interactively was used to fit all contrasts simultaneously for a given surfactant. The best fits to the neutron reflectivity profiles using this program are shown for (9) Fragneto, G.; Li, Z. X.; Thomas, R. K.; Rennie, A. R.; Penfold, J. J. Colloid Interface Sci. 1996, 178, 531. (10) Li, Y. J.; Li, P. X.; Wang, J. B.; Wang, Y. L.; Yan, H. K.; Dong, C. C.; Thomas, R. K. J. Colloid Interface Sci. 2005, 287, 333. (11) Lu, J. R.; Lee, E. M.; Thomas, R. K. Acta Crystallogr., A 1996, 52, 11.
DOI: 10.1021/la104304y
1845
Article
Li et al.
Table 2. Structure of Layers of Hydrocarbon Geminis and Gemini Surfactants with Partially Fluorinated Spacers at the Air/Water Interface surfactant
c/mM
σ/(2 A˚
δC-H/(1 A˚
δfS-H/(1 A˚
δW-H/( 1 A˚
A/( 4 /A˚2
C12-C3fC6C3-C12 C12-C4fC4C4-C12 C12-C12-C12 C12-C10-C12 C12-C6-C12 C12-C3-C12
0.35 0.27 0.28 0.63 1.03 1.1
16.25 15.25 15.25 15.5 15.25 17.75
1.25 3.0 -4.25 -5.0 -4.75 -4.0
-5.5 -1.75 -2 -2.75 -0.5 0.5
-6.3 -8.1 -1.2 -1.9 -2.0 -2.6
137 106 139 115 104 78
Figure 1. Neutron reflectivity of C12-C3fC6C3-C12 for four isotopic compositions at a concentration of 0.35 mM. The continuous lines are calculated for the best fits using the parameters in Table 2.
the C12-C3fC6C3-C12 gemini in Figure 1, and the fitted parameters for the two fluorocarbon geminis are given in Table 2. The accuracy of the fitting parameters has to be judged partially by trial and error and partially by experience. In all cases, the coverage is by far the most accurately determined quantity because it remains invariant whatever the structural parameters and is independent of the choice of the fragment volumes. The thicknesses of the individual fragments are, however, not well determined. The thickness of the fragments in a surfactant layer at the air/water interface is often dominated by a large roughness, in part caused by capillary waves.11 Given this and that the labeling scheme used for the present series of partially fluorinated surfactants used here is less extensive than for those used in the previous paper, we made the assumption that the thickness of all three surfactant fragments is the same (i.e., only one thickness parameter is needed to describe the layer). The fitting is now sensitive to this value, and this is reflected by the small value of the error given in the Table. However, this has been achieved at the expense of losing less accurate information about the detailed structure. The best quality structural information is the separation between fragment distributions. For any pair of fragments whose mutual contrast can be varied, the reflectivity will be very sensitive to their separation. For the labeling used here, the two surfactant fragment separations given in Table 2 are well determined. The reflectivity is also very sensitive to the separation of the surfactant fragments from water but the calculated values depend on the volumes assumed for the fragments. This is because the fragments displace water in the interfacial region, and this affects the reflectivity. The values are not easy to estimate and typically have an uncertainty of 10%, which introduces a small uncertainty into the final parameters. Because the main purpose of this work is to examine the effects of variable hydrophobicity of the spacer on the adsorption properties, it is useful to make the comparison with the simpler geminis studied earlier by Li et al.22 In that paper, they were modeled with a small contribution from a sublayer, a more 1846 DOI: 10.1021/la104304y
extensive parametrization of the fragments, and a less realistic model of the water distribution. However, we did not find it necessary to include any contribution from a sublayer for any of the fluorinated compounds. This and the need to make a direct comparison between the two series of compounds made us reanalyze the earlier data on the hydrocarbons using the model described above, and the results from this analysis are also included in Table 2. Apart from the main difference from the earlier analysis, which is that there is no sublayer, the basic monolayer structure is not substantially different from the earlier analysis. The limiting area per molecule for the gemini with the C3fC6C3 spacer of 137 A˚2 is similar to that of 139 A˚2 for the hydrocarbon gemini with a C12 spacer (139 A˚2), but that for the C4fC4C4 spacer is much less at 106 A˚2. Diamant and Andelman12 have produced a model that attempts to account for the variation of the limiting areas of geminis in terms of the varying interaction of the spacer with the aqueous subphase as it increases in length. The longer the spacer, the more it will tend to try to position itself in the same region of the interface as the chains. In principle, this could lead to a maximum in the area per molecule as the spacer length increases, although none was observed for the hydrocarbon series up to a spacer with 12 carbons. The partially fluorinated spacer is more hydrophobic and is therefore more likely to protrude upward into the hydrophobic region and hence bring about a reduction in the limiting area. That the difference between the two compounds is so large indicates that the flexibility of the hydrocarbon attachments to the headgroups exerts a controlling influence on the structure. The latter was shown to be very important in allowing the chain-fluorinated geminis to pack optimally at the surface.1 The C3F6 fragments at the two ends of the spacer may be too short to allow the fluorocarbon fragment of the spacer and the hydrocarbon side chains to pack comfortably to form a coherent hydrophobic region. The positions of the fragments in the layer are shown for the two partially fluorinated spacers and the simple hydrocarbon C12 spacer in Figure 2, and the large effect of partial fluorination on the position of the spacer is clear. For both fluorinated surfactants, the fluorocarbon fragment forms the outside part of the layer. For the C12 gemini, the side chains form the outer part of the layer, although as shown earlier by Li et al.22 the C12 spacer lies further out from the water than any of the shorter hydrocarbon spacers (as can also be seen in Table 2). The C4F8 fluorocarbon fragment is further from the water than the C6F12 fragment and so appears to be exerting its hydrophobicity more completely. That the limiting area per molecule is much greater for the C6F12 gemini indicates that the positioning of the C6F12 fragment on the outside is at the expense of satisfactory packing of the side chains. The value of the limiting surface tension is also higher for the C6F12 gemini, confirming that the packing of the hydrophobic fragments is less efficient in this compound. Adsorption at the Hydrophobic Solid/Aqueous Interface. The d-OTS surface was first characterized by measuring reflectivity (12) Diamant, H.; Andelman, D. Langmuir 1994, 10, 2910.
Langmuir 2011, 27(5), 1844–1852
Li et al.
Figure 2. Fragment profiles for C12-C12-C12, C12-C4fC4C4-C12, and C12-C3fC6C3-C12 at the air/water interface at their cmc’s.
profiles in three different water contrasts: D2O and water with scattering-length densities of 3.4 10-6 and 4.0 10-6 A˚-2. As always occurs with these layers in contact with water, there is a very thin layer that some attribute to contamination and others attribute to an “air” layer.13-1516 This disappears in the presence of a surface-active adsorbate. It is therefore necessary to include this layer in the basic characterization but not when the adsorbed surfactant layer is present. Deuterated OTS has a very similar scattering-length density to D2O. Adsorption of a layer containing hydrocarbon therefore gives a very strong signal. If the partially fluorinated surfactants were to adsorb with the strongly hydrophobic fluorocarbon fragment attached to the strongly hydrophobic OTS surface, then the initial part of the surfactant layer would contrast only weakly with the OTS. The main contribution to the reflectivity would come from the hydrocarbon part of the layer and the protonated TAB group. The reflectivity profile is therefore quite sensitive to the relative distribution of these two groups within the layer. The adsorbed layer was taken to consist of a hydrophobic region next to OTS and a hydrophilic region on the outside of the layer. The parameters of the silica layer and OTS were constrained to the values obtained in the absence of surfactant. The data were fitted using the optical matrix method with the usual constraints on the total volume fraction at any point in the layer. To allow for disorder in the layer, the fractions of heads and spacers in the chain region were used as additional fitting parameters. The best fits of the parameters are given in Table 3. Fits of the model to two sets of data, C12-C4fC4C4-C12 and C12-C3-C12, are shown in Figure 3. These were chosen because the former has one of the lowest coverages and the latter has the highest coverage of the geminis under consideration. (13) Poynor, A.; Hong, L.; Robinson, I. K.; Granick, S.; Zhang, Z.; Fenter, P. A. Phys. Rev. Lett. 2006, 97, 266101. (14) Mezger, M.; Reichert, H.; Schoder, S.; Okasinski, J.; Schrode, H.; Dosch, H.; Palms, D.; Ralston, J.; Honkimaki, J. Proc. Natl. Acad. Sci. U.S.A. 2006, 101, 18401. (15) Ocko, B. M.; Dhinojwala, A.; Daillant, J. Phys. Rev. Lett. 2008, 101, 039601. (16) Young-Soo, S.; Satija, S. K. Langmuir 2006, 22, 7113.
Langmuir 2011, 27(5), 1844–1852
Article
In general, it might be expected that the adsorption of an amphiphile on a hydrophobic surface of this type should be somewhat stronger than at the hydrophobic air-water interface. Thus, anionic surfactants AOT9 and SDS17 are adsorbed about 10% more on OTS than at the air/water interface (at or slightly above the cmc) whereas for nonionic C12E418 and zwitterionic dodecyl betaine17 there is a negligible difference between the adsorption on the two surfaces. The purely hydrocarbon geminis are all adsorbed much more strongly, about 20%, on the OTS surface than at the air/water interface. Thus, the areas per molecule for the hydrocarbon gemins are 66, 84, 106, and 110 A˚2 for C3, C6, C10, and C12, respectively, at the OTS surface, to be compared with 78, 104, 115, and 139 A˚2, respectively, for the air/ water interface. The area per molecule also drops about 20% for the C4fC4C4 spacer. It would then be tempting to attribute this change to charge repulsion at the air/water interface and the higher charge on the geminis. However, we will argue in the conclusions below that the change is probably associated with the high level of disorder in the layers at the air/water interface, which gives more freedom for a change in area under more hydrophobic conditions. Though the decrease in area for the hydrocarbon geminis is large, the change for C3fC6C3, 137 to 77 A˚2 is so large as to suggest that there is a fundamental change in structure, which is supported by the fragment profiles in Figure 4. First we note that, although in the previous paper on the partially fluorinated chain geminis the positions of the heads and spacer were not well resolved, here the difference in the distribution of contrast within the molecules and the level of hydration of the headgroup regions allows them to be differentiated. The most close-packed layer at both air/water and OTS/water interfaces is that of C12-C3-C12, the gemini with the shortest spacer. The fragment distribution shows that the two chains are next to the OTS and the spacer and headgroups are on the aqueous side of the interface, as would be expected. The spacer is too small to move into the hydrophobic region of the layer. However, as the spacer length increases to 12 carbon atoms, the spacer becomes long enough to penetrate progressively more into the hydrophobic region, much more so than at the air/water interface (Figure 2). The limiting area per molecule therefore decreases substantially, and this is in accord with the model of Diamant and Andelman for the air/water interface. As judged from what happens at the air/water interface, where the fluorocarbon is on the nonaqueous side of the interface, one might expect an intensification of this effect for the fluorocarbon spacers. However, it is clear from the profiles that the repulsion between fluorocarbon and hydrocarbon completely reverses the distribution of the fragments from that at the air/ water interface. At the OTS surface, the headgroups and spacer move away from the hydrophobic interface and point inward toward the water. Thus, the large limiting area at the air/water interface could be attributed to the competition of hydrocarbon chains and fluorocarbon spacer to form the hydrophobic outer part of the layer, but at the OTS interface, only the hydrocarbon chains form this part of the layer and the fluorocarbon spacer is forced out toward the aqueous region. This evidently allows a much better steric arrangement of the spacer and closer packing of the chains in the layer. C12-C4fC4C4-C12 has an intermediate structure, and its area on the OTS surface is lowered by a proportion similar to that of conventional hydrocarbon geminis. The discussion in the previous paragraph is based on the area per molecule, A, in Table 3. When the OTS layer is deuterated, the (17) Hines, J. D.; Fragneto, G.; Thomas, R. K.; Garrett, P.; Rennie, G. K.; Rennie, A. R. J. Colloid Interface Sci. 1997, 189, 259. (18) Fragneto, G.; Lu, J. R.; McDermott, D. C.; Thomas, R. K.; Rennie, A. R.; Gallagher, P. D.; Satija, S. K. Langmuir 1996, 12, 477.
DOI: 10.1021/la104304y
1847
Article
Li et al.
Table 3. Adsorption of Layers of Hydrocarbon Geminis and Gemini Surfactants with Partially Fluorinated Spacers at the Hydrophobic (OTS)/Water Interface surfactant
c/mM
τ1/A˚ (1
τ2/A˚ (1
θ ( 0.03
σ/A˚ ( 0.5
A/A˚2 ( 5%
Γ/mg m-2 ( 0.1
C12-C3fC6C3-C12 C12-C4fC4C4-C12 C12-C12-C12 C12-C10-C12 C12-C6-C12 C12-C3-C12
0.10 0.27 0.28 0.63 1.06 1.1
10.5 10.5 9.5 9.5 9.5 10.5
1.0 2.5 2.0 2.5 1.0 1.5
0.83 0.76 0.67 0.70 0.88 0.99
6.0 6.0 6.0 7.0 7.0 7.0
81 89 110 106 84 66
2.0 1.7 1.1 1.2 1.3 1.6
Figure 3. Neutron reflectivity profiles for C12-C4fC4C4-C12 and C12-C3-C12 adsorbed on deuterated OTS on silica in D2O (points) and the fitted curves using the parameters and concentrations given in Table 3 (continuous lines).
solvent is D2O, the surfactant contains a high proportion of hydrogen atoms, and A is a well-determined quantity. However, it is the average area per molecule over the whole surface, and this may not reflect the true area per molecule because OTS does not cover the surface completely and the surfactant might be expected to adsorb only on the OTS-covered parts of the surface. The true value of the area per molecule would then be obtained by multiplying the measured average A by the fractional coverage of OTS (i.e., a smaller value and an even larger change from the value at the air/water interface). However, the situation is not necessarily this simple. As has been noted in several previous publications, the fitted parameters for the OTS layer on its own show that the OTS layer is defective in some way. Three situations may then arise, and these are illustrated in Figure 5. If there are a large number of very small defects (Figure 5a), then the surfactant might be able to form a coherent layer over all of the surface. If there are a small number of large defects in the OTS layer, then the surfactant could either adsorb onto only the OTS-covered fraction of the surface (Figure 5c) or could adsorb in the holes in the OTS to form a complete monolayer over the resulting composite layer (Figure 5b). Although at this composition the reflectivity is not at all sensitive to the amount of D2O in the layer because it matches OTS, it is very sensitive to penetration of the lowscattering surfactant into the OTS layer, and it is not possible to obtain satisfactory fits for the structure shown in Figure 5b. The choice between the remaining (a) and (c) would normally be impossible on the basis of the reflectivity alone. For example, the coverage of C12-C10-C12 is 0.70, which could be either the coverage of OTS islands at a level of 0.82 and no surfactant over the OTS holes or a uniform layer containing an average water fraction of 0.3 in the chain region. However, the situation for C12-C3-C12 is special because now the observed average coverage is 0.99 (i.e., there is more or less no water at the center of the surfactant layer). This can correspond only to the situation shown 1848 DOI: 10.1021/la104304y
Figure 4. Fragment profiles for six gemini surfactants adsorbed at the aqueous OTS surface: C12-C3-C12, C12-C12-C12, C12C4fC4C4-C12, and C12-C3fC6C3-C12.
in (a). Even after allowing for some uncertainty in the assumed volumes of the various fragments, there probably cannot be more than about 5% water in the center of the chain region. This indicates that the defective structure of OTS results from small pin-prick defects, that the surfactant layer is laterally homogeneous, and that the average area A is the correct value to take for the area per molecule, not the lower value that would result from using the model shown in Figure 5c. The thicknesses of all of the layers are substantially less than at the air/water interface, confirming that capillary waves contribute significantly to the latter. They are also significantly less than the value of 14.7 A˚ for a fully extended C12 chain. This would suggest an average tilt away from the surface normal of about 52° (based on the chain thickness only) for C12-C12-C12 to 38° for C12-C3-C12 and C12-C4fC4C4-C12. For the hydrocarbon geminis, the tilt away from the normal decreases with the spacer length (i.e., decreases as the chains pack more tightly). Silica (Hydrophilic)/Water Interface. The oxide layer on the Si(111) substrate was first characterized using the neutron reflectivity profiles from Si/SiO2/D2O, Si/SiO2/water2.07, and Si/ SiO2/water4.00 interfaces, where the subscripts indicate the scattering-length densities of water. Three silicon blocks were used for Langmuir 2011, 27(5), 1844–1852
Li et al.
Figure 5. Possible structures for a surfactant layer adsorbed on a defective OTS surface: (a) a bridging surfactant layer over small defects, (b) a bridging surfactant layer over large defects filled with surfactant, and (c) surfactant only on OTS patches.
the measurements. The profiles were almost exactly the same as those obtained by Fragneto et al.9 and could be well fitted by a single uniform layer of silica of thickness 14 ( 1 A˚, scatteringlength density 3.4 10-6 A˚-2, and roughness 3 ( 0.5 A˚. The adsorption experiments were done at a single contrast (protonated surfactant in D2O) at a single concentration above the cmc. As with other surfactants on the hydrophilic surface, the use of a single monolayer to fit the data shows that the thicknesses (and also usually the coverages) of all of the adsorbed surfactants studied are too large to be considered a monolayer. The initial fitting of the data to a simple monolayer confirmed that the coverage is too high to be a monolayer and is therefore probably some kind of bilayer or fragmented bilayer structure. We therefore fitted a layer consisting of three sublayers, where the central layer consists mainly of the hydrophobic chains and the outer two layers consist of heads and spacers or just heads in the case of the single-chain surfactants. Because the fractional coverage of this bilayer is significantly less than unity, we further assumed that it occurs in patches. The average scattering-length density is then an average of the volume fractions of bilayer and water.19 The coverage (θ in Table 4) is defined as the volume fraction of surfactant at the center of the bilayer. The calculation of the reflectivity was based on the optical matrix method, and the main adjustable parameters are the thicknesses of the three layers and the coverage of the bilayer. The remaining parameters merely show that the layer is otherwise fairly disordered. Thus, portions of chains can be moved into the headgroup regions and vice versa while maintaining the correct stoichiometry and some asymmetry between the headgroup distributions can similarly be included, but these parameters are coupled in the fitting and should not be seen as very significant. The scattering properties of fragments of the layer used to fit the reflectivity profiles have already been listed in Li et al.1 As always with neutron reflectometry, the fits are very sensitive to the coverage, as illustrated by the three profiles in (19) McDermott, D. C.; McCarney, J.; Thomas, R.; Rennie, A. R. J. Colloid Interface Sci. 1994, 162, 304.
Langmuir 2011, 27(5), 1844–1852
Article
Figure 5 in which the coverage varies from 0.73 (C12-C3-C12) to 0.68 (C12-C12-C12) to 0.66 (C12-C3fC6C3-C12). For these thicker layers, the reflectivities are also sensitive to the overall thickness. However, the fits are less sensitive to the division into chain and head/spacer regions because the water fraction in the outer two layers is not large enough and the spacer and headgroup cannot be distinguished at all. The parameters of the silica layer were fixed as determined for the bare silica surface. The fits to the observed profiles for the two compounds with the partially fluorinated spacers are shown in Figure 5, and the fitted parameters are given in Table 4. It is now well established that surfactants adsorb onto hydrophilic surfaces in the form of aggregates, typically micellelike, flattened micelles, or rods. These have hydrophobic interiors from which water is expected to be largely expelled. Under these circumstances, 1 - θ is the fraction of the surface covered only by water in the center of the bilayer and therefore gives the fraction of the area of the surface covered. The amount of surfactant adsorbed depends partially on 1 - θ, but it also depends on the thickness of the layer, which is well determined. The true area per molecule in the aggregates, A*, which is the value that should be compared to the area per molecule at other interfaces, is then 2 A θ. The values of A* obtained for the series of hydrocarbon surfactants are 70, 85, 94, and 98 A˚2 in order of increasing spacer length, compared to 66, 84, 106, and 110 A˚2 on OTS, both now very much less than for the air/water interface, 78, 104, 115, and 139 A˚2. We comment further on these values in the Comparison and Conclusions section. The values of the adsorbed amount previously determined by Atkin et al.2 using optical reflectometry are given in parentheses in Table 4 and are all much lower than our values. However, their surface was prepared in a different way from ours. Ours was a thin native oxide layer present on a Si(111) surface and normally had a thickness of 10-20 A˚. Atkin et al. did not specify which silicon surface they were using, and their layer was prepared by oxidation at high temperature to give a layer thickness of 165 A˚. As far as we are aware, differences in adsorption between different silica layers on different silicon surfaces have not been explored. However, Fragneto et al.21 observed a 10% variation in the adsorption of hexadecyltrimethylammonium bromide that was due to differences in roughness in the oxide layer on Si(111), and it is plausible to expect differences between differently prepared silica surfaces and different silicon surfaces when the interaction with the surface is determined by a low concentration of surface charges and is fairly weak. The values of the area per molecule obtained by Atkin et al. are actually less than the coverage required for a monolayer. For example, they obtained an area per molecule of 195 A˚2 for C12-C12-C12 whereas the directly measured (neutron reflectometry) value is 139 A˚2. Given this low coverage and the fact that optical reflectometry does not measure the thickness of the layer, Atkin et al. could not technically distinguish between monolayer- and bilayer-type adsorption based solely on their optical reflectometry data. However, they also studied some of the systems by AFM and force balance measurements. They obtained a thickness of 35 A˚ for C12-C2-C12 and C12-C3-C12, which is slightly larger than the value of 28.5 A˚ measured here. Chorro et al.20 used an analytical method to study the adsorption of geminis on silica particles and obtained values lower than both ours and those of Atkin et al. (20) Chorro, C.; Chorro, M.; Dolladille, O.; Partyka, S.; Zana, R. J. Colloid Interface Sci. 1998, 199, 169. (21) Fragneto, G.; Thomas, R. K.; Rennie, A. R.; Penfold, J. Langmuir 1996, 12, 6036.
DOI: 10.1021/la104304y
1849
Article
Li et al.
Table 4. Adsorption of Layers of Hydrocarbon Geminis and Gemini Surfactants with Partially Fluorinated Spacers at the Silica/Water Interface surfactant
c/mM
σ1/A˚ ( 0.5
σ2/A˚ ( 1
σ3/A˚ ( 0.5
θ ( 0.03
0.10 8.0 15 3.5 C12-C3fC6C3-C12 0.27 8.0 19 3.5 C12-C4fC4C4-C12 4.0 6.5 14 4.5 C12-C12-C12 4.0 5.0 15 4.5 C12-C10-C12 4.0 4.5 17 4.5 C12-C6-C12 4.0 5.0 20 3.5 C12-C3-C12 a Surface excess values from Atkin et al.2 b Surface excess values from Chorro et al.20
0.66 0.62 0.68 0.66 0.65 0.73
A/A˚2 ( 4
Γ/mg m-2 ( 0.1
A*/A˚2 ( 4
70 58 72 71 65 48
2.3 2.6 1.7 (0.6) 1.7 (0.7,a 0.4b) 1.7 (0.9,a 0.7b) 2.2 (1.3)
92 72 98 94 85 70
Figure 6. Neutron reflectivity at the silica/water interface for C12C3-C12 at 4 mM, C12-C12-C12 at 4 mM, and C12-C3fC6C3-C12 at 0.11 mM.
The values of the overall thickness of the bilayer of all six compounds studied are relatively small. For example, for the partially fluorinated chain geminis studied in the previous paper the thinnest bilayer was found to be 32.5 A˚ whereas the thickest layer here is C12C4fC4-C4-C12 with an overall thickness of 31.5 A˚. Nevertheless, in all cases these values are approximately double the values obtained for the monolayer on OTS. It is then not easy to distinguish between a bilayer structure where the chains are interdigitated and less tilted and a structure where they remain strongly tilted as in the monolayer but are packed end-to-end in the bilayer. The fragment distributions, shown in Figure 6, show more clearly the variation in thickness of the bilayer for two hydrocarbon and two fluorocarbon geminis.
Comparison and Conclusions The values of adsorbed area per molecule obtained for the series of hydrocarbon geminis are shown in Figure 7. (The values for the hydrophilic surface are A* in Table 4.) They all increase approximately linearly with the number of carbon atoms in the spacer. The values at the two solid surfaces are on average more than 20% less than at the air/water interface. This could be because the interaction of the chains of the gemini with other hydrocarbon chains, either by self-interaction in an aggregate at the hydrophilic surface or with the hydrocarbon chain forming the surface (OTS), leads to more efficient packing than at the air/water interface. This would suggest that the packing at the air/water interface is inefficient, and this may result from the competition between spacer and chains, which will be most pronounced for the C12 spacer. The corresponding values for the two fluorocarbon geminis, C12-C3fC6C3-C12 and C12C4fC4C4-C12, are 92 and 72 at the hydrophilic silica surface, 77 and 89 at OTS, and 137 and 106 A˚2 at the air/water interface with decreases of 38 and 24% from air/water to the average solid value, respectively. The changes (in percent) are comparable with those for the C12 hydrocarbon spacer, again suggesting that the extra 1850 DOI: 10.1021/la104304y
Figure 7. Fragment profiles for four gemini surfactants adsorbed at the aqueous silica surface: C12-C3-C12, C12-C12-C12, C12-C4fC4C4-C12, and C12-C3fC6C3-C12.
hydrophobicity of the chain environment at the two solid surfaces promotes much more efficient packing. In the case of the partially fluorinated spacers, the reversal of the distribution of the fragments of the fluorocarbon geminis between the air/water and the OTS/water interfaces shows that there can be substantial structural changes between the different interfaces. An alternative explanation that cannot be eliminated is that the high electrostatic charge on the gemini is less well screened at the air/water interface and that it is electrostatic repulsion that strongly influences the area per molecule in this case. The coverages (θ) of both geminis with the fluorocarbon in the spacer and the pure hydrocarbon geminis on the hydrophilic silica are mostly significantly lower than for the partially fluorinated chain geminis, except for the two most heavily fluorinated. However, the spacer for the partially fluorinated chain compounds was C6 as compared to C12 for the partially fluorinated spacer, so one should not make too much of this comparison. However, the comparison of the pure hydrocarbon gemini with C6 is a valid and interesting one. All of the partially fluorinated compounds adsorb more strongly than does C12C6C12. Davey et al. found that fluorocarbon surfactants in competition with Langmuir 2011, 27(5), 1844–1852
Li et al.
hydrocarbon surfactants adsorbed more strongly on mica in mixtures with hydrocarbon compounds. They attributed this to a stronger dispersion interaction of the fluorocarbon with mica. At first sight, this does not seem to fit the gemini data where the more-fluoroinated chain geminis actually adsorb less than the less-fluorinated compounds. However, comparison with the singlechain partially fluorinated compounds suggests that the conformational problems associated with the geminis may partially obscure any dispersion force effects. For the hydrocarbon geminis, there is disagreement between the values of the limiting coverage as determined by the application of the Gibbs equation to surface tension measurements and as determined by the more direct neutron reflectometry method. Using a prefactor of 3 in the Gibbs equation results in the limiting area per molecule coverage being 30-40% larger than the neutron value for all of the geminis studied by both methods except for one with a xylyl spacer. This was attributed without proof to premicellar association and/or counterion association.22 The explanation is still not clear, although Zana has claimed that conductivity measurements show that neither type of association accounts for the discrepancy in the C12-Cm-C12 series of cationic geminis.23 An alternative way of examining the problem is to compare the limiting coverages at different interfaces. The surface tension method with a prefactor of 3 (the nominally correct value) gives values of 98 (105), 143 (143), and 224 (226) A˚2 for the C12 geminis with C3, C6, and C12 spacers, respectively, where the values are from Li et al.22 and Alami et al.24 (values in brackets), whereas the neutron values are 78, 104, and 139 ( 5% A˚2, respectively. The effect is therefore large and leads to conflicting conclusions about the surface packing. The neutron measurements at the two solid surfaces indicate an even higher coverage than at the air/water interface. Thus, the corresponding values for the OTS surface and the silica surface are 63, 105, and 112 A˚2 for the hydrophilic surface and 70, 94, and 98 A˚2 for OTS. If the surface tension value for the area per molecule were correct, then the decrease on going to either of the two solid surface would be about 50%, which would be quite extraordinary. Errors in the surface tension method originate entirely from the choice of the Gibbs prefactor, which in turn is based on the extent of either or both ion association or premicellar association, neither of which has proven at all easy to elucidate. Here, we focus on the possible errors in the neutron measurements, which are quite different among the three interfaces. To determine the surface excess of a surfactant at the air/water interface, neutron reflectometry measures the reflectivity from the partially deuterated or perdeuterated surfactant in null reflecting water (NRW). The measured signal comes only from the adsorbed layer, and it can be measured to an accuracy of better than 5%. The derivation in coverage normally involves no assumptions, and what is measured is the total adsorbed amount of the surfactant ion. In dilute solution, this is indistinguishable from the surface excess. Apart from the measurement error above, there may be differences from the hydrogen form of the surfactant because of isotope effects, but these have been found to be small for surfactants and usually within the measurement error (about 5%). The effects of impurities on the measurements depend on the system. The gemini surfactants are relatively easy to prepare at a high level of surface purity. However, a common source of discrepancy between neutron reflectometry and surface tension in other systems has been the presence of ionic impurities.25 In all (22) (23) (24) (25)
Li, Z. X.; Dong, C. C.; Thomas, R. K. Langmuir 1999, 15, 4392. Zana, R. J. Colloid Interface Sci. 2002, 248, 203. Alami, E.; Beinert, G.; Marie, P.; Zana, R. Langmuir 1993, 9, 1465. An, S. W.; Lu, J. R.; Thomas, R. K.; Penfold, J. Langmuir 1996, 12, 2446.
Langmuir 2011, 27(5), 1844–1852
Article
Figure 8. Limiting areas per molecule for hydrocarbon geminis of varying spacer length at three different interfaces. The area used for the silica/water interface is the reduced area A* calculated in Table 4.
cases studied, this has been due to the presence of divalent cations in nominally monovalent anionic surfactant systems. In these cases, the effect on the neutron measurement is small because the only effect of introducing a divalent ion into a saturated surfactant layer is generally to increase the packing by a small amount. In assessing the possibility of such an error for the gemini systems, it is important to note that no case of this type has yet been observed in a cationic surfactant and it is difficult to determine which negatively charged divalent ion could be a significant source of impurity. Also, the effect on the area per molecule measured by neutron reflectometry would be much smaller than the present discrepancies. One final possible error is that the layer is assumed to be detected as uniform by the neutrons. If the layer were to break up into islands on a length scale larger than the coherence length of the neutrons, then the reflected signal would be the average of the signal from the islands rather than a signal from the average layer. However, no one has yet reliably identified the former situation except in insoluble monolayers,26 and the latter situation has been shown to be correct for many surfactant systems. At the OTS interface, it is generally the protonated surfactant that is studied, and this reduces any isotope effects. The thickness of the composite OTS and surfactant layer is much larger than that of the surfactant at the air/water interface, which reduces the error in the determination of coverage to below 5%. Furthermore, the fractional coverage is close to unity, and there can be no island effect. However, the presence of highly surfaceactive impurities might have a larger effect than at the air/water interface. Measurements at the silica/water interface are again made on the hydrogenated surfactant. The sensitivity to coverage and thickness is again significantly higher than at the air/water interface. Adsorption is known to occur in patches, but all of the evidence (mainly from AFM) suggests patch sizes that are smaller than the neutron coherence length; therefore, the neutrons see the average refractive index of the layer. Thus, this method should also give an even more reliable value of the area per molecule than that obtained for the air/water interface. However, there is another crucial factor that concerns the sensitivity to impurities at this interface. Highly surface-active impurities are hydrophobic, and that is why they adsorb at the air/water and OTS surfaces. The typical impurity of this type will not be at all surface-active at the hydrophilic silica surface. It will tend to be solubilized into micelles. Although this might have indirect (26) Richardson, R. M.; Roser, S. J. Langmuir 1991, 7, 1458.
DOI: 10.1021/la104304y
1851
Article
consequences for adsorption at the silica surface, it is very unlikely that these will parallel those at air/water and OTS/water interfaces. Thus, the effect of impurities would tend to create inconsistencies between the coverages obtained for the two solid interfaces and that from the air/water interface. The consistency among the three independent measurements of the
1852 DOI: 10.1021/la104304y
Li et al.
area per molecule strongly suggests that the neutron values are correct. Acknowledgment. P.X.L. thanks the Clarendon Fund of the University of Oxford for a scholarship. We also thank ISIS for their considerable support in the form of neutron beam time.
Langmuir 2011, 27(5), 1844–1852
