Langmuir 1992,8,2206-2209
2206
Fourier Transform Infrared Spectroscopy and Monte Carlo Studies on the Dynamics of Physisorbing and Chemisorbing Polymers Terence Cosgrove,*vtClive A. Prestidge,* Stephen M. King,g and Brian Vincent? School of Chemistry, University of Bristol, Cantock's Close, Bristol BS8 lTS,U.K.,School of Chemical Technology, University of South Australia, Adelaide, South Australia 6093, Australia, and Neutron Science Division, Rutherford Appleton Laboratory, Didcot OX11 OQX,U.K. Received February 14,1992.I n Final Form: June 22,1992 Dynamic studies on the adsorption of poly(methybiloxane) onto alumina have been carried out 88 a function of the reaction temperature. The relative rates of chemisorption and phyaisorptionare markedly temperature dependent, and it is the balance of these ratee that determinesthe finalsurface d o r m a t i o n of the polymer. The experimental resulta are compared qualitatively with a dynamical Monte Carlo simulation.
Introduction Studies on the dynamics of polymer adsorption1p2have been made using a variety of experimental methods, including nuclear magnetic resonance spectroscopy,3ellipsometry," stagnant flow? and F'I'IFL6 These studies have focused on the determination of the structure under nonequilibrium conditions by essentially sampling the state of the system through the adsorbed amount, the thickness of the adsorbed layer, or the bound fraction. In this paper we report a novel approach to studying the dynamics of polymer adsorption, whereby the surface conformations of the polymers are "locked" through the progressive chemisorptionof the adsorbing polymer.6 The rate of physisorption, Rp, can be defined as the net rate at which reversible segment-surface bonds are formed. In a similar way, for chemisorption, Rc is the rate at which irreversible segment-surface bonds are formed. If Rc Rp, then irreversible surface chemical bonds are formed quickly before any substantial conformationalrelaxation can occur. In this situation the conformation of a chain in the adsorbed state will be rather similar to that in solution immediately preceding adsorption. In thia case the flux of chains to the intarface will not dramatically effect the final surface conformation, except in making access to parte of the surface sterically difficult. Q 1992 American
Chemical Society
Dynamics of Physisorbing and Chemisorbing Polymers
In this work, we have takenadvantage of the observation that the rate of chemisorption is temperature dependent to “lock” the adsorbing polymer at different stages in the progression toward full surface relaxation. The basic experimental system we have chosen for this study is the same as that in the previous study where it was discussed in some detail.6 The mechanisms of the physical and chemical adsorption of the polymer as described earlier6 are thought to proceed through interactions between the surface hydroxyl groups of the alumina and the silane groups of the polymer. A detailed FTIR study of this process9 has shown that there are corresponding changes in the intensity and position of the Si-H and methyl deformation absorption bands. The role of any surface bound water which may be present in the chemisorption process is not clear, though it has been shown that the results of the study are highly reproducible and that the polymers are irreversibly adsorbed. The chemisorbed bound fraction p c has been calculated from the reduction in intensity of the Si-H band, and the physisorbed bound fraction, pp, was calculated from the intensity of the shifted band. In order to model the experimental system, we introduce a dynamical Monte Carlo computer model which mimics the various stages involved in the relaxation and chemisorption process. Experimental Section The poly(methylsi1ane) [PMS] sample preparation has been described in detail elsewhere.1° This particular sample had a number average molecular weight, MN,of 851 and a polydispersity index of 1.02. Although this corresponds to only 14 monomers, the extreme flexibility of the Si-&Si bond confers a polymeric character to the molecule. The tetrachloromethane was AnalaR grade from B.D.H., Ltd., and after distillation was stored over phosphorus pentaoxide. The water contentwas found to be less than 15 ppm as determined by an automated KarlFisher titration using a Metrohm Model 652 coulometer. The y-alumina (PREST, Grenoble, France) had a BET surface area as determined by nitrogen gas adsorption, but an of 200 m2g’, effective area determined by measuring the adsorption of low molecular weight analogues of PMS of 21.6 m2 g’. Electron microscopy showed that the sample in fact consisted of small aggregates of -9 nm diameter primary particles. However, by choosing a consistent method for preparing the dispersions, reproducible results for adsorbed amounts could be obtained.6 It is therefore apparent that the preparation procedure did not lead to any dramatic increase in surface area and hence no substantial breakdown of the aggregates. The FTIR data were obtained using a Perkin-Elmer SP1750 spectrophotometer fitted with a fast DTGS detector and a 1 mm path length solution cell with KBr windows. Spectra were recorded with a solution of 4 cm-l. Values of the chemisorbed and physisorbed bound fractions, pc and pp, and the adsorbed amount were obtained by integration of the FTIR spectra after calibration using solutions of the polymer in CCL.
Monte Carlo Simulation In this section we briefly describe the principal features of the computer model. A detailed description is in preparation” and will be published in due course. We have investigated both the static and dynamic behavior of polymer chains of up to 99 segments generated by Nstep, self-avoiding, random walks (SAW’S) on a simple cubic lattice. Each segment is assumed to occupy one lattice site and all lattice sites not occupied by segments of a chain are deemed to contain a solvent molecule having the same volume as a segment. (10) Richards, R. C. D. Ph.D. Thesis (York), 1988. (11) Cosgrove, T.; King, S. M. Submitted to Macromolecules.
Langmuir, Vol. 8, No. 9,1992 2207
Except where indicated, the results depicted in this paper are ensemble averages over some lo4 individual chains. The algorithm used to generate each chain checks every new step in a SAW to see if it will contravene the principle of excluded volume. If it does, then that step is rejected and an alternative placement of that step is chosen at random. The entire walk is only rejected if no suitable alternative placement can be found within a reasonable number of attempts. This means that the probability of generating a chain conformation by this process is not the same as the true probability of that conformation occurring on a purely stochastic basis. This in turn introduces a considerable bias into the ensemble which we remove through the system of chain weighting introduced by Rosenbluth and Rosenbluth12and sampled by the method of Metr0polis.1~The basal face of the lattice was taken to represent an impenetrable solid surface for which the net enthalpy of adsorption, as quantified by the x s parameter,14 was always favorable. The solvancy of the chain was considered to be athermal. Once formed, each chain was artificially displaced in the direction perpendicular to the surface until at least one segment was in the first lattice layer, i.e., in contact with the surface. This represented the starting position for the simulation. Each chain was allowed to approach its equilibrium adsorbed conformation through a succession of elementary moves operating on randomly selected segments in the chain. It was found that in excess of one million moves were required before a N = 49 SAW started to adopt a stable conformation, as judged by the convergence of the bound fraction, p, with the number of moves. The set of elementary moves used in this simulation included the gauche-gauche, gauche-trans and trans-gauche, “end-ofchain” moves, and also the 180’ flip of a pair of gauche bonds known as a “kink-j~mp”.’~ In addition, we have also included the two-atom “crankshaft” motion recently shown as necessary to establish the true dynamics of these lattice models.16 We note in passing recent comments on the nonergodicity of Verdier’s elementary moved7 for three-dimensional SAW’Slonger than N = 18. However, since we are only using these moves to effect conformational changes and are not drawing quantitative time dependencies from the data, we feel justified in our use of these moves. As indicated above, a move is only allowed if it does not contravene the excluded volume criterion of the SAW. The type of move performed depends on the position of the selected segment, the local conformation of the chain, and the overall number of segment-segment and segment-solvent contacts for both the current conformation and the proposed new conformation. In this primitive model we do not impose a periodic boundary constraint18 but this will be discussed in a future paper.” Although this restriction is somewhat severe, we do not expect to see very different qualitative effects in the case of chemisorption because of the severe lateral constraints that are imposed by irreversible surface anchoring. In more traditional lattice simulations of adsorbed polymers, chains are often grown from terminal or medial attachment points on the surface. By introduction of a probability of (12) Rmenbluth, A. W.; Rosenbluth, M. N. J. Chem. Phys. 1966,23, 256. (13) Metropolis, N.; Rosenbluth, A. W.; Rosenbluth, M. N.; Teller, A. H. J. Chem. Phys. 1966,23, 356. (14) Silberberg, A. J. Chem. Phys. 1968,48, 2835. (15) Verdier, P. H.; Stockmayer, W. H. J. Chem. Phys. 1962,36,227. (16) Stokely, C.; Crabb, C. C.;Kovac, J. Macromol. Chem. 1986, 19, 860. (17) (a) Madras, N.; Sokal,A. K.J . Stat. Phys. 1987, 47, 573. (b) Verdier, P. H. J. Comput. Phys. 1969, 4, 204. (18) Clark, A. T.; Lal,M. Polymer 1976,16, 310.
2208 Langmuir, Vol. 8, No. 9, 1992
Cosgrove et al.
Normalized 8egment denelty
Bound Fnctlon 0.8
O
'
'
I
i
0.6
i
0.1
+ n
0 5
10
15 z/ layers
20
25
0
30
Figure2. Simulated normalized segment density profiiee for an isolated chain ( N = 20, X. = 0.6); ., chemisorptive model; 0 , physisorptive model.
desorption, the scheme described above makes it possible to investigatevarious mechanisms of polymer adsorption. We have considered three such models: (a) a "chemisorptive" model where segments are irreversibly adsorbed, (b) a "physisorptive" model where there is a finite probability of an adsorbed segment desorbing, and (c) a 'composite" model where an initially physisorptive interaction becomes more chemisorptive the more the chain relaxes. The probability of desorption is zero for case a and is weighted by xs in case b and by both xa and the number of moves in case c. These differences are described in terms of the probability of desorption per monomer, P d . In the case of the composite model we have chosen an exponential form for the variation of P d with simulation time
= exp[-(x, + k,m)J (1) lz, is the rate of chemisorption and m is the current number of moves. Although this is a purely empirical choice, it emphasizes the effect that after a few surface-polymer bonds have formed,desorption becomes progressively more difficult. pd
Results and Discussion The theoretical normalized segment density profiles for the chemisorption and physisorptionmodels for an isolated chain are shown in Figure 2. In both cases N,the chain length, x,, and the total number of moves have been kept the same. It can be seen that in the case of a physisorbed chain there is a monotomic decrease in the segment density profile awayfrom the surface. This is the normal situation for a chain relaxing into its equilibrium conformation at a surface. Given an infinite simulation time and a very high x8, it would be expected that the chain would lay completely in two dimensions. Similar shaped profiles have been obtained with the Scheutjens-Fleer model718 For the chemisorbed and other Monte Carlo sim~lations.~~ chain, very little surface rearrangement can take place, as those segments which chemisorb cannot desorb. Consequently, the chain retains much of its initial (bulk solution) conformation. The segment density distribution for a random coil chain in solution is approximately Gaussian. If such a random coil were to be irreversibly adsorbed, then a maximum in the segment density profile should occur away from the surface and this is in fact what is seen. The segment density profile is of course only depicting a static average of the final conformationsof all chains in the ensemble. The dynamic behavior of the (19)Cosgrove, T.J. Chem. SOC.,Faraday Tram. 1990,86, 1377.
1
I
I
20.000
40,000
80,000
100,000
80,000
I
120,OOO
No. of m o v n
Figure 3. Simulated averagebound fraction for an isolated chain 88 a function of the number of elementary moves under the physisorptive model (N = 50): xB= 1.0 (+); xI = 0.3 (m). Bound Fraction 1
t
0' 0
1
I
1000
2000
I
3000
Do
Timetmin
Figure 4. Experimental average bound fraction 88 a function of time for the adsorption of PMS [MN = 8611onto aluminafrom tetrachloromethane: +, ( p c ) ;*, ( p p ) .
chains however can be investigated through the evolution of the bound fraction of segments, p , with the number of moves. Figure 3 shows how a "running" average of p, (p), varies with the number of moves for two values of using the physisorptive model. For the first case, xais 1.0 [+I, representing a net attractive surfacefor the segments, while in the second xa is 0.3 [HI, which is close to the critical value for adsorption to occur, xsc. For xs = 1.0 ( p d is low) ( p )gradually increases with time toward a limiting value as a dynamic equilibrium between free and adsorbed segments is established. Conversely, for the second case, P d is much higher and so ( p ) decreases with time as the chain progressively desorbs. The smooth decrease of (p) is a result of the cumulative average wherein each new step is weighted with all preceding ones. These two simulations show qualitatively the effect of surface attraction on the value ofp. The combined chemisorption and physisorption in the experimental system is an intermediate case, and although xSis greater than xsc,the number of possible surface moves that can take place depends on the value of kc. The variation with time ofpc and p p in the experimental system is shown in Figure 4. The relative lability of the PMS-alumina interaction in the experimental system might be expected to result in the experimental p p exhibiting behavior typical of adsorption on a high xs surface (Figure3). Instead, asRc becomes more dominant pp is seen to pass through a maximum and then decrease. The physical picture behind this is that it is the initially physically adsorbed segmentewhich become chemisorbed.
Langmuir, Vol. 8, No. 9, 1992 2209
Dynamics of Physisorbing and Chemisorbing Polymers Gamma [mglg]
Bound Fraction
O
30t f t d
i
* I
I O0 65I
0
I
I
I
I
I
I
500
1000
1500
2000
2500
3000
l
" 8 -
0
20000
40000
60000
80000
100,000
120000
Time/min
No of moves
Figure 5. Experimental adsorbed amount as a function of time for the adsorption of PMS [MN= 8511 onto alumina from tetrachloromethane: (m) 353 K; (*) 293 K.
Figure 7. Simulated average bound fractionfor an isolatedchain as a function of the number of elementary moves under the composite model (one chain, N = 50): *, k, = 0.3; ., k, = 1.0.
Chemisorbed bound fraction 1
t
I
retaining some of its extended solution conformation on chemisorption and being prevented from relaxing into a flatter position. In Figure 6 the values ofp, obtained from the same experiment are shown, and as expected the higher values of the adsorbed amount correspond to smaller values of the bound fraction. This again emphasizes the extended conformation of the chemisorbed chain at the higher temperature and the more two-dimensional behavior a t the lower temperature. We can simulate this behavior using the composite model if we increase k,. In Figure 7 the evolution of p with simulation time is shown for two different values of k,. For simplicity the same values have been used for both k, and xs. With this model, Pd initially takes the value it would have under the physisorptive model; however the effective adsorption energy increases with time (viz. number of moves). At the higher temperature alarge value of k, leads to a more rapid convergence and a small value of the total bound fraction. This is qualitatively the same behavior as was found in the experimental system.
o,61 1 0.4
Conclusions For polymers that react chemically with surfaces the final adsorbed conformation is determined by the relative rates of physisorption and chemisorption. Data from a dynamical Monte Carlo computer model show qualitative agreement with experimental results and suggest that such models may have an important role to play in improving our understanding of the dynamics of polymer adsorption. Acknowledgment. We thank Dr. R. Richards and Dr. T. Semlyen of the University of York for the provision of the PMS sample. S.M.K. acknowledges EXXON Chemical for financial support and in particular Dr. I. More for his continuing interest in the project. C.A.P. acknowledges the SERC and IC1 Plant Protection for the provision of a CASE award and Dr. Th. F. Tadros for many useful discussions.