Stephen
King
*a,
Clive
Washington‡
b and
Richard
Heenan
a
aISIS Facility, Rutherford Appleton Laboratory, Chilton, Didcot, UK OX11 0QX. E-mail: s.m.king@rl.ac.uk; Fax: +44 (0)1235 445720; Tel: +44 (0)1235 446437
bSchool of Pharmaceutical Sciences, University of Nottingham, University Park, Nottingham, UK NG7 2RD
First published on 10th November 2004
Small-angle neutron scattering has been used to investigate the structure of the adsorbed layers formed by three poly(oxyethylene)-poly(oxypropylene) block copolymers (Poloxamer 188, Poloxamer 407 and Poloxamine 908) in two similar oil-in-water emulsions; one with a fluorocarbon oil phase, the other with a hydrocarbon oil phase. We find that the structure of the buoy layer is remarkably similar in the two cases. Since previous work has shown that the fluorocarbon/water interface is impenetrable to these polymers, these new findings also suggest that there cannot be significant penetration of the hydrocarbon/water interface. This is discussed in relation to the findings from other studies. The effects of added electrolyte and temperature on the adsorbed layer structure are also reported.
In earlier papers we have reported on our work with two PEO-PPO-PEO triblock species (Poloxamer 188 and Poloxamer 407), and a four-arm (each a PEO-PPO diblock) star-like material (Poloxamine 908). The colloidal system being stabilised by the adsorbed polymer was a perfluorocarbon-in-water emulsion.3,4 This was not only easy to prepare in a reproducible manner and well suited to our principal method of investigation, small-angle neutron scattering (SANS), but also had clinical analogues (such as the “artificial blood” substitutes).
Our interest in this system was driven by the now well established fact that the structure, and not just the chemical nature, of the adsorbed layer can have a profound influence on the in vivo fate of biocolloids.5,6 Given the extensive literature concerning the use of synthetic colloidal systems as carriers for the targeted delivery of drugs, being able to probe and describe the adsorbed layer is clearly of some importance.
In this paper we report the results of a complementary investigation designed to augment our earlier work. We have repeated our earlier SANS experiments but this time utilising hydrocarbon (as opposed to fluorocarbon) oils, to see if the changes in the Hamaker constant of the oil, the polymer–interface interaction, and perhaps the solubility of the blocks has any effect on the structure of the adsorbed layer.
Samples of Polaxamer 188 (Pluronic F68), Poloxamer 407 (Pluronic F127), and Poloxamine 908 (Tetronic 908) were donated by BASF, Parsippany, NJ, USA. The polydispersities (Mw/Mn) of the polymers were typically less than 1.3. Bulk densities have previously been measured by liquid pycnometry in ethanol. The characteristics of the polymers are summarised in Table 1.
Poloxamer 188 | Poloxamer 407 | Poloxamine 908 | |
---|---|---|---|
a Table includes data from ref. 26. b NSLD = neutron scattering length density. | |||
Empirical formula | PEO76–PPO29–PEO76 | PEO101–PPO67–PEO101 | (PEO119–PPO17)2–N–C2H4–N–(PPO17-PEO119)2 |
MPEO/g mol−1 | 6695 | 8900 | 20970 |
MPPO/g mol−1 | 1685 | 3890 | 3950 |
Fraction of PPO, νPPO | 0.16 | 0.25 | 0.13 |
Bulk density, d/g cm−3 | 1.160 | 1.141 | 0.914 |
NSLD b, ρ/cm−2 | 0.59 × 1010 | 0.56 × 1010 | 0.49 × 1010 |
Due to the need to prime the microfluidizer with polymer solution, the disperse phase fractions are subject to small compositional variations and were therefore assayed by specific gravity measurements after redispersal. The final droplet size distributions were measured by dynamic light scattering (DLS) with a Malvern 4700 correlator (using the cumulants method of data treatment and CONTIN for the size distribution analysis). Some of the hydrocarbon emulsions were prepared with sodium chloride added after homogenisation as an electrolyte. The characteristics of the stock emulsions are summarised in Table 2.
Poloxamer 188 | Poloxamer 407 | Poloxamine 908 | |
---|---|---|---|
a Parameter used to describe the form of a log-normal particle size distribution, see Appendix. | |||
Volume fraction, ϕ | 0.009 | 0.010 | 0.010 |
DLS volume mean radius/nm | 635.0 | 659.5 | 575.0 |
DLS surface mean radius/nm | 533.4 | 544.4 | 409.0 |
DLS polydispersity, P | 0.44 | 0.47 | 0.63 |
Polydispersity, β0a | 0.421 | 0.447 | 0.578 |
Neutron scattering length densities for the polymers and for the dispersion medium at 25 °C were obtained as appropriately weighted sums of the calculated values7,12 for PEO (+0.637 × 1010 cm−2 for d = 1.127 g cm−3) and PPO (+0.343 × 1010 cm−2 for d = 1.004 g cm−3) and for H2O (−0.559 × 1010 cm−2 for d = 0.997 g cm−3) and D2O (+6.355 × 1010 cm−2 for d = 1.105 g cm−3), respectively. The scattering length density of the dispersed phase was calculated from the values for decalin (−0.033 × 1010 cm−2 for d = 0.896 g cm−3) and perdeuterodecalin (+7.300 × 1010 cm−2 for d = 1.014 g cm−3) in a similar manner.
(1) |
Under the conditions that the hydrocarbon oil phase and dispersion medium are at “contrast match”, and ignoring any contribution from fluctuations in the average layer structure, the scattering from the polymer layer can be described by eqn. (2)13
(2) |
We have obtained estimates for Γ, √〈z2〉 and σ by non-linear least squares fitting of the scattering data in the region 0.04 < Q (nm−1) < 0.3 to eqn. (2), averaging over both polydispersity in R (log-normal distribution, see Appendix) and instrument resolution (rectangular distribution) at the same time. With the exception of the Q-independent background all other parameters were constrained to their experimentally determined values. Once the fit had started to converge R and ϕ were also allowed to refine. The maximum change in these values was ca. 3%.
An alternative to eqn. (2), though less satisfactory for incorporating polydispersity, is the following profile independent surface-Guinier type function, eqn. (3)14
(3) |
The polymer volume fraction profiles, Φ(z), that describe the average distribution of segments in the interfacial region15 can be obtained in two ways. One is the Hilbert transformation method of Crowley.14,16 This yields a model-independent profile consistent with the data, although considerable data manipulation is required prior to inversion and even then the resulting profile may still contain truncation artefacts. A more fundamental problem with this approach, however, is that the transform uses the interface as a phase reference point. It does not matter if the interface is invisible to the neutrons provided that the interface is impenetrable (since then the droplet radius may be used to locate it). We had reason to believe that in this work (and contrary to our experience with the fluorocarbon emulsions) the polymers might penetrate the hydrocarbon oil interface. For this reason we have used an indirect profile-fitting procedure developed by Cosgrove.17 In this approach a parameterised mathematical form for the volume fraction profile is back-transformed into scattering data which is then compared to the experimental scattering data. Iteration of the parameters then generates the profile of the chosen mathematical form that best describes the data. It must be stressed here that although modern theoretical descriptions of polymer adsorption can be used to guide the selection of profiles, the only reliable methodology is to test several individual functional forms for Φ(z) on each experimental data set. Before adopting the indirect profile-fitting procedure we also considered it prudent to assess its validity. This is discussed later. At the same time we address an oversight in our earlier work.
Given a volume fraction profile it is then obviously possible to derive other estimates of the thickness of the adsorbed layer such as the span, l, the distance from the interface at which Φ(z) ≈ 0 (or, more practically, the distance at which the integral of Φ(z) reaches ca. 95% of the total), and also the average fraction of polymer segments at the droplet interface, 〈p〉, (eqn. (4))
(4) |
Fig. 1 Comparison of the calculated volume fraction profiles for the three copolymers. The SANS data from which these are derived is that previously discussed in ref. 3 and is from fluorocarbon emulsions at 25 °C in the absence of electrolyte. The points were generated by the mathematical inversion procedure, and the lines by the indirect profile fitting procedure assuming exponential profiles. (◆) and (——) Poloxamer 188, (■) and (– – – –) Poloxamer 407, (○) and (–·–·–·–) Poloxamine 908. |
The resulting new profiles are also displayed in Fig. 1 (as the lines), and in Table 3 we compare the physical parameters derived from both analyses. The correspondence is reassuringly good. The small differences in Γ, σ and l are probably of the same order as the experimental error. The more noticeable discrepancy in 〈p〉 for Poloxamer 188 is probably because the volume fraction profile for this polymer is the least exponential like, particularly at small z.
Hilbert inversion methoda (profile-independent) | Indirect profile-fitting method (exponential profile) | ||||||||
---|---|---|---|---|---|---|---|---|---|
Γ b/(mg m−2 | σ b/nm | l/nm | 〈p〉 | /mg m−2 | σ/nm | √〈z2〉/nm | l/nm | 〈p〉 | |
a Cf. with data in Tables 4 and 5 in ref. 3. Data presented here is a re-analysis. b From model fits to eqn. (3). | |||||||||
Poloxamer 188 | 1.8 | 2.0 | 6.4 | 0.55 | 2.0 | 2.8 | 3.9 | 10.0 | 0.39 |
Poloxamer 407 | 3.3 | 3.2 | 15.3 | 0.25 | 3.7 | 4.9 | 7.0 | 15.0 | 0.24 |
Poloxamine 908 | 1.5 | 3.1 | 17.6 | 0.27 | 1.7 | 5.4 | 7.6 | 16.0 | 0.22 |
Now compare the data in Table 3 with the corresponding new data from the hydrocarbon emulsions in Table 4. Once again, exponential profiles were chosen to model the polymer distribution. This choice will be further justified later. Table 4 also contains estimates of Γ, σ and √〈z2〉 derived from fitting the SANS data to eqn. (2). The measures of the layer thickness are reasonable, but this function appears to overestimate the adsorbed amount. Fig. 2 shows a possible explanation for this. Whilst the various measures of the layer thickness are derived from the Q-dependence of the SANS data, it is the absolute magnitude of the data that determines Γ. Eqn. (2) appears to over-estimate the differential cross-section between 0.06 < Q (nm−1) < 0.09.
Fig. 2 The SANS data from the hydrocarbon emulsions with adsorbed copolymer at 25 °C in the absence of electrolyte. The experimental uncertainty in (∂Σ/∂Ω)(Q) ranges from about 7% at very low Q to around 4% out at Q ∼ 0.3 nm−1. Error bars are omitted for clarity. (·········) Fit of eqn. (2) to the Poloxamer 188 data. Other lines represent the best exponential profile fits to the experimental data. (◆) and (——) Poloxamer 188, (■) and (– – – –) Poloxamer 407, (○) and (–·–·–·–) Poloxamine 908. |
Without polydispersity (eqn. (3)) | With polydispersity (eqn. (2) or exponential profile fitting) | |||||||||
---|---|---|---|---|---|---|---|---|---|---|
χ 2 | Γ/mg m−2 | σ/nm | χ 2 | /nm | Γ/mg m−2 | σ/nm | √〈z2〉/nm | l/nm | 〈p〉 | |
Poloxamer 188 | ||||||||||
(eqn. (2)) | 622. | 532.8 | 2.8 | 5.4 | 5.6 | n/a | n/a | |||
(See header) | 2.25 | 1.8 ± 0.01 | 3.8 ± 0.3 | 0.88 | 537.7 | 1.9 | 3.6 | 5.0 | 12.0 | 0.32 |
Poloxamer 407 | ||||||||||
(eqn. (2)) | 5071. | 554.5 | 4.2 | 4.2 | 6.1 | n/a | n/a | |||
(See header) | 1.29 | 2.7 ± 0.01 | 3.8 ± 0.1 | 1.43 | 541.2 | 3.0 | 4.5 | 6.3 | 14.0 | 0.27 |
Poloxamine 908 | ||||||||||
(eqn. (2)) | 657. | 421.5 | 1.6 | 4.6 | 6.1 | n/a | n/a | |||
(See header) | 2.27 | 1.0 ± 0.01 | 4.1 ± 0.3 | 1.36 | 403.3 | 1.1 | 4.5 | 6.3 | 14.0 | 0.27 |
The first observation is that the trends in the data are identical for the fluorocarbon and hydrocarbon systems; in the case of the adsorbed amount, Poloxamer 407 > Poloxamer 188 > Poloxamine 908, whilst for measures of the layer thickness, Poloxamine 908 ≥ Poloxamer 407 > Poloxamer 188. Though the actual magnitudes of these parameters are quite similar, there are some interesting variations. For example, whilst Γ for Poloxamer 188 is essentially unchanged, the values for Poloxamer 407 and Poloxamine 908 are somewhat depressed in the hydrocarbon emulsion systems. In addition, whilst the Poloxamer 407 and Poloxamine 908 adsorbed layers are slightly contracted (or essentially unchanged), the Poloxamer 188 adsorbed layer is slightly expanded, compared to the equivalent fluorocarbon emulsion system. In our previous paper3 we have commented at length on how our data compares with that in the literature and so we shall not do so again here. However, there are two literature reports that are particularly relevant to this work.18,19 These both studied PEO-PPO-PEO block copolymers adsorbed at the hexane–water interface using neutron reflectometry (NR), and concluded that the PEO blocks were stretched beyond their solution radii-of-gyration (a PEO homopolymer with the same molecular weight as Poloxamer 188 would have an Rg ∼ 3.2–3.8 nm 20,21). Comparison of our SANS data for Poloxamer 188 in Tables 3 and 4 only suggests that the adsorbed polymer has a size comparable to Rg. However, all these values are still significantly smaller than the extended length of an (EO)76 chain, ca. 10 nm.22
Compared to what is observed in the fluorocarbon emulsion systems, the bound fraction data for Poloxamer 188 appears to exhibit a small decrease, whilst those for Poloxamer 407 and Poloxamine 908 increase slightly. The latter trend is actually not too suprising since, by analogy with the established behaviour of homopolymers, an increase in 〈p〉 should be accompanied by a thinning of the layer (fewer segments available to form the buoy layer) and a reduction in Γ (because those polymer molecules already at the interface occupy more of it1). In our earlier work we identified a clear correlation between Γ and νPPO (see Table 1), but not between 〈p〉 and νPPO.3 As a consequence, we were able to explain the observed differences in adsorbed layer structure in terms of the relative proportions of PPO (and even a small number of PEO) segments located at the interface. Here, we observe the same strong correlation with Γ which suggests that, once again, it is the PPO segments that are preferentially located at the interface. Since the 〈p〉 values for Poloxamer 188 and Poloxamine 908 are somewhat higher than their corresponding νPPO it is again likely that some PEO segments are also kept in the vicinity of the interface. However, the lower 〈p〉 of Poloxamer 188 in the hydrocarbon emulsion systems would suggest that a slightly smaller fraction of PEO segments are present at this interface. This would in fact be consistent with the observed thickening of the Poloxamer 188 buoy layer. Interestingly, the data suggests that this subtle effect is achieved without any real change in the number of adsorbed polymer molecules. This may be an effect of the lower curvature of the hydrocarbon emulsion droplets.
The best-fit exponential volume fraction profiles for the hydrocarbon emulsions are shown in Fig. 3, and should be compared with those for the fluorocarbon emulsions in Fig. 1.
Fig. 3 Comparison of the calculated volume fraction profiles for all three copolymers in the hydrocarbon emulsion systems at 25 °C in the absence of electrolyte and generated by indirect profile-fitting assuming exponential profiles. (——) Poloxamer 188, (– – – –) Poloxamer 407, (–·–·–·–·) Poloxamine 908. |
The NR studies by Clifton et al.18 and Phipps et al.19 also concluded that the copolymers PEO76-PPO38-PEO76 and Poloxamer 407, respectively, penetrated the hexane/water interface. Clifton et al. determined the extent of the polymer in the oil phase to be 0.8 ± 0.4 nm and Phipps et al. measured 4 ± 1 nm. As it is logical to assume that it is the more hydrophobic PPO block which would preferentially locate in the oil phase then the (seemingly big) difference in these values may partly be due to the difference in the number of segments in the respective PPO blocks (although νPPO is actually very similar).
However, where there is penetration of the interface, the volume fraction profile is more likely to be an asymmetric Gaussian (peaked in the oil phase but close to that side of the interface) and not the exponential form with which we have analysed our present data. However, since the oil phase and the bulk phase are at contrast match in our experiments (that is, they have been prepared such that each has the same ρ value) the neutrons cannot see the interface, only the polymer that may be spread across it. We have tried to assess the effect on the exponential profiles of polymer penetrating the oil/water interface by repeating some fits but with the droplet radii reduced by an arbitrary but physically meaningful amount (2 nm), and ϕ reduced by a corresponding factor (of R3/R03). As might be expected there is a small increase in Γ (arising out of the reduction in available surface area), but the starting height of the profiles, σ, √〈z2〉 and l are otherwise essentially unchanged. From this we conclude that our analysis is insensitive to the position of the oil/water interface and that our data does not contain sufficient information for us to distinguish between a Gaussian and an exponential profile. In any case, if a Gaussian is highly asymmetric it will approximate to an exponential. Attempts to fit some (symmetric) Gaussian profiles were not very successful; in the absence of additional a priori information the extra parameters introduced too much variation for our satisfaction. This is why we have adopted the pragmatic approach of only fitting exponential profiles. What we can say with confidence is that block (step-like) or parabolic profiles are much less satisfactory descriptions of the adsorbed polymer distribution than an exponential.
Phipps et al. modelled their NR data with three layers (of which two were in the aqueous phase) and this led to a volume fraction profile that might be described as a ‘stylised Gaussian’ (see Fig. 6 in ref. 19). For Poloxamer 407 they report Γ ∼ 2.9 mg m−2, σ ∼ 2.4 nm and l ∼ 9 nm. Whilst their adsorbed amount is in excellent agreement with our data, their thickness measures appear somewhat smaller until one adds on the thickness of the polymer in the oil phase mentioned above. The particular advantage of their experiment over ours was that they varied ρ of the oil phase in order to locate the interface and to highlight those polymer segments on the oil side.
Another consideration may be the presence of capillary waves travelling along the surface of these relatively large droplets. A rough calculation (see Appendix) suggests that if present the droplet interface would have a r.m.s. roughness of no more than 0.4 nm. This is in line with the NR data from the hexane/water interface, in the presence of Poloxamer 407, which ascribed to the interface a Gaussian roughness of ca. 1 nm.18 Both of these roughness values are actually very similar to the error bounds on the σ values derived from fitting the SANS data to eqn. (3).
In our earlier work with the fluorocarbon emulsion systems4 we found that increasing the concentration of electrolyte in the bulk phase or increasing the temperature had an effect on the structure of adsorbed layers of Poloxamer 188. Specifically, for sodium chloride concentrations up to about 0.6 M, or for temperatures in the range 25–45 °C, we observed an expansion of the adsorbed layer of 85–300% (dependent on the thickness measure used). At higher electrolyte concentrations, or higher temperatures (up to 55 °C), the adsorbed layer contracted (but did not noticeably diminish below the 0 M salt or 25 °C thickness). Throughout these changes the adsorbed amount remained effectively constant, although there was a steady but small decrease in the bound fraction. This behaviour, which is more noticeable with electrolyte, and which has many parallels in other solution studies of oxyalkylene-based surfactant systems, was attributed to the effect on the oxyethylene segments of changes in the water solvent quality.
We have repeated these studies, this time also including Poloxamer 407 and Poloxamine 908, as part of this work. Some results, for Poloxamer 188 for brevity, are shown in Tables 5 and 6. Additional data may be found in the associated electronic supplementary information.† The same trends identified in the fluorocarbon emulsion systems are just about discernible.
[NaCl]/M | χ 2 | /nm | Γ/mg m−2 | σ/nm | √〈z2〉/nm | l/nm | 〈p〉 | |
---|---|---|---|---|---|---|---|---|
Poloxamer 188 | 0.09 | 1.12 | 536.4 | 1.9 | 3.4 | 4.8 | 11.6 | 0.34 |
0.36 | 0.94 | 531.8 | 1.9 | 3.6 | 5.0 | 12.0 | 0.32 | |
0.56 | 0.79 | 524.7 | 1.5 | 3.7 | 5.1 | 12.2 | 0.32 | |
0.76 | 0.84 | 533.5 | 1.7 | 3.3 | 4.6 | 11.2 | 0.35 | |
1.02 | 0.78 | 521.9 | 1.7 | 3.3 | 4.6 | 11.2 | 0.35 | |
1.20 | 0.96 | 524.0 | 2.0 | 3.3 | 4.6 | 11.2 | 0.35 | |
1.41 | 0.67 | 535.0 | 1.9 | 3.1 | 4.4 | 10.8 | 0.36 |
T/°C | χ 2 | /nm | Γ/mg m−2 | σ/nm | √〈z2〉/nm | l/nm | 〈p〉 | |
---|---|---|---|---|---|---|---|---|
Poloxamer 188 | 30.0 | 0.90 | 536.6 | 1.9 | 3.6 | 5.0 | 12.0 | 0.32 |
35.0 | 1.36 | 537.5 | 2.0 | 3.4 | 4.8 | 11.6 | 0.34 | |
40.0 | 1.07 | 535.3 | 2.0 | 3.5 | 4.9 | 11.6 | 0.33 | |
45.0 | 1.16 | 536.3 | 2.1 | 3.5 | 4.9 | 11.8 | 0.33 | |
48.5 | 1.12 | 535.9 | 2.0 | 3.4 | 4.7 | 11.4 | 0.34 | |
50.0 | 0.96 | 534.1 | 2.0 | 3.5 | 4.9 | 11.8 | 0.33 |
Interestingly, the repeat runs on the different electrolyte samples (Table 5) have consistently lower Γ values than the initial measurements, though 〈p〉 and the various measures of the layer thickness remain quite similar. Since there is no clear pattern to the small changes in the droplet size, Ostwald ripening does not seem to be responsible and so we conclude that polymer has desorbed from the interface. In each instance the repeat measurement commenced almost exactly 6 h later. In our previous work with the fluorocarbon emulsions4 scattering data was typically accumulated as four 1 h runs but spread over a much longer duration. Those data did not show any comparable change in adsorption. Electrolyte-induced desorption would therefore appear to be a feature of the hydrocarbon emulsion systems. One may speculate that this is related to the differences in segment–interface interaction parameters resulting from the reduction in the polarizability of the C–H bond compared to the C–F bond.
We have previously investigated adsorption in the fluorocarbon emulsion system in some detail3,4 and concluded that the polymer segments did not penetrate the interface. This was in keeping with both the general insolubility of hydrocarbon polymers in fluorocarbons and our own observations (e.g., attempting to dissolve the polymers in perfluorodecalin). In contrast, there are literature reports18,19 that suggest these polymers should be capable of penetrating the hydrocarbon/water interface. We have been unable to find any unambiguous evidence of this, indeed the polymers are also insoluble in normal decalin, which suggests that if there is indeed penetration it can at best only be of the order ≤1 nm (given that the resolution of our measurements is ca. 0.1 nm.
Apart from a slight reduction in the amount of polymer adsorbed at the interface, our data suggest that there is otherwise considerable similarity in the structure of the adsorbed layer (and particularly that of the stabilising PEO blocks) between fluorocarbon–water and hydrocarbon–water emulsion systems. In both types of emulsion the PEO buoy blocks appear to extend distances comparable to their solution radii-of-gyration.
As found in the earlier fluorocarbon emulsion study, changes in temperature or electrolyte concentration are able to bring about subtle modifications of the layer structure.
(A.1) |
ln = ln Rm + 1·5 β02 | (A.2) |
β = Rm[exp(4β02) − exp(3β02)]1/2 | (A.3) |
(A.4) |
(A.5) |
(A.6) |
Footnotes |
† Electronic supplementary information (ESI) available: A schematic representation of a copolymer-stabilised emulsion droplet and the corresponding volume-fraction profile (Fig. S1); comparison of parameters derived from the SANS data from the hydrocarbon emulsions as a function of electrolyte concentration or temperature using the profile-fitting method of analysis with exponential profiles and incorporating droplet size polydispersity (Tables S1 and S2). See http://www.rsc.org/suppdata/cp/b4/b414175j/ |
‡ Present address: AstraZeneca, Macclesfield Works, Hurdsfield Industrial Estate, Macclesfield, Cheshire, UK SK10 2NA. |
This journal is © the Owner Societies 2005 |