Malcolm J. W. Povey*a, Scott A. Hindlea, John D. Kennedya, Zoe Steca and Richard G. Taylorb
aUniversity of Leeds, Leeds, UK LS2 9JT. E-mail: m.j.w.povey@leeds.ac.uk; Fax: 44 (0) 1132332982; Tel: 44 (0) 1132332963
bDow Corning, Central Research (BA104) Cardiff Road, Barry, South Glamorgan, Wales CF63 2YL. E-mail: r.taylor@dowcorning.com; Fax: 44 (01446) 747944; Tel: (01446) 723841
First published on 19th November 2002
The velocity of sound has been determined in pure liquid n-alkanes (pentane to hexadecane), 1-alcohols (methanol to 1-dodecanol) and dimethylsiloxanes (200 fluid (L2, 0.65 cSt) to 200 fluid (5000 cSt)) at 20°C. Corresponding density data have been taken from the literature and the adiabatic compressibility determined. The measured adiabatic compressibility has been compared with two molecular models of sound velocity, the Schaaffs model and a development of the Urick equation. The Urick equation approach is based on a determination of the compressibility of the methylene or siloxane repeat units which make up the chains in these linear molecules. We show that the Urick equation approach accurately predicts sound velocity and compressibility for the higher members of each series, whilst the Schaaffs approach fails for the 1-alcohols. We suggest this is because of the influence of the hydroxyl end group on nearby molecules through hydrogen bonding. The Schaaffs approach does not take into account interactions between the end units and the chain repeat units. This interaction modifies the derived compressibility of the methylene groups reducing their compressibility as compared to their compressibillity in the n-alkanes. The technique described provides valuable new insights into end-group, inter-molecular and intra-molecular interactions in liquid linear-chain molecules. We suggest that it provides a powerful new method for characterising polymeric fluid materials.
(1) |
A semi-phenomenological molecular model of sound velocity was devised in 1941 by Schaaffs4,5 as shown in eqn. (2).
v=WρB/M−W/(1+B/β) | (2) |
B=S1(2n+2)+S2n+S3 | (3) |
The Schaaffs approach requires that the product (V·v) of molar volume (V) and sound velocity be plotted against ‘molecular’ volume B, the slope giving the coefficient W. The value of W is usually between 4000–4500 m s−1, indicating the velocity of sound within a molecule, where frequency tends to zero. In general the three coefficients in eqn. (3) must be predetermined and these are tabulated elsewhere.4,5 So this approach requires the determination of four separate coefficients.
In our alternative approach, we assign a compressibility and density to each repeat unit in the series. The repeat unit is a {CH2} (methylene) unit in both the n-alkane and the 1-alcohol series and a {SiMe2O} (dimethlysiloxane) unit in the pmds series. We assign a separate compressibility and density to the terminating groups: both are {CH3} (methyl) in the n-alkanes, one is methyl and one is {OH} (hydroxy) in the 1-alcohols, and both are {SiMe3} (trimethylsilyl) in the pmds series.
The compressibility and density are combined using the Urick equation7 which is based on an idea suggested by Wood.8 Simply put, in a multicomponent system the compressibility and density in the Laplace eqn. (1) are replaced by their volume average values (eqns. (4) and (5)).
(4) |
(5) |
κ=(n−ne)κrϕr+neκeϕe | (6) |
ρ=(n−ne)ρrϕr+neρeϕe | (7) |
Later we will see that it is not so straightforward thus to identify and account for the end units, since the effect of the end units on the compressibility of the repeat units extends quite far down the chain. On the other hand, we shall see that the molar volume of the repeat units is hardly affected. Initially, we chose ne=4 to exclude repeat units attached to end units in our determination of repeat unit properties. However, the application of the Urick model to experimental results showed that the final value for ne varied and so in practice ne was chosen to minimise the standard error of the regression (see Table 2). It was thence found that ne (which appears in Table 2 as nregression) varied from zero, in the case of the fit of molar volume to carbon/silicon number, to six in the case of the compressibility in the n-alkanes.
We can see already a merit of the new approach in comparison to the previous Schaaffs method, in that ordinary measureable thermodynamic properties are assignable to individual chemical moieties.
Data analysis proceeds by first determining the molar volume as a function of chain length: linear regression of density against chain length gives the molar volume of each repeat unit and hence the fractional molar volume ϕr of the chain occupied by repeat units (eqn. (8)) and the fractional molar volume ϕe occupied by end units (eqn. (9)).
(8) |
(9) |
The compressibility of the repeat unit is then determined by linear regression of measured compressibility against the volume fraction of the chain occupied by repeat units. Hence the new method gives, explicitly, the adiabatic compressibility of the repeat unit and its molar volume. Examination of eqns. (6) and (7) indicates the physical basis for the Schaaffs method. If the contribution of the end units is ignored, then substitution of eqns. (6) and (7) into eqn. (1) gives:
(10) |
In other words, we expect a plot of the product of sound velocity and molar volume against the sound velocity of the repeat unit to yield a straight line whose slope is characteristic of the chemical repeat moiety. A corrollary of this would be that there would be a constant intercept contribution for a given molecule type that would vary according to the chemistry of the end units. Interactions between end units and neighbouring molecules are not explicitly accounted for in the Schaaffs model and the coefficients S1, S2 and S3 are a semi-phenomenological attempt to make the model fit the data.
In this regard our new method differs from the old, in that the effects of the molar volume and compressibility of the repeat and end units are explicitly accounted for. As will be seen this gives a more accurate model whose physical basis is more easily understood. This provides a firmer basis for comparison with other molecular models.
Sample | Silicon/carbon number | Measured velocity of sound/m s−1 | Density/g m−3 | Compressibility/Pa−1 | Molecular wt./g | Molar volume, V/cm3 | D/nm | (n−ne)·ϕr | B/cm3 |
---|---|---|---|---|---|---|---|---|---|
200 fluid (L2, 0.65 cs) | 2 | 912.5 | 761000 | 1.58E-09 | 162 | 212.9 | 1.24 | 0 | 12.48 |
Decamethyltetrasiloxane | 4 | 952.2 | 854000 | 1.29E-09 | 310 | 363.0 | 2.49 | 0 | 22.84 |
Dodecamethylpentasiloxane | 5 | 963.3 | 875000 | 1.23E-09 | 384 | 438.9 | 3.11 | 0.17 | 28.02 |
L6 PDMS (Me3SiO(Me2SiO)4SiMe3 | 6 | 971.8 | 887000 | 1.19E-09 | 458 | 516.3 | 3.73 | 0.29 | 33.2 |
DMS-T03 (3.0 cSt) | 7 | 976.6 | 898000 | 1.17E-09 | 550 | 612.5 | 4.51 | 0.40 | 39.64 |
DMS-T05 (5.0 cSt) | 10 | 985.0 | 918000 | 1.12E-09 | 770 | 838.8 | 6.36 | 0.56 | 55.04 |
DMS-T07 (7.0 cSt) | 13 | 989.3 | 930000 | 1.10E-09 | 950 | 1022 | 7.87 | 0.64 | 67.64 |
DMS-T11 (10 cSt) | 17 | 996.9 | 935000 | 1.08E-09 | 1250 | 1337 | 10.39 | 0.72 | 88.64 |
DMS-T12 (20 cSt) | 27 | 1006.9 | 950000 | 1.04E-09 | 2000 | 2105 | 16.70 | 0.82 | 141.1 |
DMS-T50 (50 cSt) | 51 | 1012.5 | 960000 | 1.02E-09 | 3780 | 3938 | 31.67 | 0.91 | 265.7 |
DMS-T21 (100 cSt) | 80 | 1015.2 | 966000 | 1.00E-09 | 5970 | 6180 | 50.09 | 0.94 | 419.0 |
DMS-T22 (200 cSt) | 131 | 1017.0 | 968000 | 9.99E-10 | 9730 | 10052 | 81.71 | 0.96 | 682.2 |
DMS-T23 (350 cSt) | 184 | 1017.6 | 970000 | 9.96E-10 | 13650 | 14072 | 114.7 | 0.97 | 956.6 |
DMS-T25 (500 cSt) | 233 | 1018.0 | 971000 | 9.94E-10 | 17250 | 17765 | 145.0 | 0.98 | 1208.6 |
DMS-T31 (1000 cSt) | 378 | 1017.6 | 971000 | 9.95E-10 | 28000 | 28836 | 235.4 | 0.99 | 1961 |
DMS-T35 (5000 cSt) | 667 | 1019.5 | 973000 | 9.89E-10 | 49350 | 50719 | 414.9 | 0.99 | 3455 |
Pentane | 5 | 1034.7 | 626200 | 1.49E-09 | 72 | 115.0 | 1.86 | 0.14 | 28.02 |
Hexane | 6 | 1103.6 | 659330 | 1.25E-09 | 86 | 130.4 | 2.23 | 0.25 | 33.2 |
Heptane | 7 | 1155.0 | 683750 | 1.10E-09 | 100 | 146.3 | 2.60 | 0.33 | 38.38 |
Octane | 8 | 1196.0 | 702670 | 9.95E-10 | 114 | 162.2 | 2.97 | 0.40 | 43.56 |
Nonane | 9 | 1233.1 | 717720 | 9.16E-10 | 128 | 178.3 | 3.34 | 0.45 | 48.74 |
Decane | 10 | 1256.2 | 730120 | 8.68E-10 | 142 | 194.5 | 3.71 | 0.50 | 53.92 |
Undecane | 11 | 1279.2 | 740200 | 8.26E-10 | 156 | 210.8 | 4.08 | 0.54 | 59.1 |
Dodecane | 12 | 1297.8 | 748750 | 7.93E-10 | 170 | 227.0 | 4.46 | 0.57 | 64.28 |
Tetradecane | 14 | 1332.1 | 762550 | 7.39E-10 | 198 | 259.7 | 5.20 | 0.62 | 74.64 |
Hexadecane | 16 | 1358.4 | 773530 | 7.01E-10 | 226 | 292.2 | 5.94 | 0.66 | 85 |
Methanol | 1 | 1124.8 | 791310 | 9.99E-10 | 32 | 40.4 | 0.37 | 0 | 11.83 |
Ethanol | 2 | 1164.9 | 789370 | 9.33E-10 | 46 | 58.3 | 0.75 | 0 | 17.01 |
1-Propanol | 3 | 1225.0 | 803750 | 8.29E-10 | 60 | 74.7 | 1.12 | 0 | 22.19 |
1-Butanol | 4 | 1257.9 | 809700 | 7.81E-10 | 74 | 91.4 | 1.50 | 0 | 27.37 |
1-Pentanol | 5 | 1293.3 | 814800 | 7.34E-10 | 88 | 108.0 | 1.87 | 0.15 | 32.55 |
1-Hexanol | 6 | 1319.9 | 819800 | 7.00E-10 | 102 | 124.4 | 2.25 | 0.27 | 37.73 |
1-Heptanol | 7 | 1344.6 | 822300 | 6.73E-10 | 116 | 141.1 | 2.62 | 0.35 | 42.91 |
1-Octanol | 8 | 1364.6 | 825800 | 6.50E-10 | 130 | 157.4 | 2.99 | 0.42 | 48.09 |
1-Nonanol | 9 | 1382.7 | 828000 | 6.32E-10 | 144 | 173.9 | 3.37 | 0.48 | 53.27 |
1-Decanol | 10 | 1397.7 | 829700 | 6.17E-10 | 158 | 190.4 | 3.74 | 0.52 | 58.45 |
1-Undecanol | 11 | 1408.7 | 832400 | 6.05E-10 | 172 | 206.6 | 4.12 | 0.56 | 63.63 |
1-Dodecanol | 12 | 1433.2 | 834300 | 5.84E-10 | 186 | 222.9 | 4.49 | 0.59 | 68.81 |
Symbol | Name | Material | |||
---|---|---|---|---|---|
n-Alkanes | 1-Alcohols | pmds | Units | ||
a se - standard error of the regression. | |||||
n | Number of members in series | 10 | 12 | 16 | |
End unit data | |||||
Chemistry | (CH3)...(CH3) | CH3...OH | SiO(CH3)3...Si(CH3)3 | ||
Ve | Partial molar volume | 33.4 | 24.9 | 64.3 | cm3 |
sea | 0.4 | 0.3 | 2.9 | cm3 | |
nregression | 10 | 12 | 16 | ||
ke | Compressibility | 1.370 | 0.793 | 1.293 | 10−9 Pa−1 |
sea | 0.010 | 0.006 | 0.003 | 10−9 Pa−1 | |
nregression | 5 | 6 | 12 | ||
Repeat unit data | |||||
Chemistry | CH2 | CH2 | SiO(CH3)2 | ||
Vr | Partial molar volume | 16.14 | 16.54 | 76.00 | cm3 |
sea | 0.04 | 0.04 | 0.01 | cm3 | |
nregression | 10 | 12 | 16 | ||
kr | Compressibility | −1.01 | −0.34 | −0.31 | 10−9 Pa−1 |
sea | 0.016 | 0.013 | 0.003 | 10−9 Pa−1 | |
nregression | 5 | 6 | 12 | ||
Schaaffs data | |||||
W | Ideal sound velocity | 4566 | 4496 | 14964 | m s−1 |
sea | 34 | 44 | 1 | m s−1 | |
nregression | 10 | 12 | 16 | ||
S1 | Schaaffs coefficients | 1.06 | 1.06 | 1.06 | m3 |
S2 | 3.06 | 3.06 | 3.06 | m3 | |
S3 | 4.59 | 4.53 | 0 | m3 |
Fig. 1 Measured and predicted compressibilities for n-alkanes, 1-alcohols and polydimethylsiloxanes. |
Fig. 2 Difference between the measured compressibility and that determined by the Urick equation, plotted against estimated half chain length. |
Fig. 3 Measured and predicted velocity of sound plotted against carbon/silicon number. |
Examination of Fig. 1 indicates that the Urick approach is accurate for the higher numbers in all three series, whilst the Schaaffs approach works well for pmds and n-alkanes but not for 1-alcohols. Urick is consistently more accurate than Schaaffs but fails, not unexpectedly, for the first few members of each series. In terms of intramolecular geometrical structure, the 1-alcohols are identical to the n-alkanes apart from the hydroxyl group which replaces one hydrogen atom on one methyl end group. There will, however, be considerable differences in intermolecular structure, particularly for smaller carbon numbers. In the alcohols, the hydrogen-bonding ability of the hydroxyl group will generate inter-hydroxy intermolecular association with the terminal hydroxy groups of adjacent molecules, effectively binding molecules together to increase effective molecular chain-length. Purely statistically, effective dilution of the hydroxy groups to ensure that none are adjacent will not occur until the carbon number exceeds eleven, although lone-pair orientation requirements for strong O–H–O interaction may effectively reduce this to six or seven. The low carbon-number deviations from the Schaaffs approach are in apparent agreement with this last. This hydrogen-bonding effect is obviously not a factor in the other two comparator series.
In sum, since the Schaaffs approach does not account for the effect of the end-groups on compressibility, it is least accurate for the 1-alcohols, especially at lower carbon numbers. On the other hand, it is clear that if it were possible to measure the 1-alcohols up to sufficiently high carbon numbers, then the effect of the end-groups should be greatly reduced by effective dilution, as is apparent for pmds in Fig. 2. It is apparent from Fig. 2 that, for the extent of chain-length examined, neither the 1-alcohol series nor the n-alkanes have sufficient methylene group content for the effects of the end-groups to become negligible. In the case of the n-alkanes and 1-alcohols, it would be necessary to repeat the measurements at higher temperatures (e.g. 60°C) in order to include the members with higher carbon numbers in the series that are solid at room temperature. The derived values for the compressibilities of the repeat units in the cases of the 1-alcohols and n-alkanes are therefore strongly influenced by the end units. This is especially so in the case of the 1-alcohols, where hydrogen bonding and other dipolar effects22 between adjacent chains in the 1-alcohols may have a strong influence on the derived compressibility of the repeat {CH2} unit. This may arise from a number of factors. In general we expect the bulk compressibility to be influenced by interactions with nearest-neighbour molecules. This will be particularly so when there is chemical bonding between neighbours—for example, hydrogen bonding can have values of up to ca. 30 kJ mol−1 or so, and compressibility transmitted through such a bond will clearly be different to that between molecules and to that within more strongly covalently bonded molecules. Intermolecular interactions involving the {CH2} repeat units may also have some significance; here there will be intermolecular Van der Waals attractions of up to 5 kJ mol−1 or so between pairs of methyl and/or methylene groups, but this would be expected to be approximately constant and identical, per repeat unit, for a standard assembly of such repeat units. Interaction of methylene and methyl hydogen atoms with the hydroxy group, which may have a stronger dipolar compenents and also some OH–HC dihydrogen-bonding character, could on the other hand have a much more significant influence on the derived methylene-unit compressiblities.
It is noteworthy that the effects of the end-groups disappear completely at approximately 10 nm into the chain in the case of pmds. This also appears to be the case in the n-alkanes and 1-alcohols, although we await further measurements to confirm that this is so. This value of 10 nm appears not to be a function of the number of repeat units, corresponding as it does to carbon/silicon numbers of six (pmds), eleven (n-alkanes) and six (1-alcohols), see also Fig. 3. It may perhaps be more closely related to the number of chemical bonds along a chain, which has some independence from chemical composition. For example, each silicon number represents a {SiMe2O} unit that is two-bonds long, whereas one {CH2} unit is one-bond long. The lower alkanes may be anomalous in this regard because they are closer to their boiling points at the temperatures of the measurements reported here, and therefore have a looser liquid-state ensemble.
From the present experiments, the value determined for the derived compressibility of the repeat units can only be regarded as a constant in the case of pmds. The derived values for the 1-alcohol and n-alkane series will change if more, higher, members of these series could be included. This is despite the apparent quality of the fit of the model to the experimental data. It is important to recognise that in no case can this derived compressibility value be regarded as intrinsic to the chain, affected as it is by end-group and nearest-neighbour interactions. For example, the methylene repeat units in the n-alkane and 1-alcohol series engender completely different derived compressibilities, despite being chemically identical. This is at first sight counter-intuitive. However, these derived compressibilities are not true methylene compressibilities, as the values will also constitute error-sinks for assumptions in the various models. As pointed out earlier, this will be partly due to hydrogen bonding between adjacent molecules in the case of the 1-alcohols, and other dipolar effects may also be expected.22 As a consequence, we cannot extrapolate from the behaviour of the short chain 1-alcohols and n-alkanes to the long chain behaviour, because we cannot be sure that the end effects have been correctly accounted for in our mathematical model. In addition, at chain numbers above around 100 in n-alkanes, chain folding may have significant effects.23 The partial molar volume, on the other hand, is nowhere nearly as sensitive to these chain–chain interactions. Whilst the Urick model suggests that the velocity will flatten out at high carbon numbers for both the n-alkanes and 1-alcohols just as the dimethylsiloxanes do (Fig. 3), we cannot predict precisely how this will occur for the reasons given above. We can, however, say that the effects of compressibility are much more significant in the chain-length dependence of all three series than they are in the density dependence, the latter being hardly affected by end-group effects. This is reasonable on the basis that the density can be regarded essentially as a geometrical packing phenomenon, whereas the compressibility will have two components, intramolecular, associated with the flexibility of the covalent bonding chain, and intermolecular, associated with the flexibilities of a variety of intermolecular interactions.
In summary, these data provide a unique and valuable insight into chain–chain and chain–end-group interactions in liquids, an insight which has hitherto been unavailable.
We have already used these data to determine molecular weight on-line during polymerisation interactions and have shown that the technique can easily detect the change from cyclic siloxane to linear polymer in ring-opening polymerisation to cylic polymerisation during the polymerisations of silicones. We intend to extend this work to higher members of the 1-alcohol series by working at 60°C, to confirm that the effect of the hydroxyl end-group diminishes as carbon number increases. We also intend to compare these data with molecular-modelling predictions of intrinsic repeat unit compressibility in an attempt to deduce the magnitude of the intermolecular effects.
There are very few methods available which are sensitive to the interactions in fluids between end units, chain repeat units and their nearest neighbours and we envisage that the ultrasound velocity method, with its ease of use and in-line capability, can provide a powerful new method of characterising polymeric fluids, as well as contributing to the understanding of fluid compressibility and of the propagation of sound in terms of specific intermolecular and intramolecular physico-chemical behavior. The weaker molecular forces involved are difficult to probe by other methods and we hope that this method will offer new perspectives on the molecular basis (molecularics) of the liquid state.
We find that the compressibility of the methylene unit in the n-alkanes is −1.01±0.01 10−9 Pa−1, in 1-alcohol −0.34±0.01 10−9 Pa−1 and in pmds the compressibility of the siloxane unit is −0.31±0.003 10−9 Pa−1.
This journal is © the Owner Societies 2003 |