Reactive scattering as a chemically specific analytical probe of liquid

The arguments outlined in section 3.4.1 point strongly to a predominantly direct reaction mechanism. Consequently, it can safely be assumed that the reaction observed is occurring at the extreme outer layers of the liquid surface, as any trapped OH would be expected to react further to form H2O, given the modest reaction barriers for this

reaction. The direct reaction at the interface can therefore be used as a new type of analytical tool to characterise the ionic liquid surface. It can be assumed, based on reaction barrier heights (section 3.4.1), that the O(3P) atoms react exclusively at the secondary CH units. As such, the O(3P) atoms act as a site-specific probe of the interface, providing information on the extent to which the interface is occupied by the Cn alkyl chain of the cation.

The observation of OH is, in itself, evidence of the presence of alkyl units at the interface. Figure 3.4 clearly shows substantially different OH yields from the family of ionic liquids. The integrated appearance profiles are proportional to the OH yield. The differences in OH yield, relative to the reference liquid squalane (table 3.1) are too extreme to be accounted for by differences in molecular stoichiometry. This is emphasised in figure 3.13 where the OH yield per reactive hydrogen atom is plotted. This was calculated by dividing the area under the appearance profile (proportional to amount of OH formed) by the number of ‘reactive’ hydrogen atoms present in the liquid. In this work, only the secondary CH units on the Cn alkyl chain on the cation

were assumed to be reactive.

A purely stoichiometric dependence would correspond to a horizontal line on figure 3.13. Instead, there is a dramatic, non-linear chain length dependence with a significant enhancement in OH yield with increasing chain length. The results show that [emim][NTf2] is the least reactive towards O(3P). In the context of the applicability of

this family of ILs being used in thruster systems for spacecraft it is clear that, all other factors being equal, [emim][NTf2] would indeed be the best candidate of those studied.

It can also be predicted, on the basis of these results that the shortest member of the family, [mmim][NTf2], which was not included in this work, would be a good candidate

also. From a defence point-of-view it would be important to establish which ILs used might produce an OH plume which could be detected by an enemy. Conversely, it

would be useful to detect the presence of an enemy’s satellites by knowing that an OH plume might be suggestive of ILs being used in their thrusters.

Included in figure 3.13 is the reactivity per reactive hydrogen observed in the complementary molecular beam experiments of Minton and co-workers[125]. The reactivity has been normalised with respect to the peak signal from [C12mim][NTf2] in

both the molecular beam (Minton) and photolytic (this work) experiments to allow for direct comparison. In the molecular beam case, the terminal CH3 groups are also

presumed to be reactive as a result of their increased collision energy (~520 kJ mol-1); the barrier to abstraction for a primary C-H is ~42 kJ mol-1[47].

In this case, there is clearly still a distinct, non-stoichiometric variation in the OH yield for the more restricted measurements on the first and last members of the series ([emim][NTf2] and [C12mim][NTf2]). There is agreement therefore between the two

studies that the availability of the aliphatic chains at the interface increases dramatically with chain length. There is a less extreme suppression of reactivity towards [emim][NTf2] in the molecular beam experiments (figure 3.13). This suggests that the

higher energy probe (520 kJ mol-1) used in the molecular beam experiments samples a different range of depths than the lower energy probe (16 kJ mol-1) used in the photolytic experiments presented here. In the higher energy case, it is more likely that OH formed at greater depth will be able to escape before reacting to form water on account of it having sufficient energy to overcome the attractive projectile-surface interactions.

The conclusions of both experimental scattering studies are further strengthened by the molecular dynamics simulations, carried out by Schatz and co-workers, co-reported with the Minton experimental results[125]. The simulations predict little preference for the ethyl groups in [emim][NTf2] to occupy the surface. For the [C12mim][NTf2]

surface on the other hand, there is a clear selectivity towards the alkyl groups preferentially occupying the interface. This is shown in the representative snapshots of the [emim][NTf2] and [C12mim][NTf2] surfaces shown in figure 3.14. The final six

0 2 4 6 8 10 12 0.0 0.2 0.4 0.6 0.8 1.0 OH y ie ld p er a vaila bl e hydr og en a to m

Number of carbon atoms in 1-alkyl chain

Figure 3.13: Relative OH yield per available hydrogen as a function of alkyl chain length for a

series of [Cnmim][NTf2] ionic liquids as measured by photolytic-spectroscopic experiments

(blue open circles with 2σ error bars) and the complimentary molecular beam approach (red closed squares)[125]. Yields per atom are normalised for n = 12 for both experiments. The available hydrogen atoms are assumed to be restricted to CH2 groups in the photolytic

experiments but include CH3 and CH2 groups in the molecular beam experiments. This reflects

the relationship between collision energies and barrier heights for the different C-H bond types.

The authors[125] go further and predict that the alkyl chains (or the outer 6 carbon atoms) lie, on average, parallel to the plane of the surface. This is consistent with the photolytic experimental results presented here. A chain lying parallel to the surface would expose the secondary C-H units. A chain perpendicular to the surface in a monolayer-type structure would preferentially expose the least reactive primary groups. This might be expected to result in a suppression of reactivity in the photolytic experiments described here. Some clustering of the chains is predicted also for the longer chain IL, at the interface and the bulk (figure 3.14).

It has clearly been established here that there is preferential occupation of the interface by alkyl groups as the chain length is increased. This is in agreement with the majority view from many of the more traditional experimental surface analysis techniques described in the introduction (section 3.1)[160,161,177,178,180] and theoretical studies[166,168,169,175,176]. A discrepancy remains with the direct recoil spectroscopy measurements of Seddon and co-workers[156,157] who predict no ion segregation. This may be a consequence of the photolytic method presented here (and

even the molecular beam) being ‘gentler’ than direct recoil spectroscopy as previously implemented with higher energy particles such as Ar+ and Ne+ ions[156]. There is evidence that there is a critical chain length of n ≥ 4 carbon atoms, beyond which chain segregation is more pronounced [161,179]. The results presented here (figure 3.13) would be consistent with this, although in the absence of data for [C3mim][NTf2] it is

impossible to conclude whether this onset is sudden above n = 3. There is however, clear preferential occupation of the interface by alkyl units above n = 4.

The advantage of using chemical ‘probes’ such as this to interrogate the chemical nature of the interface is that specific target groups can be identified; such as the alkyl segment in this case. A consequence of this is that the absolute surface coverage by the target alkyl groups can also be estimated. The difference in the energy transfer between the O and OH impulsively scattering from the liquid surface (described in section 3.4.1) in the Minton experiments confirms that the [C12mim][NTf2] surface is not saturated with

alkyl groups.

The surface coverage was estimated in this work by comparing the OH yield from each IL to that of the reference liquid hydrocarbon squalane. McKendrick and co-workers previously used molecular dynamics simulations to predict the fraction of the squalane surface at which H-abstraction reactions could take place[121] i.e. the fraction of the surface which is occupied by secondary and tertiary C-H units. They predicted that 60% of the squalane surface was potentially reactive; a fraction which is close to the bulk composition of squalane but with a slight preference for primary CH3 units to

occupy the surface.

The OH yield from each IL relative to squalane is known (table 3.1), and this relative yield as a fraction of the potentially reactive surface coverage of the squalane reference gives an estimated surface coverage (by alkyl CH2 units) of ~35% for [C12mim][NTf2],

~16% for [C8mim][NTf2], ~5% for [C5mim][NTf2] , ~2% for [bmim][NTf2] and ~0.2%

for [emim][NTf2]. These surface coverage estimates agree qualitatively with the

majority of previous experimental and theoretical work (that there is a preference for the alkyl chains to occupy the surface). Although direct quantitative comparison is not possible at this time it would be an interesting and reasonably straightforward quantity to test in future MD simulations.

Figure 3.14: Representative snapshots from molecular dynamics simulations of the surfaces (at

323 K) of [emim][NTf2] (top) and [C12mim][NTf2] (bottom) ionic liquids, showing the anion

(green), imidazolium ring (red), alkyl chains (blue) and hydrogen atoms (white). Reproduced from reference[125].