On the second day of WATSURF 2013, the topic of the day was nominally "Water on solid substrates".
Alan Soper (ISIS, UK) talked about "The structure of water in bulk and in confinement by neutron and x-ray scattering". He explained a lot of the mechanics of measuring structure factors (F(Q)) in neutron scattering experiments. One part of the talk was how to get pair correlation functions (g(r)) from structure factors. Historically, you would try to IFT F(Q), but you don't have the full Q range and you have to make an estimate of the self-scattering background, so doing this directly is very error-prone and the resulting g(r) are full of truncation ripples [and in the case of multi-component mixtures, you still need to disentangle partial pair-correlation function of each pair of components]. More complications: neutron scattering (especially) and XRD (a bit) are not really elastic, so there isn't an exact 1-1 mapping between momentum transfer Q and scattering angle theta. Of course, van Hove (1954) wrote down exactly how to handle inelastic scattering if you can record both scattering angle *and* energy. Unfortunately, no experiment actually gives you d^2 sigma / dOmega depsilon with any accuracy (and getting it inaccurately would take months). Moreover, looking at either constant scattering angle or constant neutron time-of-flight gives you an integral of d^2 sigma / dOmega depsilon over an extremely inconvenient curve in (Omega, epsilon) space. Not all is lost. Placzek (1952) showed that int_(const Q) epsilon S_d(Q, eps) d eps = 0 (S_d is distinct scattering factor), whereas int_(const Q) eps S_s(Q, eps) d eps hbar^2 Q^2 / 2 M (S_s is self-scattering factor), so you have some clue to disentangle distinct and self scattering.
Soper's preferred method of extracting information from scattering data is the Empirical Potential Structure Refinement (EPSR) method. For a given model interaction potential, use MC to measure g(r) in simulation. Details: harmonic constraints define molecules, use existing "reference" potential to generate g(r), perturb potential to make simulated structure factor look more like measured data. Details at http://disordmat.moonfruit.com, or PRB 72, 104204 (2005). Summary of results: very good agreement, most of the low-Q discrepancies between the experiments and simulations are due to inaccurate subtraction of self-scattering intensities in the data. Peak of g_OO(r) in water at STP, derived from all the X-ray and neutron data, is around 2.4 (very similar to Arten & Levy data from the 1970s).
Looking beyond g(r), you can imagine a g(r, Omega) ("a spatial density function"). Technically, done by projecting components of g(r, Omega) onto spherical harmonics, then reconstruct g(r, Omega) assuming components for high moments are zero (i.e., hard to measure accurately). The result as a function of larger and larger isosurfaces. First, large density of waters opposite OH bonds. Next, much lower density around where the lone pairs would be. As you draw larger and larger isosurfaces, the lone-pair lobes eventually join into a single lobe, while the waters opposite the OH bonds are still well separated. If you continue onto larger isosurfaces, you start seeing "interstitial waters" in the first hydration shell (lobes opposite H's still well defined). As you pressurize cold water (268K) up to 2 kbar, the interstitial shell comes in closer to the central O, the lobes due to H-bonding are not very perturbed. (Mancinelli et al, PCCP 07 (2007))
To summarize, Soper has produced a nice review recently (2013), titled something like "is there anything we can say for sure about the structure of water?" He presents something close to "the definitive data set for g_OO, g_OH and g_HH".
What about water in confinement? In MCM41 (hexagonal array of cylindrical pores of amorphous silica), Soper reviewed the claims made about supercooled water in confinement [cite DL about what may be going on]: a so-called "fragile-to-strong transition" upon supercooling, and existence of a density minimum. What do the experiments actually say? Scattering of dry MCM41 already has a lot of information, so you have to see differences between dry and wet MCM41. Experimentally, when you add D2O vs H2O to dry MCM, the (100) peak of scattering intensity is about a factor of 4 smaller in D2O. But interpreting peak heights as changes in material densities is not straightforward. More difficulties: no two samples of MCM41 are the same (100 peak positions and heights differ from experiment to experiment). Claimed literature values for amount of water absorbed (e.g. 0.4-0.5 g / g) are inconsistent with small pore size obtained by gas adsorption measurements (reported ~7.5 A, but must be at least ~15-20 A). Soper is doing EPSR analysis of water-in-MCM41, in a hexagonal box of unit cell size 33.1 A containing a pore of about 25 A radius. The simulated pore is ~148 A long. For dry material, 25 A (not 23 A or 21 A) pore radius gives best agreement to data. Wet MCM41 at 298K shows good (not perfect) agreement between simulation and data. At 210 K, with no change in density in the simulation, reproduces the experimental data quite well (=> heights in density peaks don't really indicate density changes). They have plots of coordination numbers of waters as function of distance from pore walls: significantly below bulk, even at center of pore. O-O-O angle distributions change upon supercooling, more tetrahedrality around the intermediate region between the pore center and walls, but drastic dropoff of tetrahedrality towards pore walls. Conclusion: "when someone shows you results about water structure in confinment, make sure they show you both the experimental scattering data *AND* a detailed model that explains the data: inferring conclusions just by "looking" at the data is very unreliable."
Alenka Luzar (Virginia, US) asked about what is "the definitive water model that reproduces the data". Soper was reluctant to put his foot down (none fits the data perfectly), but suggested TIP4P-2005 is one of the better ones available nowadays. Potentials that put charges on O atom (e.g. SPC/E) instantly make to peak in g(r) too high.
Jan Swenson (Chalmers, Sweden) asked a question alluding to whether confined water could just be thought of as bulk water that doesn't freeze. Soper came down firmly on the side that confined water is so perturbed that most phenomena on it are due to surface perturbations. Swenson also commented that many people nowadays no longer think there is a fragile-to-strong transition in confined supercooled water, even though there's lots of older papers that make that claim. Soper made a comment about how scary it would be to base your entire career on the idea that confined water is just like bulk water that doesn't freeze, and use it to make "deep" statements about bulk water.
The next talk was by Toshio Yamaguchi (Fukuoka University, Japan), "Thermal behaviour, structure and dynamics of low-temperature water confined in mesoporous materials". He planned to talk about water confined in three different environments: (a) a hydrophilic one (MCM41); (b) a hybrid hydrophilic-hydrophobic one (mesoporous organosilica) and (c) a hydrophobic one (ordered mesoporous carbon) [however, ran out of time for the last part]. MCM41 is produced by templating silica onto surfactants that have self-assembled into hexagonal arrays of cylinders. If you use C10, C14 or C18 (C_n = alkyl group with n carbons) as surfactants, you get a pore size of 21 A, 28 A and 37 A, respectively. In MCM41 made with C10, differential scanning calorimetry shows no phase transition in C10 MCM41 (but at ~220 K for C14 MCM41 and at ~240 K, there is a freezing peak in C_P). At least this is the initial observation, but he claims that if you do things very very carefully, you *do* see a peak in DSC for C10 MCM41.
MCM41 C10 has uniform and independent cyl. channels of 2.1 nm in diameter, large surface area (~ 1096 m^2 /g) and an interior surface that is hydrophilic (3 OH groups per 1 nm^2). When adsorbing water to dry MCM41 through gas adsorpotion/desorption (increasing vapour pressure), first get a mono-layer of water (MLW) on the silica surface, then get capillary condensation (CCW) (slight hysterisis). In old work (JPCB, 2000), they did XRD of MLW and CCW in MCM41 C14. Structure of MLW is tetrahedral-like, very distorted, and almost independent of water. For CCW, tetrahedral-like structure, distorted compared to bulk water, and tetrahedrality increases with lowering temperature. They've also done some EPSR modeling of their data to get density profiles of water in confinement. For C10 pores, density of water at interfaces doesn't change 296 K -> ~170K, but interior water becomes more layered. They have some spatial density functions (a la Soper) of "HDL" and "LDL", which shows that HDL has same interstitial waters as bulk water, but LDL has no such interstitial waters.
Second part of the talk: PMO (Periodic Mesoporous Organosilica). Same synthesis as MCM41, but embed organic groups in silica matrix to make partially hydrophobic wall. There's lots of freedom to tune organic groups. Adsorption curves for these materials with only 8% of the surface sites decorated with hydrophobic groups results in much less adsorption (0.1 g/g) than in ordinary MCM41 (0.5 g/g). The results of XRD and neutron scattering went by me too quickly to process.
Jan Swenson (Chalmers, Sweden) commented on the liquid-liquid transition: they have studied this possibility, but in clays (will talk later in the meeting). The advantage of using clay is that if the density of water changes, the layer-to-layer spacing of clays changes. They only ever see gradual, continuous changes, consistent with slight thermal expansion of clay. Argues against density changes around T_LLCP.
After getting lost walking towards the village of Les Houches (and failing to get there in less than an hour!), the afternoon session got started with a talk by Lars Pettersson (Stockholm, Sweden), titled "Fluctuations in Ambient Water". The title slide had two density isosurfaces of water at 253 K showing that regions of "high density" and "high tetrahedrality" are anticorrelated there. He summarised the usual story of anomalous properties of water and enhancement of response functions around "Widom line". The main things he has to say are: a) X-ray adsorption spectroscopy suggests that every water molecule has only slightly more than one donating H-bond on average (Wernet et al Science 304 (2004) 995); b) X-ray emission spectroscopt suggests two motifs: strongly tetrahedral ("LDL") and very disordered ("HDL") (Tokushima et al, Chem. Pys. Lett. 460 (1008) 387; c) SAXS shows density fluctuations enhanced upon cooling (Huang et al PNAS 106, 15214 (2009), JCP 133, 134504 (2010)); d) XRD shows continuous transition on supercooling to T ~ 223 K (from LCLS, Huang et al submitted).
I've seen most of this talk before, but they've done some recent work that's worth thinking about. First, they compute S(Q) to low Q for a "simple" liquid (ethanol) to show that the rise in S(low Q) that they reported in PNAS 2009 for water does not occur for a simple liquid. Second, they simulated water at its "first" critical point and did their S(Q) Ornstein-Zernike analysis to determine the "growing correlation length" there. They get xi = 5.9 A, whereas if you look at the structures by eye, the more natural size of gas-like or liquid-like domains is ~ 30 A. So they argue that even though their analysis for supercooled water measures a "small" correlation length, the "actual" correlation length could be bigger. They've also shown that the S(Q) rise is sort-of reproduced for TIP4P/2005. Finally they've shown that if they apply their "local-structure index" characterisation of the local structure of water molecules to the *inherent* structure of water, molecules clearly divide into two populations. Something to chew on...
L. Bove (ILL, Grenoble?) talked about "Salty water under pressure". The received wisdom is that salt is excluded from "open" structures of ice (e.g., ice Ih, LDA). Instead, they have experimental evidence that there exists a "salty ice VII", realized by taking salty water (H2O + LiCl) through an amorphous phase first (S. Koltz, L. E. Bove, et al, Nature Materials 2009). The ions do not have long-range positional order.
G. Cassone (joint PhD student at Messina + Paris IPMC) talked about "Ab Initio of proton conduction mechanism in ice (I_h and Ice XI)". Under a strong (~ 0.2 - 0.3 V / A) E-field, sustained flow of protons through ice I_h and Ice XI.
OK, that's the latest from Les Houches. My talk is tomorrow morning, and then the meeting becomes a bit more relaxing from my end.
No comments:
Post a Comment