Prof. Moffatopened the discussion of Dr Martlew's paper: The Roman and Syrians made superb glass two millennia ago—but not window glass. Who is the “Faraday” of window glass making? When was the modern process invented? Where? What were the key novel features?Dr Martlewreplied: It may have been the expansion of the Roman Empire to these chilly islands which provoked the Roman glassmakers to develop methods of creating glass sheets to block up apertures in buildings. Casting molten glass onto a flat rock slab then rolling it out into a sheet was a technique used by the Romans, to create a translucent rather than transparent pane, appropriate to the glare of the Mediterranean lands!Glass blowing was invented at some time about the beginning of the Christian Era, and naturally provided an easy way to make objects of spherical or cylindrical symmetry or derivatives therefrom. Blowing a glass sphere then flattening it to make a window was quite challenging, and three methods developed. One involved manipulating the sphere to make it into a cylinder, which could then be split axially and flattened to make a flat sheet. A second method (which never became really popular) involved blowing the glass into a suitable mould to create a bottle with four flat sides and a flat base; one could then cut these up to create small flat sheets of glass. The third method, known in England as the Crown Glass process, used the centrifugal forces generated by spinning a sphere on its supporting rod to open the hot glass shape into a flat disc from which small window panes could be cut. The individual inventors of these techniques are not known.By the middle of the twentieth century continuous glassmaking on an industrial scale had replaced these ancient hand methods of production. Two main kinds of process were used. If cheap window glass was needed it could be drawn vertically upwards from a tank of molten glass in a continuous ribbon, being cooled and annealed in a tower before being cut into handlable sheets at the top. Glass thickness was limited to about 4 mm, but the product answered domestic needs very well, aside from the optical distortion generated in the forming process.Thicker glass with better optical characteristics was in demand for shop windows and the large windows and mirrors desired by the affluent classes. Polished plate glass to satisfy this demand was made by continuously rolling the stream of molten glass to create a ribbon as thick as needed. To remove the surface scarring introduced by the rolling process it was then necessary to grind and polish both sides of the glass to create optically perfect plates, a tremendously expensive process, and very wasteful of glass.When the float process was announced in 1959 it was revolutionary in that it amalgamated the strengths of each of the earlier processes. The perfect flatness of polished plate glass was achieved by floating the molten glass on a layer of molten tin. Because the new ribbon didn't touch anything solid whilst it was still deformable, the as-formed surfaces were as brilliant as those of sheet glass drawn up from the melt.We may not know the individual inventors of antiquity, but we do know that the “Michael Faraday” of industrial window glass production was undoubtedly Sir Alastair Pilkington.Prof. Hounslowasked: You show particle tracks, yet your velocity maps indicate a great deal of mixing. Would there not be a distribution of residence times?Dr Martlewanswered: There is indeed a very wide spread of residence times over all possible particle tracks. Mixing does occur, but with the generally very high viscosities of the molten glass and the very low Reynolds numbers involved, mixing tends to be slow and relies heavily on chemical diffusion. Homogenising the glass and avoiding the visible effects of non-uniformity of refractive index in the product has exercised glassmakers for several thousand years.For the present work it has been necessary to prepare several thousand particle tracks for each case, to establish which tracks within the furnace result in the smallest residence times. Dissolution of sand grain residues is favoured by long times and higher temperatures. Generally the particle tracks which have the smallest times yield the worst dissolution efficiency; temperatures are not sufficiently high along these tracks to compensate. For the exploratory research presented here the method of calculation is sufficiently cumbersome to make it necessary to limit the number of tracks studied to a very few. Having established the method and demonstrated some usefulness, the next step is to code the calculation within the 3DGLASS post processor module, so that more complete explorations can be carried out economically.Dr Colesaid: When you add impurities to the glass in order to generate a particular colour of glass, how will this affect the reference data that you have regarding the stability/quality of the glass relating to composition? Do you thus obtain glasses of certain colours being more stable/fragile,etc.than clear glass?Dr Martlewreplied: We have talked already about the corrosive nature of molten glass. Though prejudicial to the longevity of the capital equipment needed to make glass products industrially, this attribute enables us to explore many different compositional variants in the quest for desirable properties and performances.Colouring oxides (notably those of the transition elements) dissolve fairly readily in the molten glass, unconstrained by any rigid stoichiometry. Ligand field effects then do the rest, creating the colour. More often than not, commercial colours require only very small percentages of colouring oxides, so the key macroscopic physical properties are not affected.Prof. Helliwellasked: I would like to make a connection with Greaveset al.'s paper at thisFaraday Discussionand ask how the new characterisation details are informing new or better products. Also are the SR central facilities currently available, adequate, or needing further development to meet your needs as an industrialist?Prof. Wilsonasked: You mentioned the importance of the surface layer. Could you comment on the use of surface scattering methods to examine the nature of glass surfaces and interfacese.g.neutron or X-ray reflection methods, both specular and off-specular.To follow up—is this important enough to justify obtaining access to these centralised, often expensive, facilities?Dr Martlewanswered: I may not comment in any detail about novel products which may be emerging from better understanding of glass structures, particularly the structure of glasses within the vicinity of the glass surface. However it is well known that the vagaries of the strength of glass products may be rationalised in terms of Griffith flaws (these are envisaged as sub-microscopic flaws in the glass surface which concentrate any tensile stress being experienced, starting a crack and causing failure). Any techniques which can be used to investigate glass structure at the molecular level are very desirable. With any such technique, however, it seems that the larger the scale of examination the more expensive is the experimental facility. My personal view is that we need to be looking at mesoscale structures in order to gain the working understanding that is needed to be the fundamental foundation of new commercial products. In the current economic climate, manufacturing industry cannot justify spending large amounts of private money on equipping for such techniques. If the United Kingdom is to take the lead in these matters, I believe that some way of financing such expensive facilities needs to be undertaken at Government level to support British industry in the increasingly competitive global marketplace. Having created such capital resources, the Government must ensure that the accountancy conventions used to finance their application to industrial exploitation must be friendly to the industrial budgetary processes. Sadly, the pressures now faced by industry in the current climate mean they can no longer be thought of as any kind of “cash cow”!Dr Maycommented: At the ILL (Institut Laue-Langevin), there are no “tickets” to get neutron beam time. You just need to propose an experiment, and you can perform it for free if it is accepted by the relevant subcommittee. The condition is that the results get published. However, one has to pay for secret research.Dr Martlewreplied: Thank you for this helpful comment. The crunch is often the tension between the publication requirement to obtain access to centrally funded capabilities, and the secrecy necessary to protect intellectual property. Defensible IP rights often crucially depend on being able to demonstrate a stance of confidentiality during the early period of the inventive step. Publication of results is directly opposed to this stance.Prof. Finneyopened the discussion of Prof. Greaves' paper: (1) The INEL detector resolution is quite poor, and you are getting a very restricted range inqfrom it atλ = 1.5 Å. Is this adequate to be sure of phase identification from one peak for each of two phases (stuffed quartz and spinel)?(2) Similarly, the detector limits you to only a small part of the Debye–Scherrer rings and hence quantifying amounts of each of the two phases on the basis of integrating under one peak leaves you open to problems of preferred orientation. How can you be sure your quantative phase estimates are not significantly affected by preferred orientation?(3) In Section 4 you say that the proportions of amorphous content are estimated from integrating the amorphous background. (a) How good do you think this estimate is? (b) Do you have the data available? It is not plotted with the other quantities in Fig. 4 of the paper. (c) Can the variation of the “amorphous content” give a useful constraint on the accuracy of the other quantities (e.g.by some kind of sum rule of total material quantity)?Prof. Greavesanswered: (1) The identification of phases were first made throughex situmeasurements using standard high resolution powder diffraction with a fullqrange. The INEL system on the station 8.2 at the Synchrotron Radiation Source was then used to providein situfingerprints. It is worth reporting that in the last few weeks the RAPID II curved multi-wire proportion counter has been commissioned at the SRS on station 6.2. This has a much-improved resolution compared to the INEL, a count rate three decades higher and an overlap geometry with respect to the SAXS detector offering a fullqrange to the smallest angles. In future it will be possible to identify phasesin situas well as to track their development as a function of time and temperature.(2) Of course we cannot be sure. However, in recent measurements at the ESRF on BM26B we found that the intensity of the quartz pattern was extremely erratic compared to that of spinel, strongly supporting the view that nucleation of stuffed quartz occurs at the surface and for monolithic specimens is prone to preferred orientation. Regarding the quantification of phases at each stage, we have not used the intensity of diffraction lines but have relied on the good agreement between Cr EXAFS and SAXS data (Porod regime plus Invariant) to determine the crystalline fraction,v.(3) We have chosen here not to useIamorph(t) to square withInlm(t) from the two phases present because the composition of the glass is changing. In the past, though, we have explored the sum rule approach in crystallisation from gels, for example, where the crystalline phase totally replaces the amorphous one or in amorphisation where the reverse occurs (e.g.refs. 1 and 2).1 J. C. Fernandes, D. A. Hall and G. N. Greaves,Mater. Sci. Forum, 1996,228–231, 411–416.2 G. N. Greaves, inFrontiers of High Pressure Research II: Application of High Pressure to Low Dimensional Novel Electronic Materials, NATO Advanced Research Workshop, Kluwer, Dordrecht, 2001, p. 53.Dr Sankarcommented: The main limitation in the angular range is primarily due to thein situhigh temperature facility which has a small window region that permits only up toca.60° 2&thetas;. In addition only a few strong reflections appear in this 2&thetas;range.It appears from Fig. 4 (top) of your paper that the stuffed quartz phase is stable. It has been observed in other cordierite-forming systems, for example Mg2+-exchanged zeolite B, that the stuffed quartz phase appears above 900 °C and disappears as soon as the cordierite phase starts to appear. It appears from these studies that the stuffed quartz phase is unstable, whereas this phase appears to be stable in this investigation. Is that due to the presence of chromium?Prof. Greavesresponded: Our view is that the stuffed quartz phase here is not nucleated from Cr at these temperatures but from defects at the glass surface, which will therefore offer some physical stability. However, with accurate unit cell parameters, the composition of stuffed quartz can be determined from the tabulations of Schreyer and Schairer (ref. 12 of the paper) and may not always coincide with stoichiometric cordierite Mg2Al4Si5O18. From additional unpublished data the stuffed quartz composition in Cr-doped cordierite glass heat treated at higher temperatures than reported falls on the quartz-rich side of cordierite. This may well be because of the twist in the residual glass composition resulting from the earlier formation of spinel (see eqn. (8) of the paper). Also, energy dispersive XRD results on powders of Cr-doped cordierite glass have shown that for isothermal crystallisation at 1250 °C, 200 °C above the temperatures employed here, the stuffed quartz phase is eventually replaced by cordierite, but that this conversion is slow, suggesting that the stability of stuffed quartz relates to the fact that it does not have the composition of cordierite. In the example of Mg-exchanged zeolite B, as far as I understand, the composition of the stuffed quartz phase is closer to Mg2Al4Si5O18which may explain why the final conversion to cordierite is more rapid and your observation that the stuffed quartz in this case is unstable.Prof. Wilsonsaid: I am interested in the fact that the time profiles for evolution of the various types of scattering are similar, indicating that equilibration occurs on all length scales simultaneously. That is, are the formation of 210 Å particles and the formation of crystalline phases governed by the kinetics of a single process?Prof. Greavesreplied: Yes, insofar as the spinel phase is concerned, as this almost completely dominates the SAXS. If the composition determined from the final Cr EXAFS spectrum, MgCr0.18Al1.82O4, is the same at nucleation, then the single process is the one described by eqn. (8) in the paper. This of course raises the question as to whether the changing glass composition remains uniform on all length scales as growth advances.Prof. Ryanasked: The data presented in the paper are truly beautiful and show scattering patterns that look like the form factor of a sphere. The paper indicates that the phase that is growing has a different composition to the glass and that the scattering is dominated by the local Cr concentration. Detailed analysis of the scattering is required, however, as the nucleation and growth process has a dense particle surrounded by a depletion layer, the shape of the depletion layer depending on the thermodynamic driving force for crystallisation and the diffusion coefficients of the components. This effect is not so important at the beginning of the crystallisation and when the reservoir of Cr is exhausted the equilibrium structure could well be dense spherical structures in a uniform background. It is in the intermediate stages (Fig. 1) that a more complex electron density profile needs to be considered.Plot of electron density throughout the growth process.Prof. Greavesanswered: I agree, we have just looked at the start and finish. With improved data coming from the ESRF BM26B and in due course from the SRS 6.2, we can look forward to the possibility of exploring the complete progression from nucleation to full devitrification of Cr.Prof. Hounslowsaid: (1) You indicate in the paper that the scaling of radius of gyration with time indicates a diffusion limited process, but that at long times depletion of free Cr slows the process. Does the overall balance on Cr indicate that the amount of free Cr decreases to zero?(2) For a process with monodisperse particles whose growth is limited by diffusion, is it not surprising that no Ostwald ripening is seen?Prof. Greavesreplied: (1) TheR2gvs.tplot is very sensitive to errors inRg. Nevertheless the initial linear rise is clear, as is the levelling off around the stage at which the intensity ofIspinellines are beginning to saturate. The evidence for the eventual removal of Cr from the glass matrix comes from the composition of the spinel and from the agreement of the crystalline fraction,v, from eqn. (8) in the paper with that obtained from SAXSviaeqn. (7) in the paper.(2) Absolutely, but it needs stressing that the greatest degree of monodispersion occurs part way through the heterogeneous growth process,i.e.before all the Cr is crystallised and probably at the point at whichR2gvs.tceases to be linear.Prof. Ryancommented: The formation of monodisperse particles during thermally induced phase separation of a mixture of two polymers has been observed.1Mixtures of polyisoprene and ethylene–propylene copolymer were studied using time-resolved elastic light scattering. For off-critical quenches highly monodisperse spheres were observed whose radii grow with a power law. The monodispersity and growth law are rationalised as a heterogeneous nucleation process in a similar manner to the paper under discussion.1 Cumming, Andrew, Wiltzius, Pierre, Bates and S. Frank,Phys. Rev. Lett., 1990,65(7), 863–866.Prof. Helliwellasked: How has this detailed structural characterisation improved the knowledge of the function of glass?Prof. Greavesanswered: The traditional functions of glasses and ceramics, many with ancient pedigree, have generally developed from the craft sciences. What combinedin situstructural methods, like those presented here, can now provide is direct observation of the development of high temperature structural chemistry as the optimum conditions that result in a particular function are reached. With serendipity new functions may emerge. For example, fine glass ceramics have been developed for their optical transparency, mechanical and thermal properties, for which monodispersion is not critical. The discovery of monodispersed nanocrystals in Cr-doped cordierite glass, however, promises a bulk quantum dot system with the function of non-linear optical response mouldable into any shape.Dr Dentopened the discussion of Dr Chen's paper: What are the differences between the electrochemically generated species and the laser induced species?If the edge data are the same, then how can the EXAFS be different, especially given the signal to noise limitation?Dr Chenreplied: (1) Electrochemically generated Cu(ii) species from the starting Cu(i)(dmp)2+species had one electron removed from the Cu(ii), so the total number of electrons in this species is one less compared to the starting material. The laser excited Cu(i)(dmp)2+, however, underwent an intramolecular charge separation process where one electron was transferred from Cu(i) to one of the ligands, forming the Cu(ii) species. Therefore, the total number of electrons was unchanged. In fact, the metal–ligand charge transfer (MLCT) transition is a reversible process. As the excited state decayed, the intramolecular charge separation recombined.(2) The edge data for the MLCT state and the electrochemically generated Cu(ii) species are almost identical, indicating the generation of Cu(ii) species by the laser. However, the edge data for the ground state and the laser excited state are not the same, as indicated by Fig. 5 in the paper. The EXAFS data in Fig. 6 of the paper which we were comparing were the ground state Cu(i)(dmp)2+and laser excited mixture with 80% ground state and 20% MLCT state. They are different and from these differences, we extracted the MLCT structure.Prof. Coppensasked: (1) With acetonitrile you get a shortening on the Cu–N bonds on excitation, while in toluene you get a lengthening. Is it possible that you are looking at the complex before exciplex formation, as we calculate the 4-coordinate Cu(ii) excited state to have a C–N bond length shortening of 0.04 Å?(2) Do you identify the very short-time species visible in the fast laser experiment with the excited state before exciplex formation?Dr Chenanswered: (1) The process of thermal equilibration of the MLCT state takes place on a subpicosecond to a few picosecond timescale. Therefore, it would be too fast for current 100 ps X-ray pulses to probe.(2) In a recent fs pump–probe laser transient spectroscopic study, we observed two very short components in the kinetic decay trace in addition to the ns longer lifetimes commonly referred to in the literature. We measured transient spectra of Cu(dmp)2+in acetonitrile and in ethylene glycol, where both measurements gave a subpicosecond rise time that was longer than our instrumental response time, and a few picoseconds fast decay time. We are in the process of further investigating the origins of these components in the kinetics of the MLCT state. It is likely that they may be related to generation of the thermally equilibrated MLCT state and exciplex formation.Prof. Evanssaid: (1) This is a very impressive experiment, but the intrinsic problem is the 2 shell fit for the 2 Cu–N sites. Have you examined the correlation between these 2 shells to make a good error estimate?(2) It seems surprising that an expansion is observed in the Cu–N distance. The MLCT process would remove an electron from a t2M–L σ* orbital in Cu(i) and transfer it into a phen π* ligand. So the Cu(ii) transient centre would be expected to have a smaller covalent radius.(3) The BArF anion is an extremely weak binder, with the aryl fluorines the only plausible donors. Acetonitrile, or a linear triatomic ligand, will provide a distinct multiple scattering fingerprint to make it identifiable if it is the fifth ligand.Dr Chenreplied: (1) Yes. In our data analysis, we assumed two Cu–N distances that were associated with the ground state and the MLCT state respectively. From the transient optical absorption experiment and the model calculation with experimental parameters, such as laser pulse energy, sample concentration, sample dimension,etc., we obtained the fraction of the excited state at around 20%. Therefore, in data fitting, we fixed the ratio of the ground state and excited state to 80%vs.20%. TheE0edge shifts in the fittings for the two Cu–N bond distances were kept the same. The difference in Cu–N bond lengths resulted from such fittings with the above precautions.(2) That's a very good point. However, the coordination of Cu also changed from tetrahedral to penta-coordinated. I would expect the steric hindrance could force the Cu–N bond to be longer despite the effects that you mentioned.(3) We are not sure what the fifth ligand was at this point. In the acetonitrile case, you raised a good point.Dr Techertasked: Despite the fact that triplet annihilation processes quench the population of the triplet state (and therefore the EXAFS difference signal), how would this annihilation process change the structure of the energy-transferring moieties and therefore the EXAFS signal?Dr Chenanswered: The triplet–triplet annihilation due to the adjacent molecules being simultaneously excited will certainly quench the excited state population and shorten the lifetime. Thus, the laser excitation will become less efficient and laser photons will be wasted. More importantly, the excited state annihilation will change fundamental aspects of the photoexcitation of the molecules as the interactions between the same kind of molecules can no longer be neglected. The triplet–triplet annihilation could also result in singlet states at a much later time than the initial photoexcitation. Therefore, it is the fundamental aspects of photochemistry that have been changed due to the strongly interacting molecules. Of course, the XAFS signals expected will no longer be limited to those from isolated molecules due to their photochemical reactions, but come from a mixture of isolated and strongly interacting molecules. Therefore, we can no longer claim the signals are from excited molecules of a certain kind, but from a collection of molecules with a certain configuration.Dr Sagicommented: Doing a single energy experiment may reveal some correlation between the structural kinetics and the overall kinetics of the system, thus proving that one can identify distinct features in the single energy experiments on EXAFS or edge regions that are directly associated with the structure, for example the transition from one coordination number to another during the excitation process.Dr Chenreplied: In small molecules rather than proteins, the actual atomic movements after the photoexcitation take place in fs or a few ps. After that, the excited state is thermally equilibrated. Therefore, strictly speaking, our experiment is not time resolved, but provides snapshots of the excited state. However, there are circumstances where different excited states could be generated sequentially, such as a singlet state initially generated by photoexcitation, and a triplet generated by intersystem crossing from the initially generated singlet. In this case, we will have to adjust the time delays corresponding to the optimal population of the singlet and the triplet states respectively in order to capture their structures. If this is successful, we will be able to associate certain distinctive spectral features that appear at a certain time after the photoexcitation to particular excited states.Prof. Helliwellasked: How often does the Advanced Photon Source (APS) offer the “special timing mode”? Is it a limiting factor?Also the APS operates a “top up mode”. How do you allow for that in your experimental design?Dr Chenresponded: APS provides the hybrid fill timing mode about 4 weeks per year. It was split into about one week at a time. So we have the special timing mode every three months. The “top-up” mode from the APS did not affect our experiments, because we average over a long time, and fluctuation caused by the top-up was not obvious to us. We like this mode because it provides 30% more X-ray photon flux.Prof. Coppensopened the discussion of Prof. O'Hare's paper: We have some experience with hydrothermal synthesis in crystal engineering, where one works with multicomponent (>2) systems. Since we cannot predict which phases will be formed, the stoichiometry, and therefore the products change with time. Have you looked at cases where completely different phases are formed as the cooling process proceeds?Prof. O'Harereplied: We have not observed any changes in crystalline phase composition of these reactions on cooling.Prof. Wilsonasked: What is the limitation in time resolution for the neutron and X-ray experiments? How well is this matched to “typical” reaction rates in your processes,i.e.do you need second-scale time resolution?Prof. O'Hareanswered: The current setup using the GEM diffractometer at the ISIS facility gives us a time resolution of minutes while the synchrotron at the SRS gives us a time resolution of seconds. For most hydrothermal reactions and conventional solid–solid processes this is adequate. For other reactions we would ideally need better time resolution.Prof Evanssaid: Have you been able to investigate the solution phase pre-nucleation processes by other techniques, say by high temperature NMR?Prof. O'Harereplied: We have not. Prof. Taudelle and co-workers at the University of Strasbourg have recently usedin situNMR to complement our studies on gallium and aluminium phosphate crystallisation.Prof. Ryanasked: What evidence is there for precursor structures in hydrothermal synthesis?Prof. O'Hareresponded: We cannot detect precursor structures in our experiment. We need crystalline domains which can diffract X-rays.In situNMR experiments by Taudelleet al.suggest the existence of four-membered Al2P2rings in solution prior to crystallisation of aluminium phosphates.Mr Robbsaid: You seem to follow the structure by Bragg peak evolution. Is the data you collect sufficiently good for complete structure refinement and, for example, can you obtain atomic displacement parameters for your material and what does this tell you?Prof. O'Harereplied: The energy dispersive X-day diffraction data we measure is not suitable for complete structure refinement. Different X-ray photon energies are absorbed from the incident and diffracted white X-ray beam by the cell materials and the sample. This means that we cannot define the incident beam profile.Prof. Finneyasked: How much do we know about the variation of conditions within the cell? If there is significant non-uniformity (e.g.in temperature or concentration), is there scope for probing spatially as well as temporally?Prof. O'Harereplied: We try to stir the contents of the hydrothermal cell using a magnetic stirrer which attempts to average out any non-uniformity in temperature or local reagent concentrations.Prof. Cerniksaid: Have you used the information from the three-element SSD to show crystallisation differences within the hydrochemical cell? The intensity overlaps should sample different regions of the crystallite distribution and therefore pinpoint non-homogeneous growth.Following on, do you translate the sample during the experiment to probe preferred regions of synthesis?Prof. O'Hareanswered: The three element detector does allow us to look at differentd-spacing ranges simultaneously. In the systems we have studied using the three element detector where these regions overlap the variation in the intensity of the common Bragg reflections is identical which suggests isotropic crystal growth.Prof. Moffatasked: Might the geophysicist's style of large-volume press that affords simultaneous control of pressure and temperature be of use to your style of experiments?Prof. O'Harereplied: The reagents used in hydrothermal synthesis are quite corrosive viscous gels (acid or alkali) and I wonder whether this would be compatible with their equipment.Prof. Hounslowasked: (1) Fig. 2 of the paper contains a curve fit—of what form?(2) Can you quantify the rates in terms of rate constants and orders of reactions?(3) Are Avrami kinetics appropriate in this case?Prof. O'Hareresponded: (1) The fit is to the Avrami–Erofe'ev equation:α = 1 − exp{−(k(t − t0))n}.(2) Yes, the Avrami–Erofe'ev fit gives us an exponent,n, and a rate constant,k.The value ofn, the Avrami exponent, is believed to contain information about the mechanism of reaction. The model was originally developed to describe the growth of crystallites in solid–solid reactions, and assumes the formation of nucleation sites in a uniform mixture of reagents from which crystal growth occurs.(3) I am not sure; it is a standard analysis but I am not confident that it actually tells us much about the chemistry.Prof. Ryanasked: What is the path length through the cell?Prof. O'Harereplied: 20 mm.Prof. Sir John Meurig Thomasopened the discussion of Prof. Ryan's paper: One wonders whether by changing the conditions (in a well-defined fashion) of precipitation so as to produce aragonite (instead of calcite) you would get the same kind of precursor “phases” being formed.Prof. Ryanreplied: We suspect that a poorly ordered precursor phase could well be found prior to the precipitation or aragonite. We have been able to produce mixed aragonite calcite mixtures under conditions where a precursor phase is observable but cannot distinguish whether there are two precursor phases.Prof. Finneycommented: I would like to add a comment concerning metastable forms occurring in the early stages of crystallisation which then transform to the stable form.The system is the formation of normal hexagonal ice Ih from ice II (a phase stable at higher pressures) recovered to ambient pressure at 77 K.1Neutron powder diffraction on the resulting structure looks like the so-called cubic form Ic, though the resolution of the measurements showed (a) the presence of the 100 hexagonal peak and considerable broadening of the higher angle shoulder of the 111 (cubic) peak. Moreover, various peaks showedhkl-dependent broadening suggesting not only a small particle size of about 160 Å, but also significant stacking faulting. On heating to around 160 K, the structure transforms to normal hexagonal ice, without evidence of any significant disorder.The simple interpretation of these results would be that the ambient pressure ice phase nucleates first as very small crystallites of the cubic phase, which then transforms as the crystallites grow on heating to the normal stable hexagonal structure ice Ih. However, considering that the difference between the cubic and hexagonal structures is that of different stackings of the hexagonal layers (ABCABC stacking for cubic, ABABAB for hexagonal), an explanation that takes account of the varioushkl-dependent broadenings observed suggests that the initial small crystallites may not be identifiable as cubic ice Ih but a structure of random stacking of hexagonal layers. In fact, work on colloidal systems shows similar diffraction features to those observed in the ice case, in particular the simultaneous presence of 111 cubic and 100 hexagonal features, and a high angle shoulder on the 111 peak. Only when the temperature is increased do the molecules have sufficient mobility to begin to anneal out the irregular stackings to approach the stable hexagonal structure.Thus there is evidence in even this apparently simple system for a metastable phase forming in the early stages of crystallisation. In this case, the metastable phase is clearly not amorphous (though such might have occurred earlier but not been detected), but it is, in comparison to the stable phase, clearly disordered. Noting electron microscope observations three decades ago of icosahedral structures formed in the early stages of growth of an even simple crystal (fcc gold), we can perhaps suspect that an initial metastable (with respect to the equilibrium phase) structure may be a common occurrence. In fact, there is perhaps little good reason to expect very small microcrystallites to have the structure of the stable phase, considering the large surface/volume ratio compared to the equilibrium (extended) crystal.1 P. N. Pusey, W. van Megen, P. Bartlett, B. J. Ackerson, J. G. Rarity and S. M. Underwood,Phys. Rev. Lett., 1989,63(25), 2753–2756.Prof. Ryananswered: Thank you for pointing out the relationship to the field of supercooled water and colloidal crystals. The formation of dense, random hexagonally-close-packed structures prior to cubic crystallisation is indeed observed in a wide variety of systems and has been predicted theoretically (see for example refs. 1 and 2).1 I. Kusaka, D. W. Oxtoby and Z.-G. Wang,J. Chem. Phys., 2001,115(15), 6898–6906.2 V. Talanquer and D. W. Oxtoby,J. Chem. Phys., 1998,109(1), 223–227.Prof. Finneyasked: Distinguishing a genuinely amorphous structure from a very small crystal by the presence of clear Bragg peaks can often be controversial: once the “crystallite” is <3–4 unit cells across, the peak broadening may make it difficult to distinguish the scattering from that of an amorphous structure. How can you be sure that your amorphous precursor phase really is genuinely amorphous and not a very small crystalline nucleus?Prof. Ryanreplied: This is indeed a difficult problem and was the matter of considerable debate atFaraday Discussion112.1My colleague Nick Terrill did some modelling, using the scattering geometry of beamline 8.2, and estimated that Bragg peaks could be resolved from a collection of 3 × 3 × 3 unit cells at 1% by volume. Obviously there is considerable line broadening with such small crystals and the wide angle scattering from a metastable, lower-order precursor phase would be even more difficult to resolve. There are some semantic problems in the polymer crystallisation field and the distinction between a collection of chains and a nucleus is often blurred by beliefs.1 A. J. Ryan, J. P. A. Fairclough, N. J. Terrill, P. D. Olmsted and W. C. K. Poon,Faraday Discuss., 1999,112, 13–29.Prof. Evanscommented: The local site symmetry of the carbonate anion differs considerably between calcite and aragonite such that these forms can be differentiated by IR spectroscopy. Perhaps this technique could be used to probe the structure of the pre-crystallisation phase.Prof. Ryananswered: We are currently working on the use of FTIR and FTIR microscopy to differentiate between aragonite and calcite in our crystalliser geometry. We would anticipate probing the structure of the precursor phase by IR or Raman spectroscopy in the not-too-distant future.Prof. O'Hareasked: Have you performed an Avrami-type analysis of the crystallization data described in Fig. 13b of the paper?Prof. Ryanreplied: We have indeed performed an Avrami-type analysis on our inorganic crystallisation data but, as was pointed out by Prof. Hounslow, this is not an appropriate model to use for these crystallisation conditions. I bow to his superior knowledge in this regard and leave it to him to explain the reason why the Avrami model is inappropriate and describe the correct model to use.Dr Chensaid: I am impressed by your results of simultaneous collection of SAXS/WAXS data. As a general question, I would like to ask your perspective on using such techniques on other systems, such as protein folding, crystallisation and molecular self-assembly to form nanostructure materials?Prof. Ryanreplied: Thank you, we have worked hard at developing the SAXS/WAXS instrumentation to such a level that it is widely applicable to a wide range of systems. A good account of the work on protein in solution can be obtained from the Daresbury Annual Reports and in refs. 1–6.1 M. Hirai, H. Iwase, T. Hayakawa, K. Miura and K. Inoue,J. Synchrotron Radiat., 2002,9(4), 202–205.2 W.-Y. Choy, F. A. A. Mulder, K. A. Crowhurst, D. R. Muhandiram, I. S. Millett, S. Doniach, J. D. Forman-Kay and L. E. Kay,J. Mol. Biol., 2002,316(1), 101–112.3 W. Zheng and S. Doniach,J. Mol. Biol., 2002,316(1), 173–187.4 R. Russell, I. S. Millett, S. Doniach and D. Herschlag,Nat. Struct. Biol., 2000,7(5), 367–370.5 D. I. Svergun, G. Zaccai, M. Malfois, R. H. Wade, M. H. J. Koch and F. Kozielski,J. Biol. Chem.2001,276(27), 24 826–24 832.6 D. I. Svergun, M. V. Petoukhov and M. H. J. Koch,Biophys. J., 2001,80(6), 2946–2953.Prof. Helliwellcommented: Regarding the literature on protein assembly, protein crystal nucleationetc.I refer you to the following authors:(a) M. H. J. Koch: assembly of multi-macromolecular structures(b) A. Tardieu: protein crystal nucleation(c) D. I. Svergun: combined SAXS/WAXS and SANS for protein fold discoveryProf. Coppensopened the discussion of Prof. Harrison's paper: The discrepancy between the thermal parameters and the cell dimension temperature seems paradoxical, as expansion is driven by anharmonicity, which increases when the thermal parameters increase. So the discrepancy must be in the experiment. My question is: were the crystals in the microwave experiment and the independent measurement the same, or could there be a difference in crystal quality and thus extraction? Or could such a difference be a result of the microwave treatment?Prof. HarrisonAlso Dr Whittaker, University of Edinburgh.replied: The neutron diffraction experiments were performed on the same sample of BaTiO3, first with conventional heating to 1173 K and then, after slow cooling back to 295 K, with microwave heating. We would not expect that this treatment would lead to any significant change in crystallinity between the conventional and microwave heating measurements, and therefore one would expect little change in factors such as extinction and strain broadening as a consequence of the heat treatment. Indeed, when the sample was remeasured at 295 K just before the microwave heating measurement, the peak widths were essentially unchanged. It should be stressed that the limited time available for neutron data acquisition led to relatively fast scans, and the precision of many of the refined parameters is less than ideal. The parameter that appears to show the most distinct change when the microwave and conventional heating are compared is the peak width parameter,σ1, as described in our paper.Prof. O'Hareasked: (1) Have you looked at the effects of using variable microwave frequencies?(2) Do different solids absorb the microwave energy in different ways?(3) Do the thermal parameters give you any indication of the types of motion that may be excited by the microwave field?Prof. Harrisonreplied: (1) No we haven't: so far our work has been restricted to frequencies of 2.45 GHz, which is one of the small number of frequencies allocated for non-telecommunications work. Microwave sources for those other prescribed frequencies are far more expensive, and the range of devices available at those frequencies is much more limited.(2) Absolutely: there are several distinct mechanisms of microwave heating, treated in detail in standard references (for example refs. 1 and 2). The most important modes of heating involve coupling of the microwave radiation with phonons (for example the present case of BaTiO3) or mobile species (electrons, for example in metals, or ions, for example in fast-ion conductors), but defects may also play an important role,3as may the coupling between the magnetic component of the microwave field and materials with high magnetic susceptibility (such as some ferrites).(3) In the first instance one would expect rapid equilibration of microwave energy among the various modes, which one would expect to be manifested through increased thermal parameters as an increase in anisotropic displacement parameters, just as one expects with conventional heating. However, one could imagine cases where the thermal parameters respond to the microwave field in a manner that is different from what one would expect from a simple increase in temperature. Where the rate at which particular modes are excited by microwaves is significantly higher than the rate of repartition, and where these modes are either polarised (as might happen when a single crystal is exposed to a linearly polarised microwave field) or in the case of a molecular solid localised to a particular region of the constituent molecules, it is conceivable that there is an anomalous response of either certain components of the thermal parameters, or of thermal parameters for a particular group of atoms.1 D. M. P Mingos and A. G. Whittaker, inChemistry Under Extreme or Non-Classical Conditions, ed. R. van Eldik and C. D. Hubbard, Wiley, Chichester, 1997, ch. 11, pp. 479–514.2 Y. V. Bykov, K. I. Rybakov and V. E. Semenov,J. Phys. D.: Appl. Phys., 2001,34, R55–57.3 J. H. Booske, K. I. Rybakov, V. E. Semenov, S. A. Freeman, J. H. Booske and R. F. Cooper,Phys. Rev. B, 1997,55, 3559–3567.Prof. Wilsonsaid: (1) Is it possible to decouple microwave heating effects from the electric-field induced issues?(2) Can you envisage a use for some sort of “spatial scanning” resonance radiography as a means ofin situtemperature measurement, helping to pin down the temperature fluctuations locally within the samples?Prof. Harrisonresponded: (1) Yes. This has been in a particularly elegant piece of work to probe this effect,1the essence of which was to study ion migration between similar compounds of different elemental composition in a linearly polarised microwave field. The result was an enhanced diffusion coefficient at a given temperature in the direction of polarisation.(2) I could imagine that such a technique could provide a spatially resolved probe of temperature, but the length scales involved could well be too large for some of the large and highly localised thermal gradients that can arise in microwave-heated systems, so it is likely to provide only a partial solution to some of the problems we presented.1 A. G. Whittaker and L. Cronin, inProceedings of the Second International Conference on Microwave Chemistry, ed. A. Gourdenne, 2000, Institut National Polytechnique de Toulouse, on behalf of AMPERE, Cambridge, UK.Prof. Wilsonsaid: I am interested in the coupling of the microwave frequencies with directional modes within the sample: are these compatible, and is such coupling possible?Presumably the energy redistribution within the sample/system results from partitioning of “thermal” motion into the various modes.Prof. Harrisonreplied: The microwave radiation couples with the solid through a multiphonon process, such that radiation whose frequency is of the order of GHz is able to excite modes whose frequency is the order of THz (see for example ref. 1). If the microwave field is linearly polarised, one would expect the phonons that are excited to be anisotropic. However, the energy put into these modes is rapidly repartioned through anharmonic processes with a time-scale typical of the phonon frequencies, that is of the order of THz (ps). For a recent consideration of this sort of process, see ref. 2.1 M. Sparks, D. F. King and D. L. Mills,Phys. Rev. B, 1982,26, 6987–7003.2 Y. V. Bykov, K. I. Rybakov and V. E. Semenov,J. Phys. D.: Appl. Phys., 2001,34, R55–57.Dr Techertasked: What is the expected order of magnitude of the phonon–phonon coupling ine.g.BaTiO3?Prof. Harrisonanswered: I don't know, and I imagine that it would be very difficult to determine this quantity accurately. When a material such as BaTiO3is heated with microwaves, phonons are excited through a multiphonon process, and this energy is then repartitioned over the phonon spectrum through anharmonic effects. One therefore needs to know which phonons are involved, and then calculate the appropriate anharmonicity, which is not trivial for a material containing heavy atoms. However, it has been estimated that for electric field strength and temperatures typical of those used in microwave processing, deviations of the population of high-energy phonons from Boltzmann values are very small indeed, implying that this process is fairly efficient in redistributing energy.11 Y. V. Bykov, K. I. Rybakov and V. E. Semenov,J. Phys. D.: Appl. Phys., 2001,34, R55–57.Dr Techertasked: Since the water in crystalline protein samples cannot rotate freely, as in solution, the energies of the corresponding libration modes should be shifted to higher frequencies. Do any calculations exist concerning these values?Prof. Harrisonreplied: The motion of water molecules bound to protein molecules in a variety of ways has been the subject of extensive studies by NMR with complementary modelling, and reveal a range of amplitudes and correlation times for libration (see for example refs. 1 and 2) extending from relatively fast, small-amplitude motion with a correlation time of the order of 0.07 ps, to slower, larger-amplitude motion correlated with the motion of portions of the molecule, with correlation times in range 1–10 ps. Note that pure water has modes in the region 0.05 ps that have been attributed to libration of H-bonded molecules.1 V. P. Denisov, K. Venu, J. Peters, H. D. Hörlein and B. Halle,J. Phys. Chem. B, 1997,101, 9380–9389.2 V. P. Denisov and B. Halle,J. Am. Chem. Soc., 1995,117, 8456–8465.Dr Grantcommented: Michael Levitt published a paper in 19851covering calculations (NMA) he made on bovine pancreatic tripsin inhibitor (BPTI) and he found that the collective motions of groups of up to ten residues have frequencies of 2 cm−1to 10 cm−1. 2 cm−1 ≃ 60 GHz, only about one order of magnitude larger than microwave frequency (2.45 GHz for a domestic microwave oven). Therefore the discrepancy between the THz frequency of many of the modes in a protein, and microwave oven and mobile phone frequency (1.8 GHz) is not such a large gap to bridge. The other point is that water in a protein is known to play an integral role in the unfolding and folding of a protein and water obviously couples with 2.45 GHz microwave radiation, so maybe we should be looking at the role of H2O in conformational changes induced by microwaves as providing a bridge to causing structural changes in the host biomolecule.1 M. Levitt, C. Sander and C. Stern,J. Mol. Biol., 1985,181, 423.Prof. Helliwellsaid: (1) You stimulate me to wonder whether the scattering pattern of sectioned brain, with and without microwaves, has been measured.(2) Assessing the risk of mobile phones, another possible “sample” would be the impact of microwaves on the inner ear. The physiology of the inner ear is studied using guinea pig inner ear, as it is easy to section.11 F. Mammano, IUPAB 'Prof. Harrisonanswered: (1) I don't believe that experiment has been tried yet, and I imagine it would be difficult to get conclusive information about microwave-induced changes in brain tissue from such measurements because of the complexity of the sample. At the moment the relatively small amount of work that is being done on the effect of microwaves on materials or biological systems, and which involves spectroscopic or structural measurements, has involved relatively simple model systems.(2) This would indeed be an interesting system for study in that it has been shown to be sensitive to irradiation; however, it is also a relatively complex system and I think it will be some time before conclusive measurements could be performed on it.