Prof. Helliwellopened the discussion of the Introductory Lecture: Your conversion efficiency in the Pt–Pt complex experiment was only 2%; have you explored,e.g.by simulation or calculation, the limits of applicability to lighter,i.e.weaker scattering, elements than Pt due to this?Prof. Coppensresponded: We believe that we can reach higher conversion percentages by working with smaller crystals at more intense sources, or by using more complex solids in which the active species are diluted. The ‘limiting percentage’ depends on the distortion, for example, for the large change in dihedral angle predicted to occur (for the isolated molecule) on excitation of benzophenone, the limiting percentage will be lower than for small changes. We have done some simulations; you are right that in general the conversion percentage must be higher when the scattering is not dominated by one or a few strongly scattering atoms.Prof. Bürgiasked: An illuminated crystal of TEA3HPt2(H2P2O4)4is composed of two species, 98% ground state molecules , 2% excited state molecules. Does the difference in Pt–Pt distances give rise to a diffuse scattering signal?Prof. Coppensanswered: We have assumed here and in our work on transition metal linkage isomers that the distribution of the excited molecules in the crystal is random. This assumption is supported by the lack of additional diffraction spots and the results of the structural analyses. So there should be a diffuse scattering contribution, but we have not looked for it.Dr Schmidtasked: (i) Did you see changes in the crystal parameters, since your observed atomic displacements are significant relative to the unit cell parameters? (ii) If yes, how did you correct for these?Prof. Coppensreplied: Yes. In the case of the binuclear Pt complex the excited state population was too small to observe an effect on the cell dimensions. However in our work on photoinduced metastable transition-metal nitrosyl and sulfur dioxide complexes, the conversion percentages were much higher (sometimes close to 50%), and significant cell dimension changes were observed. The sign of such changes correlates with the change in the molecular shape. For example, we observed the largest effect in (pentamethylcyclopentadienyl)nitrosyl-nickel, [Ni(NO)(η5-Cp*)], in which the NO groups are closely aligned with theb-axis. This axis shortens by 0.28 Å on photoinduced conversion of 47% of the molecules to the metastable state, while thec-axis shows a smaller decrease. In the combined analysis of the light-on and dark data it is necessary to take such changes into account. For the photodifference maps in which the ground state density is subtracted, the fractional coordinates must be converted to the new cell in order to prevent a bond length change on transfer to the new cell. In least squares refinement of the response ratios we use two sets of cell dimensions, in the metastable state work we use a slightly different procedure in which the ground state molecule is allowed to undergo rigid body motions after it has been included with corrected fractional coordinates to preserve the intramolecular distances.Dr Chensaid: I am not sure about the Pt–Pt distance in the compound, but we have encountered difficulties in determining the Rh–Rh distance in another similar compound. This distance was too long for EXAFS in solution at room temperature. (i) What were your excited state structure calculation results on the nearest neighbour to Cu distance? (ii) Would you please comment on any problem in making comparisons between your calculations that were carried out on a single molecule and your experimental results that were obtained in single crystals where the effect of crystal packing was significant?Prof. Coppensreplied: (i) We find that the Cu–N distance decreases by about 0.04 Å upon metal-to-ligand charge transfer. The shortening is as expected from the change in oxidation state of copper. We are working on the calculation of the five-coordinate exciplex, and hope to have results shortly. (ii) There are crystal structures for three different salts of the binuclear Pt complex. They are very different in terms of intermolecular interactions, two are acid salts with infinite hydrogen bonded networks, while the third, the potassium salt, is held together by cation–anion contacts. The variation in ground state distances between the three structures is about 0.01 Å at most, which makes it seem unlikely that the excited state distances would be very much affected by the crystal packing, in this case at least. The variation between distances from the various theoretical calculations for either the ground or excited state is about ten times larger.Prof. Helliwellopened the discussion of Prof. Wulff's paper: Your beamline at ESRF encompasses the spectrum of chemistry fields from biochemistry through to “pure” chemistry what features are common to all the chemical experiments in terms of structure to function rather than ‘simple’ characterization?Prof. Wulffresponded: The main focus of our work has been to follow structural changes following breakage or formation of a chemical bond in condensed systems. For systems such as myoglobin and iodine, the cage around the photo-activated molecule determines the reaction pathways and to some extent their time course.Dr Nibberingsaid: Most of the I2recombines very fast because of the cage effect, and what you observe is only the minor fraction that can escape the cage, diffuse away and finally recombine again. Would it not be beneficial if one were to persue an experiment where I2is inside a nanocavity, so that all I2molecules can dissociate to larger internuclear distances, and the expected signals in X-ray diffraction will be more pronounced.Prof. Wulffreplied: The photo-induced signal indicates that 15–50% of the photolysed iodine molecules dissociate (solvent dependent),i.e.break through the solvent cage. The signal from atomic iodine is therefore already quite strong.Dr Techertsaid: One advantage of time-resolved X-ray scattering is the extendedqrange, which can be scanned in one measurement. Therefore, it is possible to determine intra- as well as inter-molecular distances of a system within one experiment, in contrast to spectroscopic methods. This is summarised in Figs. 12 and 13 of your paper. In the reported experiment a ns laser was used as excitation source. In Fig. 13, the experimental difference maps at lowqrange are interpreted as thermal heating effects. How can the found thermal effects be affected by energy flow? Can, in general, energy flow be studied by time-resolved scattering techniques? How does the thermally induced change in the Boltzmann distribution function (see the paper) influence the difference map at highq?Prof. Wulffanswered: When two iodine atoms recombine to form a molecule, the binding energy is transferred to the solvent,i.e.the solvent cools the nascent molecule. The flow of energy is slaved to the recombination. It proceeds though at three stages: direct recombinationviavibrational cooling on the X-state potential (100 ps), intersystem crossing from the A-state (500–2700 ns) and atomic recombinationviadiffuse motion (μs). The result is that the solvent starts to expand. If the expansion was instantaneous, the distance between nearest neighbours, as measured by the angular position of the liquid peak, would be a direct measure of the energy flow. In a real liquid, the expansive motion and its time scale have to be included. The expansion is seen starting from 25–50 ns onwards. After about 1 μs, the expansion stops,i.e.the system reaches thermal equilibrium. We are currently working with Savo Bratos, University of Paris, to separate the thermal expansion from the energy flow. Concerning your second question, if an iodine molecule is (thermally) excited to a vibrational level near the ground state, its average bond length will increase. A distribution of bond lengths will shift and broaden the difference oscillations and the effect is largest at highq. However this broadening does not prevent us from seeing oscillations up to aQ = 9.0 Å−1in iodine and we have the impression that the effect is small in iodine. The effect should be treated as a Debye–Waller factor as in crystallography.Dr Mayasked: Is there any chance of activating the C–Cl bond and what would be the consequences?Prof. Wulffreplied: From recent studies of CH2I2in the solvent CH2Cl2, we know that this solvent can be activatedviatwo-photon absorption. It is our impression that the solvent CCl4is completely transparent to the (softer) 515 nm laser pulse used here. We are planning an experiment to check this point.Prof. Greavessaid: You use the molecular potential for iodine to interpret your laser excited X-ray scattering data. Do these new experiments enable you to refine the potential in any way or do they simply demonstrate its correctness?Prof. Wulffreplied: Our initial fit functions, based on simple molecular form factors with the solvent cage, are not very sensitive to the bond length of the A/A-state. We will soon include cage effects and hope that that will enhance our sensitivity. In addition, new multi-layer optics will reduce the polychromatic bandwidth from 3–4% to about 1%. Combined, we hope to enhance our model selectivity.Prof. Coppensasked: Is your spatial resolution affected by the use of a 3% band width. As the lineshape of your undulator harmonic is asymmetric does it introduce a bias?Prof. Wulffanswered: The polychromatic X-ray spectrum from a single-harmonic undulator leads to a slight increase in the background of the radial intensity on a CCD detector and that dampens the oscillation amplitude in the difference pattern. Simulations show that the damping is small and greatly compensated by the 250–500 fold gain in flux over conventional monochromatic methods.Dr Techertopened the discussion of Dr Nibbering's paper: In the reported work the out-of-plane-twisting mode was found to be 60 cm−1, which is surprisingly high if it is compared with the twist motion of stilbene (ca.11 cm−1). Is this high value due to the rigidity of the system, which enhances the harmonic contribution (in contrast to the rotational contribution in stilbene) to this vibrational mode?Dr Nibberingresponded: The high value is indeed due to the strong intramolecular hydrogen bond, making the molecular structure more rigid. We compared the out-of-plane mode of 2-(2′-hydroxyphenyl)benzothiazole (HBT) with that of its anion (where the proton is removed by a strong base) in quantum chemical calculations. In the case of the anion, where no hydrogen bond exists, the frequency of the out-of-plane mode is very close to that of stilbene.Prof. Moffatasked: Have similar ultra-fast spectroscopic measurements been made either in single crystals or in HBT bound to a small protein? In other words, how does the “solvent” environment affect the mechanism and rates of reaction?Dr Nibberinganswered: The observed H-transfer reaction in HBT in tetrachloroethene represents a clear case of intramolecular H-transfer with minor solvent contributions. When dissolving HBT in polar solvents, like DMSO, alcohols or water, the efficiency of intramolecular hydrogen bond formation will be affected and compete with intermolecular hydrogen bonds with solvent molecules. Since these changes in the hydrogen bond have pronounced effects on the optical spectra, HBT in polar solvents has been studied in optical pump–probe experiments, where different proton transfer times have been deduced from the observed transients. We have not performed femtosecond IR experiments on HBT in polar solvents yet. We expect to see different dynamics dependent on the hydrogen bond configurations.Dr Techertasked: How isexcitonic couplingdefined in the liquid phase (or does this only refer to IR studies on peptides)?Dr Nibberingreplied: Excitonic coupling refers to coupling of vibrational modes within a larger molecular system, that may itself be in the condensed phase (liquids, proteins, solids) or even in the gas phase. These vibrational modes have a mutual coupling when their vibrational bands spectrally overlap or are at very small detuning. Coupling might take place by a through space transition dipole–dipole coupling mechanism with angle and distance dependences. Another mechanism might involve a through bond coupling scheme. Examples include the amide I band in amino acid subunits in peptides, or hydrogen bonded OH-groups in dimeric systems, such as the acetic acid dimer or the DNA base pairs in the double helix.Dr Chenasked: How do you distinguish anharmonic coupling in the proton transfer processes that you are probing from heterogeneity of H-bonding distribution that was disturbed by the laser excitation, which may also result in the spectral shift of the O–H shielding band?Dr Nibberinganswered: The theory of O–H stretching vibration line shapes shows that anharmonic coupling of the high-frequency O–H stretching mode to other vibrational modes leads to (i) a red-shifting, a consequence of the weakening of the O–H stretching vibration; (ii) a Franck–Condon progression due to coupling with underdamped low-frequency modes; (iii) a broadening, due to coupling to overdamped low-frequency modes; (iv) level splitting due to Fermi resonances with vibrational overtone or combination modes. We probe here the carbonyl stretching mode marking the formation of the keto*-product state. We observe in a clear way the effects of anharmonic coupling of coherently excited underdamped intramolecular low-frequency modes. We observe these coherently excited low-frequency modes also in the O–H/N–H stretching region. Interpretation is more problematic though due to the above mentioned other contributions to the overall line shape of O–H and N–H stretching bands. Since the O–H and N–H stretching bands spectrally overlap to a significant extent, it is difficult to distinguish the coherent motions in the keto*-product and enol-ground states.Prof. Coppenssaid: Is the time scale of H-transfer distance dependent? Would it be slower for intermolecular transfers?Dr Nibberingreplied: The time scale of H-transfer is first and foremost determined by the shape of the potential energy surfaces. If barriers exist, reaction times may be longer, and even the surrounding solvent may play a decisive role. For the photoacid molecules we currently study the dynamics appear to take place on picosecond time scales. However, even in these cases recent findings suggest that the H-transfer distance is not the most important parameter, but a more complex relaxation scenario through different electronic states may be the rate-limiting step.Prof. Wilsonsaid: You are correct that determination of hydrogen atoms is difficult with X-ray diffraction but the accurate “end-point” structures can be examined by a “slow technique”—neutron diffraction. Such precise/accurate structural techniques provide the ideal basis for setting upe.g.theoretical chemical modelling as identified by Prof. Coppens in his Introductory Lecture. In addition, such static methods can show “precursor” effects which set up conditions for processes such as proton transfer. For example, variable temperature or variable pressure studies can begin to probe the average shapes of potentials, for example in low barrier potentials.Dr Nibberingresponded: I agree that for a full understanding of the dynamics one needs to determine the structural information in real time. While awaiting the development of time-resolved structure resolving techniques with femtosecond time resolution, such as ultrafast X-ray spectroscopy, X-ray diffraction, electron diffraction or even neutron diffraction, we pursue femtosecond vibrational spectroscopy. However, for a full interpretation of our studies the understanding of the potential energy surfaces is a necessary requirement. Here the mentioned “slow techniques”, or I would rather say “steady-state techniques with averaging over a broad range of time scales” may be helpful in giving part of that information on the steady-state configurations, with high accuracy as well as the option to study temperature or pressure dependences. Nevertheless, for the study of transient states time resolved techniques are the only experimental approaches that potentially reveal the structural dynamical information.Dr Coleasked: If you were to isotopically enrich the subject sample and perform an otherwise identical experiment to that described, could you use the comparison of the data,viaexploitation of the isotope effect, to draw out more quantitative information from the results,e.g.with regard to force constants, and the alike? If experimental complications could be overcome, I wondered if it would prove very useful, not only for this quantification of the results, but also to help highlight the harmonic frequencies which could be used to give further consistency to the results in hand.Dr Nibberinganswered: It is known that in numerous H-transfer studies a kinetic isotope effect exists. Isotopic labelling will then give information on the potentials (i.ethe force constants). For the case of HBT the currently adopted model ascribes the finite H-transfer time to the motion of a low-frequency mode modulating the hydrogen bond distance. Since in this case the dynamics involve relative motions of the O and N atoms constituting the hydrogen bond, no significant H/D-exchange effect on the reaction time scale can be expected. This is reflected by the fact that the low-frequency modes do not alter signifcantly their frequencies upon H/D exchange.Prof. Helliwellsaid: You are advocating in your conclusions the femtosecond IR spectroscopy method but what systems do you have in prospect to apply it to? (Finding systems that are appropriate can be difficult and of course it is the proposition of this meeting that such research can be scientifically rewarding at the “Smoking Gun” level.)Dr Nibberingreplied: We are currently studying the dynamics of intermolecular proton transfer of photoacids in liquid solution, and we aim to extend these studies to photoacids in protein surroundings. In a different line of experiments we now also use femtosecond IR spectroscopy in the study of the photoinduced dynamics of photochromic switches. In principle the method can be extended to any chemical process induced by an ultrafast trigger, that may be an optical light pulse (for dynamics in electronic excited states), but could also be an IR pump pulse, initializing reactions in the electronic ground state.Prof. Helliwellopened the discussion of Prof. Bürgi's paper: Central to your approach is that the structure database scatter plot reveals sampling of metastable intermediate structures but, if I may use the analogy of jumping off a cliff, at the cliff and on the beach you have a larger population of structures than you do falling to the beach. How do you really reveal snapshots between the top and the bottom? (After all to isolate structural intermediatesviatime-resolved experiments requires considerable ingenuity to trap them in time.)Prof. Bürgiresponded: Structural databases sample stable structures only. For a set of structures with a particular fragment in common, the structures of the common fragment—the only part comparable among the members of the set—may vary considerably. The variation in bond lengths and angles reflects the influence of those parts of the structures, which are not common to the set,i.e.the influence of the environments. Our approach does not presume one cliff and one beach, but rather a series of closely related energy surfaces whose minima delineate a (more or less well defined) region of parameter space such as a reaction path (see Fig. 6 of our paper, for example).Dr Sagiasked: How can a dynamic process be elucidated from a collection of data points that were taken in different crystallization conditions. For example, ionic strength, pH, solventetc.Prof. Bürgireplied: Techniques in time resolved chemistry localize chemical processes along the arrow of time. In our approach we attempt localization in the structural domain. The approach is based on the hypothesis that a given class of ground state dynamic processes is associated with a common reaction path on a generic energy surface. The details of the path and the surface for a specific process,i.e.its minima and transition states, depend somewhat on the specific molecule undergoing this process. Our hypothesis implies that the stationary points characterizing the various molecules tend to congregate in the low-lying parts of the generic surface, which—by definition—includes the reaction path of the process. They can be said to ‘map the reaction path’ along the structural coordinate (principles of structure–structure and structure–energy correlation, see section 3.1). Empirically the hypothesis has been found to hold for many different chemical systems (see ref. 11 of the paper). (My answer to Prof. Helliwell's question is also relevant.)Dr Techertasked: How would the presented simulation change if the assumptions of the Eyring theory/transition state theory (equilibrium reactant ↔ TS) were to break down?Prof. Bürgianswered: I assume that you refer to the empirical correlations between ground state structures obtained from crystal structure analyses and activation energies derived from reaction rate constants measured in solution (see section 5 of the paper). These correlations apply to thermally activated processes and illustrate the Hammond principle from a structural point of view. They are similar in a way to correlations that can be understood in terms of a Marcus relationship, which emphasises the energy point of view. The reactions you are interested in are initiated by a photo-excitation and followed by vibrational relaxation. These processes lead to a more or less long lived intermediate associated with a structure on an excited-state surface that usually differs significantly from the original ground state structure. Therefore the simple correlations described in section 5 will not apply.Dr Hirstsaid: One section of your paper mentions the non-planarity of amides, referring to deviations of up to 10° in proteins. Do you believe that protein crystallography can resolve such deviations reliably? And if so would you like to expand on your speculation that this extra degree of freedom is important in protein folding?Prof. Bürgireplied: To measure the non-planarity of amides accurate positions of the atoms Ci−1′, Ni, αCiand HNiare needed. For a number of high-resolution X-ray structures several torsion anglesω(αCi−1–Ci−1′–Ni–αCi) have been reported to deviate from 180° by more than 10° (for a recent example see ref. 1). The position of HNicannot be determined accurately from even the highest-resolution X-ray diffraction experiments, but could be determined by high-resolution neutron diffraction (<1 Å). The hinge-like flexibility associated with this degree of freedom could help to minimize strain energy of protein folds, especially with respect to N–H⋯X hydrogen bonds. It could also facilitate domain motions associated with protein function. Even though the energy increment associated with the flexibility of a single N–H⋯X hydrogen bond may be small, a substantial contribution to the total energy may result, because of the large number of such hydrogen bonds. However, these are speculations and require experimental verification.1 J. Symersky, Y. Devedjiev, K. Moore, C. Brouillette and L. DeLucas,Acta Crystallogr., Sect. D, 2002,58, 1138–1146.Prof. Sir John Meurig Thomascommented: Prof. Burgi's persuasive arguments about how much we can learn about fast chemical processes from slow diffraction experiments prompts me to draw to the Discussion's attention to how much it becomes possible to gain quite useful insights into the process of catalytic turnover at a well defined active site from steady-state measurements of X-ray absorption fine structure coupled with density functional theory computations. In papers that my colleagues and I have published elsewhere1–3we showed that the commercial Ti–SiO2catalyst for the epoxidation of olefins has an active site which is a Tiivion tripodally attached ,viaoxygens, to the underlying silica support. This ion has also attached to it an OH group. In the presence of hydroperoxide oxidant and the alkene, the XANES fingerprint of the Tiivunmistakably shows it to be 6-coordinated,3and EXAFS analysis yields precise bond distances and quite good bond angles in which the Tiivis involved (see the figures in the paper). These tally well with the corresponding values deduced from DFT. Taking all experimental factors into consideration we may plausibly portray the key catalytic act as the “plucking away” of one of the oxygens of the bound peroxide, which leaves the active site in its original 4-coordinated state, ready for further catalytic turnover. Fuller details are described in a recent paper.4(A fictional animation of the conversion of cyclohexene to its epoxide was shown during this contribution.)1 T. Maschmeyer, G. Sankar, F. Ray and J. M. Thomas,Nature, 1995,378, 159.2 J. M. Thomas, G. Sankar and C. R. A. Catlow,Top. Catal., 2000,10, 225.3 J. M. Thomas and G. Sankar,Acc. Chem. Res., 2001,34, 571.4 G. Sankar, J. M. Thomas C. R. A. Catlow, C. M. Baker, D. Gleeson and N. Kaltsoyannis.Prof. Greavessaid: In your intriguing video which models the structural changes taking place during the catalytic event around Tiiv, the start and end configurations come directly from static experiments.Prof. Wilsoncommented: To emphasise the point ofin situversusquenching experiments, oftenin situexperiments with extrapolation can give more accurate (and less misleading) information than more traditional methods which often involved quenching and the implicit assumption that the state frozen-in by the quenching process is actually of relevance to the reaction under study.Prof. Helliwellopened the discussion of Prof. Moffat's paper: (i) Re your target proteins, how many are membrane bound (i.e. which would then be very difficult to crystallise)? (ii) Re Fig. 1 of your paper. A monochromatic chirping approach will involve a profile method. What accuracy of intensity measurements are to be expected?Prof. Moffatresponded: (i) The target proteins are identified only by sequence homology. Some such as the photosynthetic reaction center and light-harvesting complexes are indeed integral membrane proteins; others such as phototropin are loosely membrane-associated and can readily be solubilized, yet others are probably cytoplasmic and soluble in aqueous media. The experimental distribution into these three subsets, to my knowledge, has not been established. My point is that experimental organisms such asArabidopsisoffer several light-sensitive target proteins (and even more targets if domains of the full-length proteins are considered, or of the light-insensitive components that lie further downstream in the overall, light-driven signal transduction pathways). (ii) This has not yet been explored. The accuracy will depend on, among other factors, the nature of the spatial chirp, the mosaicity of the crystal and the resolution of the spot being considered. One way to approach this problem is to introduce time-dependent Ewald spheres into the static treatment for oscillation spot size and shape developed 20 years ago by, for example, Greenhoughet al.1–31 T. J. Greenhough, J. R. Helliwell and S. A. Rule,J. Appl. Crystallogr., 1983,16, 242–250.2 T. J. Greenhough and J. R. Helliwell,J. Appl. Crystallogr., 1982,15, 493–508.3 T. J. Greenhough and J. R. Helliwell,Prog. Biophys. Mol. Biol., 1983,41, 67–123.Dr Mayasked: (i) You mentioned that there are 58 photosensitive proteins inArabidopsis. How many proteins are not photosensitive? Only direct photoreactions are very rapid. (ii) You may not be able to observe all reactions (and their consequences) that are going on in the protein within a crystal, because the movement may not be possible in the crystal.Prof. Moffatanswered: (i) The exact number of proteins inArabidopsisthat are not photosensitive is now known; but it is certainly orders of magnitude larger than 58. That is, only a small fraction of proteins in an organism are naturally photosensitive. Yes, only direct photoreactions are very rapid, but this is characteristic of all naturally light-sensitive proteins. Attempts to confer light sensitivity on otherwise light-inert reactions by preparing so-called “caged” compounds, such as caged ATP, suffer at present from the fact that the initial, direct photoreactions are followed by much slower, dark reactions that ultimately liberate the desired product, such as ATP. It may be possible to develop other forms of caged compounds that do not suffer from this limitation.(ii) Yes, but the intermolecular forces that stabilize the crystal lattice in biological macromolecules are weak, roughly 1 kcal mol−1per interface. I find it worrying that activity in the crystalline state is seldom directly examined, let alone quantitated; we blithely assume that crystal structures always provide an accurate foundation for assessing mechanism. In any case, it is highly desirable to check by optical means that the reaction in the crystal firstly can proceed at all, secondly that it does so by the same reaction mechanism as in dilute solution, and thirdly that the rate coefficients associated with each step in this mechanism are not significantly different from those in dilute solution. (“Significant” here must be thought of in energetic terms; by how much is the free energy of activation for the particular step affected?)Prof. Sir John Meurig Thomassaid: Do you feel that, in using Laue time-resolved and other X-ray crystallographic techniques on enzyme crystals, one is addressing the active sites and other regions of the enzyme under realistic (pseudo-physiological) conditions?Prof. Moffatreplied: Yes—provided always that one can demonstrate that chemical or biochemical activity is quantitatively retained in the crystalline state. If this is NOT the case then time-resolved experiments are of no relevance and indeed, the static enzyme structure itself comes into question.A historical note on this point—I was first stimulated to consider time-resolved crystallography in 1969 by a key paper by Parkhurst and Gibson1who examined quantitatively the reactivity of haemoglobin towards carbon monoxide in the physiological environment of the erythrocyte, in the biochemist's environment of dilute solution, and in the crystallographer's environment of the intact crystal. They concluded that, at least for the reaction studied, the kinetic behaviour of haemoglobin was very closely similar in these three, very different environments. This provided strong supporting evidence for the physiological relevance of the crystal structures of haemoglobin then being determined by Perutz and colleagues. It took a further 25 years before it became clear that the kinetic properties of haemoglobin are in fact significantly different, both qualitatively and quantitatively, in the crystal compared with dilute solution.21 L. J. Parkhurst and Q. H. Gibson,J. Biol. Chem.,1967,242(24), 5762.2 C. Rivetti, A. Mozzarelli, G. L. Rossi, E. R. Henry and W. A. Eaton,Biochemistry, 1993,32(11), 2888–2906.Prof. Greavescommented: If you can bring the time scale of diffraction experiments into the regime of ps or shorter, will this possibly open the door to penetrating the co-operative elements of photostructural phenomena in macromolecules.Prof. Moffatadded: It might bring coherent structural transitions into view—an exciting challenge.Dr Chensaid: I am always very impressed by the movies of protein movement from your studies. However, I would like to understand what happens in the later time after the initial coherent atomic movement triggered by the laser pulse are lost in the protein molecules. One can envision by the potential energy landscape in the protein that multiple conformations exist after the initial protein quake?Prof. Moffatreplied: Coherence is presumably maintained only for a very short period of time after the laser pulse, perhaps a few hundred fs but in any case, a time much shorter than the present time resolution of our experiments. Thereafter, we believe that structural processes evolve purely stochastically; coherence is lost. That is, the subsequent time evolution arises from the build-up and decay of the populations of intermediate structures, exactly as in dilute solution. A consequence is that at all time points, there is indeed a mixture of multiple conformations present. Our challenge is to unscramble this mixture (for example, by singular value decomposition and associated techniques) and to recover the time-independent structures of all intermediates. This is in progress, as illustrated by the poster by Schmidt, Rajagopal, Ren and Moffat presented at this meeting.Prof. Bürgiasked: How does the information from a chirped pulse in a time slice of 1 ps, say, compare with that of a typical Laue experiment?Prof. Moffatanswered: In a time slice of 1 ps, say at a timetafter the laser pulse, only a subset of the Laue spots on a “normal” Laue image would be stimulated. This subset corresponds to those spots stimulated between an X-ray energyEandE + dE, where the values ofEand dEdepend ont, dt(here, 1 ps) and the nature of the chirp. Thus the information content would be a subset of that present in a “normal” Laue image. However, I emphasize that each “chirped” Laue image contains all the spots that would be present in the corresponding “normal” Laue image.Prof. Greavessaid: You say that the time resolution for Laue diffraction from a chirped source in principle is greater than that from an unchirped source even though the former is far longer than the latter. Is this because you have created a more manageable instrument function?Prof. Moffatreplied: The time resolution from an unchirped pulse is today set by the total duration of the X-ray pulse, say 100 ps. If such a pulse is manipulated to produce a (much longer) chirped pulse, the time resolution can never be better than 100 ps since phase space volume must be conserved. Equally, if a 100 fs unchirped pulse from the proposed SPPS or LCLS (Linac Coherent Light Source) is stretched to provide a 1000 times longer, chirped pulse of 100 ps total duration, it may be possible to retain near-100 fs time resolution by the strategies outlined in the main paper. This is not really due to an “instrument function”; rather, it results from the experimental design that maps time into X-ray energy and X-ray energy into space, on the detector.Prof. Finneyasked: Do you have any good physical evidence to justify a statement that, as you improve the resolution towards ∼100 fs, you will have a coherent target system? Light takes 1 ps to cross a 0.3 mm crystal. Does this not raise a basic problem with approaching 100 fs time resolution?Prof. Moffatanswered: Yes it does; the physical dimensions of the crystal (likely to be somewhat smaller than 0.3 mm, say 0.075 mm in typical dimension) does enter into the time resolution. Rather than a disadvantage this fact may be exploited as proposed by Neutze and Hajdu;1but this will be very challenging experimentally.1 R. Neutze and J. Hajdu,Proc. Natl. Acad. Sci. USA,1997,94, 5651–5655.Dr Nibberingasked: Did you consider the effects of group velocity mismatch between the optical pump pulse and the X-ray probe pulse? How thin do your crystals have to be to maintain the anticipated time resolution of about 100 fs (100 μm or less)?What do you expect will happen when probing a molecular wave packet motion that will spread out in the course of time? Will the diffraction signal initially be strong and wash out with the dephasing time (typically 0.5–2 ps)?Prof. Moffatreplied: As in my reply to Prof. Finney, crystal dimensions do inescapably enter into the time resolution.The initial diffraction signal will presumably arise from coherent motion of the excited species; and then evolve (but not “wash out”) as coherence is lost over the few ps time frame when stochastically independent structural processes begin to come into play. (My replies to Prof. Greaves and to Dr Chen are also relevant here.)Prof. Greavessaid: If the pump pulse occurs during the chirped pulse and the X-ray detector has no time resolution, how are the ensuing changes in diffraction differentiated from those that are unperturbed if neither are known beforehand—or is this simply a question of normalising against a pulse-free pattern?Prof. Moffatresponded: Laser-pulse-free patterns will normally be acquired, interleaved in time with the laser-pulse-present patterns. Comparison of the structure amplitudes derived from the latter (which correspond to time zero) with those of the former (which correspond to different times after the laser pulse depending on their X-ray energy) will yield the desired, time-dependent differences in structure amplitudes.Prof. Wilsonsaid: As I understand it, there is no time (energy) or space resolution on the detector. Also, the method does not improve the overall data collection time resolution, but instead is improving the intrinsic time resolution of the Laue sections taken in each shot.Prof. Moffatreplied: There is indeed no time resolution in the detector but it certainly has excellent spatial resolution; and energy resolution is afforded by indexing the Laue pattern and hence assigning an energy (or more accurately stated, a small energy range) to each spot. In conventional Laue experiments, the time resolution is at present limited by the total duration of the X-ray pulse from the synchrotron, typically 100 ps. The time resolution of a Laue experiment conducted with a chirped X-ray pulse is not limited by the total duration of this pulse but can be much shorter. Exactly how much shorter depends on many factors, among them the nature of the chirp itself.