X-ray crystallography is a form of microscopy that allows the three-dimensional arrangement of atoms belonging to molecules within crystals to be determined. In this method, a crystal is illuminated with a beam of X-rays and the diffracted amplitudes resulting from the illumination are measured and computationally processed to enable the electron density of the unit molecule, or the unit cell, constituting the crystal to be calculated. The recent development of the X-ray free-electron laser (XFEL) provides new routes for determining molecular structures via its ability to generate intense but brief X-ray pulses. These new instruments enable diffraction measurements to be obtained from crystals that have a small number of unit cells, referred to as nanocrystals, and molecular structure determination via this technique is known as serial femtosecond nanocrystallography (SFX). This thesis is concerned with the characterisation of diffraction data obtained from SFX experiments and the techniques for reconstructing the electron density of the molecule from such data. The noise characteristics of diffraction measurements from nanocrystals is developed. Methods for directly inverting nanocrystal diffraction to obtain the electron density of the molecule are analysed and an approach to ameliorate the effect of noise is proposed and evaluated by simulation. A model for diffraction by nanocrystals that include the effects of different unit cell arrangements and incomplete unit cells on the crystal surface is also developed and explored by simulation. The diffraction by finite crystals is shown to be equal to the incoherent average over a set of unit cells that contain different molecular arrangements related to the symmetry of the crystal at hand. The problem of image reconstruction under this circumstance is investigated. The more general problem of reconstructing multiple, unrelated, objects from their averaged diffraction is also explored and uniqueness properties along with reconstruction algorithms developed. The problem of reconstructing multiple, related, unit cells is studied and preliminary results are obtained. These results show that iterative phase retrieval algorithms can in principle be adapted to reconstruct the electron density of a crystalline specimen from the data obtained in SFX and the retrieval of phases from the diffracted intensity averaged over multiple objects is feasible.

abstract: The superior brightness and ultra short pulse duration of X-ray free electron laser (XFEL) allows it to outrun radiation damage in coherent diffractive imaging since elastic scattering terminates before photoelectron cascades commences. This “diffract-before-destroy” feature of XFEL opened up new opportunities for biological macromolecule imaging and structure studies by breaking the limit to spatial resolution imposed by the maximum dose that is allowed before radiation damage. However, data collection in serial femto-second crystallography (SFX) using XFEL is affected by a bunch of stochastic factors, which pose great challenges to the data analysis in SFX. These stochastic factors include crystal size, shape, random orientation, X-ray photon flux, position and energy spectrum. Monte-Carlo integration proves effective and successful in extracting the structure factors by merging all diffraction patterns given that the data set is sufficiently large to average out all stochastic factors. However, this approach typically requires hundreds of thousands of patterns collected from experiments. This dissertation explores both experimental and algorithmic methods to eliminate or reduce the effect of stochastic factors in data acquisition and analysis. Coherent convergent X-ray beam diffraction (CCB) is discussed for possibilities of obtaining single-shot angular-integrated rocking curves. It is also shown the interference between Bragg disks helps ab-initio phasing. Two-color diffraction scheme is proposed for time-resolved studies and general data collection strategies are discussed based on error metrics. A new auto-indexing algorithm for sparse patterns is developed and demonstrated for both simulated and experimental data. Statistics show that indexing rate is increased by 3 times for I3C data set collected from beam time LJ69 at Linac coherent light source (LCLS). Finally, dynamical inversion from electron diffraction is explored as an alternative approach for structure determination. / Dissertation/Thesis / Doctoral Dissertation Physics 2016

京都大学 / 0048 / 新制・課程博士 / 博士(理学) / 甲第22239号 / 理博第4553号 / 新制||理||1654(附属図書館) / 京都大学大学院理学研究科物理学・宇宙物理学専攻 / (主査)准教授 松田 和博, 教授 田中 耕一郎, 教授 佐々 真一 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM

abstract: Serial femtosecond crystallography (SFX) uses diffraction patterns from crystals delivered in a serial fashion to an X-Ray Free Electron Laser (XFEL) for structure determination. Typically, each diffraction pattern is a snapshot from a different crystal. SFX limits the effect of radiation damage and enables the use of nano/micro crystals for structure determination. However, analysis of SFX data is challenging since each snapshot is processed individually. Many photosystem II (PSII) dataset have been collected at XFELs, several of which are time-resolved (containing both dark and laser illuminated frames). Comparison of light and dark datasets requires understanding systematic errors that can be introduced during data analysis. This dissertation describes data analysis of PSII datasets with a focus on the effect of parameters on later results. The influence of the subset of data used in the analysis is also examined and several criteria are screened for their utility in creating better subsets of data. Subsets are compared with Bragg data analysis and continuous diffuse scattering data analysis. A new tool, DatView aids in the creation of subsets and visualization of statistics. DatView was developed to improve the loading speed to visualize statistics of large SFX datasets and simplify the creation of subsets based on the statistics. It combines the functionality of several existing visualization tools into a single interface, improving the exploratory power of the tool. In addition, it has comparison features that allow a pattern-by-pattern analysis of the effect of processing parameters. \emph{DatView} improves the efficiency of SFX data analysis by reducing loading time and providing novel visualization tools. / Dissertation/Thesis / Doctoral Dissertation Biochemistry 2019

abstract: Serial femtosecond crystallography (SFX) with X-ray free electron lasers (XFELs) has enabled the determination of damage-free protein structures at ambient temperatures and of reaction intermediate species with time resolution on the order of hundreds of femtoseconds. However, currently available XFEL facility X-ray pulse structures waste the majority of continuously injected crystal sample, requiring a large quantity (up to grams) of crystal sample to solve a protein structure. Furthermore, mix-and-inject serial crystallography (MISC) at XFEL facilities requires fast mixing for short (millisecond) reaction time points (𝑡"), and current sample delivery methods have complex fabrication and assembly requirements. To reduce sample consumption during SFX, a 3D printed T-junction for generating segmented aqueous-in-oil droplets was developed. The device surface properties were characterized both with and without a surface coating for improved droplet generation stability. Additionally, the droplet generation frequency was characterized. The 3D printed device interfaced with gas dynamic virtual nozzles (GDVNs) at the Linac Coherent Light Source (LCLS), and a relationship between the aqueous phase volume and the resulting crystal hit rate was developed. Furthermore, at the European XFEL (EuXFEL) a similar quantity and quality of diffraction data was collected for segmented sample delivery using ~60% less sample volume than continuous injection, and a structure of 3-deoxy-D-manno- octulosonate 8-phosphate synthase (KDO8PS) delivered by segmented injection was solved that revealed new structural details to a resolution of 2.8 Å. For MISC, a 3D printed hydrodynamic focusing mixer for fast mixing by diffusion was developed to automate device fabrication and simplify device assembly. The mixer was characterized with numerical models and fluorescence microscopy. A variety of devices were developed to reach reaction intermediate time points, 𝑡", on the order of 100 – 103 ms. These devices include 3D printed mixers coupled to glass or 3D printed GDVNs and two designs of mixers with GDVNs integrated into the one device. A 3D printed mixer coupled to a glass GDVN was utilized at LCLS to study the oxidation of cytochrome c oxidase (CcO), and a structure of the CcO Pr intermediate was determined at 𝑡" = 8 s. / Dissertation/Thesis / Supplementary Video D.1 - Droplet formation in a 3D printed droplet generator / Doctoral Dissertation Chemistry 2019

京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第20071号 / 医博第4164号 / 新制||医||1018(附属図書館) / 33187 / 京都大学大学院医学研究科医学専攻 / (主査)教授 黒田 知宏, 教授 松田 文彦, 教授 楠見 明弘 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

Röntgenlicht-Freie-Elektronen-Laser (XFEL) schaffen neue Möglichkeiten für die molekulare Strukturbestimmung in Einzelmolekülexperimenten. In dieser Arbeit stelle ich zwei alternative bayessche Verfahren vor, das Orientational Bayes und das Structural Bayes Verfahren, die das Extrahieren der Strukturinformationen aus dünn besetzten und verrauschten Streuungsbildern ermöglichen. Im ersten Verfahren wird ein "Seed"-Modell verwendet, um die zugrunde liegende molekulare Orientierung für jedes aufgezeichnete Streuungsbild separat zu bestimmen. Eine verbesserte molekulare Transformation der bestrahlten Moleküle wird durch Ausrichten und Mitteln dieser Bilder im dreidimensionalen reziproken Raum erhalten. Im Structural Bayes Verfahren wird ein Realraum-Strukturmodell optimiert, sodass es am besten zum gesamten Streuungsbildersatz passt. Auf diese Weise wird ermöglicht, zwischen verschiedenen Strukturmodellen zu unterscheiden. Ich habe die Auflösung bei der Abbildung einzelner Moleküle mit unterschiedlichen Massen für verschiedene XFEL Strahlintensitäten abgeschätzt. Die Ergebnisse zeigen, dass die erreichbare strukturelle Auflösung mit der Molekülmasse wie M^{-1/ 6} steigt. Laut dieser Skalierung ist hierbei, im Gegensatz zur traditionellen Röntgenkristallographie, die hochaufgelöste Strukturbestimmung kleiner Einzelmoleküle, im Vergleich zu großen Molekülen, schwieriger. Als Machbarkeitsnachweis des Orientational Bayes Verfahrens wurde beispielhaft die Elektronendichte eines Glutathion-Moleküls aus 20.000 synthetischen Streuungsbildern, mit durchschnittlich 82 aufgezeichneten elastisch gestreuten Photonen und bis zu 50% zusätzlichem Hintergrundrauschen pro Bild, berechnet. Um die Anwendbarkeit des Structural Bayes Verfahrens in einer de novo Strukturbestimmung zu testen, wurde zudem die Struktur des Glutathion-Moleküls in einer Monte Carlo-Verfeinerungs-Simulation gelöst, für die zufällige Aminosäure-Konformationen als Ausgangsmaterial verwendet wurden. Um zusätzlich zu prüfen, ob mehrere Längenskalen umfassende Strukturänderungen in einem komplexen Molekül unter Verwendung des Structural Bayes Verfahrens rückverfolgbar sind, wurden Konformationsänderungen von drei Immunglobulin-Domänen eines Titin-Moleküls sowie der tRNA-Translokationsvorgang im Ribosom untersucht. Die Ergebnisse zeigen, dass es möglich ist sowohl zwischen unterschiedlichen molekularen Konformationen zu unterscheiden als auch kleinere strukturelle Änderungen, die mit der tRNA-Translokation assoziiert sind, zu erkennen. Insgesamt betrachtet deuten die Ergebnisse dieser Arbeit darauf hin, dass sich mithilfe der beiden hier vorgestellten bayesschen Verfahren die Struktur einzelner Moleküle mit atomarer Auflösung von dünn besetzten und verrauschten Röntgenstreuungsbildern aus XFEL-Einzelmolekülexperimenten für ein breites Spektrum von Molekülmassen bestimmen lässt.

abstract: Phase problem has been long-standing in x-ray diffractive imaging. It is originated from the fact that only the amplitude of the scattered wave can be recorded by the detector, losing the phase information. The measurement of amplitude alone is insufficient to solve the structure. Therefore, phase retrieval is essential to structure determination with X-ray diffractive imaging. So far, many experimental as well as algorithmic approaches have been developed to address the phase problem. The experimental phasing methods, such as MAD, SAD etc, exploit the phase relation in vector space. They usually demand a lot of efforts to prepare the samples and require much more data. On the other hand, iterative phasing algorithms make use of the prior knowledge and various constraints in real and reciprocal space. In this thesis, new approaches to the problem of direct digital phasing of X-ray diffraction patterns from two-dimensional organic crystals were presented. The phase problem for Bragg diffraction from two-dimensional (2D) crystalline monolayer in transmission may be solved by imposing a compact support that sets the density to zero outside the monolayer. By iterating between the measured stucture factor magnitudes along reciprocal space rods (starting with random phases) and a density of the correct sign, the complex scattered amplitudes may be found (J. Struct Biol 144, 209 (2003)). However this one-dimensional support function fails to link the rod phases correctly unless a low-resolution real-space map is also available. Minimum prior information required for successful three-dimensional (3D) structure retrieval from a 2D crystal XFEL diffraction dataset were investigated, when using the HIO algorithm. This method provides an alternative way to phase 2D crystal dataset, with less dependence on the high quality model used in the molecular replacement method. / Dissertation/Thesis / Doctoral Dissertation Physics 2016

All across physics research, incoherent and coherent light sources are extensively utilized. Especially highly brilliant X-ray sources such as third generation synchrotrons or free-electron lasers have become an invaluable tool enabling experimental techniques that are unique to these kinds of light sources. But these sources have developed to large scale facilities and a demand in compact laboratory scale sources providing radiation of similar quality arises nowadays. This thesis focuses on Traveling-Wave Thomson-Scattering (TWTS) which allows for the realization of ultra-compact, inherently synchronized and highly brilliant light sources. The TWTS geometry provides optical undulators, through which electrons pass and thereby emit radiation, with hundreds to thousands of undulator periods by utilizing pulse-front tilted lasers pulses from high peak-power laser systems. TWTS can realize incoherent radiation sources with orders of magnitude higher photon yield than established head-on Thomson sources. Moreover, optical free-electron lasers (OFELs) can be realized with TWTS if state-of-the-art technology in electron accelerators and laser systems is utilized. Tilting the laser pulse front with respect to the wavefront by half of this interaction angle optimizes electron and laser pulse overlap by compensating the spatial offset between electrons and the laser pulse-front at the beginning of the interaction when the electrons are far from the laser pulse axis. The laser pulse-front tilt ensures continuous overlap between electrons and laser pulse while the electrons cross the laser pulse cross-sectional area. Thus the interaction distance can be controlled in TWTS by the laser pulse width rather than laser pulse duration. Utilizing wide, petawatt class laser pulses allows realizing thousands of optical undulator periods. This thesis will show that TWTS OFELs emitting ultraviolet radiation are realizable today with existing technology for electron accelerators and laser systems. The requirements on electron bunch and laser pulse quality of these ultraviolet TWTS OFELs are discussed in detail as well as the corresponding requirements of TWTS OFELs emitting in the soft and hard X-ray range. These requirements are derived from scaling laws which stem from a self-consistent analytic description of the electron bunch and radiation field dynamics in TWTS OFELs presented within this thesis. It is shown that these dynamics in TWTS OFELs are qualitatively equivalent to the electron bunch and radiation field dynamics of standard free-electron lasers which analytically proves the applicability of TWTS for the realization of an optical free-electron laser. Furthermore, experimental setup strategies to generate the pulse-front tilted TWTS laser pulses are presented and designs of experimental setups for the above examples are discussed. The presented setup strategies provide dispersion compensation, required due to angular dispersion of the laser pulse, which is especially relevant when building compact, high-yield hard X-ray TWTS sources in large interaction angle setups. An example of such an enhanced Thomson source by TWTS, which provides orders of magnitude higher spectral photon density than a comparable head-on interaction geometry, is presented, too

abstract: Scientists have used X-rays to study biological molecules for nearly a century. Now with the X-ray free electron laser (XFEL), new methods have been developed to advance structural biology. These new methods include serial femtosecond crystallography, single particle imaging, solution scattering, and time resolved techniques. The XFEL is characterized by high intensity pulses, which are only about 50 femtoseconds in duration. The intensity allows for scattering from microscopic particles, while the short pulses offer a way to outrun radiation damage. XFELs are powerful enough to obliterate most samples in a single pulse. While this allows for a “diffract and destroy” methodology, it also requires instrumentation that can position microscopic particles into the X-ray beam (which may also be microscopic), continuously renew the sample after each pulse, and maintain sample viability during data collection. Typically these experiments have used liquid microjets to continuously renew sample. The high flow rate associated with liquid microjets requires large amounts of sample, most of which runs to waste between pulses. An injector designed to stream a viscous gel-like material called lipidic cubic phase (LCP) was developed to address this problem. LCP, commonly used as a growth medium for membrane protein crystals, lends itself to low flow rate jetting and so reduces the amount of sample wasted significantly. This work discusses sample delivery and injection for XFEL experiments. It reviews the liquid microjet method extensively, and presents the LCP injector as a novel device for serial crystallography, including detailed protocols for the LCP injector and anti-settler operation. / Dissertation/Thesis / Doctoral Dissertation Physics 2015