| In comparison to visible light and electrons, X-ray ha s shorter wavelength and longer penetration depth, which may be an ideal light source for high resolution imaging without sample section. With the development of third synchrotron radiation facility, much brighter and coherent x-ray light sources are available. The construction of forth generation synchrotron radiation facility-X-ray free electron laser (XFEL), the brightness of the X-ray light source is further improved and the full coherent X-ray light source is possible. The more excellent properties of the XFEL is the ultra-fast pulse with femtosecond duration, which has potential applications in radiation damage free imaging, dynamic imaging and live cell imaging. The traditional X-ray microscopes are mostly based on the X-ray focus devices, such as K-B mirrors and Fresnel zone plate. But compared with electrons and visible light, X-rays are more difficult to be focused and the focusing devices fabrications are complex. The imaging resolution are mainly restricted by the focusing devices. Taking advantage of improvement of the high brightness and coherence of the synchrotron radiation X-ray light sources, the coherent x-ray diffraction imaging, extended from the x-ray crystallography, is a lensless imaging method, which is free of focusing devices. Thought it is extended from the X-ray crystallography, the samples are not limited to crystals and the high-resolution quantitative imaging of the non-crystalline samples structures is also available. By illuminating of the sample with a coherent x-ray beam, the coherent diffraction pattern of the sample could be recorded by a two-dimensional detector in the far field. By iterating the diffraction patterns between the reciprocal space and real space with a phase retrieval and reconstruction algorithm, the un-recorded phases of diffraction patterns could be recovered and the high-resolution images of the samples could be reconstructed. Because of free of X-ray focus devices during the imaging process, the theoretical resolution is only related to the x-ray wavelength and diffraction angles. Theoretically an atomic resolution is achievable.For now, CDI has been widely used in biology, physics and materials science. More CDI methods were extended from the plane wave CDI, such as Bragg-CDI, Scan-CDI (Ptychography), Fresnel-CDI, etc. By focusing the X-ray beam in front of the sample to increase the x-ray brightness, a 2D reconstruction image of silver nanocube with 2nm resolution was achieved using the third synchrotron radiation facility. With the XFEL-CDI single pulse imaging, a 3D structure of the gold nanocrystal with 5.5nm resolution was performed. But due to the small coherent scattering cross-section of the biomaterials, it is difficult to get a high signal-to-noise ratio diffraction patterns for the low scattering biomaterials. However, the resolution of the reconstructed images is determined by the diffraction signals. Increasing the exposure time could increase the diffraction signal in some sense, but the radiation damage would be a serious problem for biomaterials. XFEL-CDI single ptulse imaging could record the diffraction pattern from the sample before the radiation damage occurs and was regarded as one promising radiation free imaging method. At present, even with the ultra-high bright XFEL single pulse, the resolution of the biomaterials is not high enough. How to improve the imaging resolution of biomaterials with CDI is a scientific problem for researchers. At the same time, a noise robust^and accurate phase retrieval and image reconstruction algorithm is very important for biomaterials to retrieve the precise phase information and get the high resolution reconstruction images.As the diffraction intensity of CDI measured diffraction patterns come from the elastic scattering between the x-ray photons and the sample electrons, theoretically, it is possible to calculate the sample absolute electron density from the diffraction patterns. Base on the electron density, more information of the sample could be extracted, such as component, hided structures. By combining with tomography, high-resolution 3D quantitative images of the samples could be reconstructed from the multiple diffraction patterns, recorded by rotating the sample without the section of the sample.Base on the advantages of CDI and the scientific problems for low scattering biomaterials to overcome, here, we focus on the resolution enhancement of the low scattering materials and 3D high-resolution imaging of biomaterials both with synchrotron radiation and XFEL based CDI in this dissertation:(1) By optimizing the data collecting methods, we quantitatively analyzed the structures of magnetotactic bacteria (MTB) with high resolution. During the data collection, we recorded the low-resolution diffraction pattern and high-resolution diffraction pattern of the sample separately by changing the exposure time and moving the beamstop. After merging the low-resolution and high-resolution diffraction patterns, the missing data in the center of the diffraction pattern was confined in the central speckle and the signal-to-noise ratio of the diffraction pattern at high spatial frequency was improved. For the reconstruction, a precise tight support was employed to increase the convergence speed and retrieved phase accuracy. Finally, a 2D projection of the MTB was achieved with an 18.6nm half period resolution. Depending on the incident x-ray intensity and diffraction pattern intensity, for the first time, we calculated the average mass density of the MTB to be 1.19g/cm3. Some inner cellular structures of the MTB were identified, including the magnetosomes chains.(2) The diffraction signal and reconstruction image resolution enhanced XFEL-CDI single pulse imaging methods was demonstrated using the staphylococcus aureus (S. aureus) as a model by labeling with and without the gold nanoclusters. With the XFEL-CDI single pulse imaging, the diffraction patterns could be recorded before the radiation damage occurred and radiation damage problem for biomaterials was overcome. The signal intensity and the power spectrum density (PSD) of the labeled and control samples shown that, by labeling with gold nanoclusters, the diffraction signals could be increased obviously and the cutoff spatial frequency of the diffraction patterns was increased by around 2 times. With the tight support, the images of the labeled and control samples were reconstructed from the diffraction patterns. The phase retrieval transfer function (PRTF) of the reconstruction images shows that, by labeling with gold nanoclusers, the reconstruction resolution of labeled sample was enhanced by a factor of 2.6. A further discussion of the enhancement of diffraction signal and reconstruction resolution was also conducted.(3) A noise and missing data robust guided oversampling smoothness (GOSS) phase retrieval and reconstruction algorithm was built. The GOSS algorithm is created by applying the denoising technique of oversampling smoothness (OSS) algorithm and the searching optimal solution theory of guided hybrid input-output (GHIO) algorithm to the hybrid input-output algorithm (HIO). The simulated diffraction patterns with different noise levels and different centre missing data size were reconstructed using HIO, GHIO, OSS and GOSS separately. By comparing these reconstruction results, GOSS algorithm has a similar performance at the denoising with OSS algorithm, but have^much better performance at missing data robust, reconstruction consistence and reconstruction resolution than HIO, GHIO and OSS. The reconstruction results of the experimental diffraction pattern of Candida albicans with HIO, GHIO OSS and GOSS shown that the GOSS algorithm also had a good performance for the experimental diffraction patterns. Compared with HIO, GHIO jand OSS, the GOSS reconstruction result had the best imaging resolution and more cellular structures could be identified.(4) The 3D high resolution structure of the mineralized fish bone was studied with synchrotron radiation based CDI. The diffraction patterns at 0°,45° and -45° of the high mineralized fish bone and 27 diffraction patterns between -69.4° to 69.4° of low mineralized fish bone were collected by rotating the samples depending on the equally sloped rotation angles. The 2D projection reconstructions from these diffraction patterns were performed with GOSS algorithm with tight supports that had a similar profile but a few pixels larger with the sample. The highly mineralized fish bone reconstruction half period resolution was better than 15nm and the low mineralized one is better than 26nm. The 3D high resolution of the low mineralized fish bone was reconstructed from the 27 projections by using of the precise Fourier transform based equally sloped tomography (EST) 3D reconstruction algorithm. Compared with the 2D reconstruction results, more information could be identified, benefit from the elimination of thickness effects from the 3D reconstruction results. |
PDF Full Text Request