| This dissertation includes three main parts: studies on coherence requirements for the diffraction microscopy experiments, ice formation on frozen-hydrated sample during data collection, and centering of the diffraction data sets. These three subjects are all in support of our groups overall goal of high resolution 3D imaging of frozen hydrated eukaryotic cells via x-ray diffraction microscopy. X-ray diffraction microscopy requires coherent illumination. However, the actual degree of coherence at some beamlines has never been tested. In research on coherence, our first aim is to determine the transverse coherence width at the sample plane at BL 9.0.1 at the Advanced Light Source in Lawrence Berkeley National Laboratory. An analytical calculation of the coherence at the sample plane is presented. Experimental diffraction patterns of pinhole-pair samples were also taken at the beamline to determine the coherence. Due to the irregular shape of the pinholes and other optics complexity, it was very difficult to fit the data with known theoretical equations as it was traditionally done with 1D data. However, we found out that the auto-correlation function shows clearly three spots. Theoretical calculation have been carried out to show that the degree of coherence can be obtained from the intensities of the three spots. These results are compared with the results from the analytical calculation. We then perform a simulation, showing the required transverse coherence width for reconstructing samples with a given size.The last part of the research is based on simulations and a real data set collected on beamline 9.0.1 at the ALS in Berkeley. In X-ray diffraction microscopy, one of the major challenges when processing the data is to accurately determine the true center of the recorded data that is, the zero spatial frequency position. Simulations of reconstructing shifted data show that if the center of a 2D diffraction pattern is shifted by more than 3 pixels from its true center, the positivity constraint to the phase, which otherwise might be applied to improve the convergence of the reconstruction algorithm, cannot be imposed. Moreover, the phase unwrapping problem may appear during the reconstruction. These issues undermine the quality of the reconstruction of 2D data. Furthermore, the individual shift in each 2D pattern will lead to errors when assembling a 3D diffraction data cube, making the 3D reconstruction very difficult. We developed a method which uses power spectra of the partial diffraction pattern to pre-align the data. A reconstruction without severe phase unwrapping can then be obtained from the pre-aligned data. Next, the precise zero spatial frequency position can be found by examining the linear ramp present in the reconstructed phase. This method was applied to a freeze-dried yeast data set to show that this approach is effective with experimental data.Ice accumulation has been a major problem in X-ray diffraction microscopy with frozen hydrated samples. Since the ice structure is different from point to point, we cannot subtract the scattering from ice, nor assume a completely "empty" region outside the finite support constraint area as required for reconstruction. Ice forms during the sample preparation and transfer. However, from the tests we did in September 2007, we found that the ice layer thickens significantly during the data collecting process. One of the tests we did was putting a dry room-temperature grid into the beam, cooling it down to liquid nitrogen temperature, and then collecting the diffraction pattern of it over time. This test showed that, after the cold grid remained in the chamber for a while, a ring could be observed in the diffraction pattern. The time necessary for this ring to be visible is highly dependent on the pressure and vacuum history of the chamber. We will discuss how the chamber pressure influences the ice accumulation rate, how an anti-contamination device can help to reduce the rate, and how this ring forms. |
