The efficiency of different spectroscopic techniques are examined through four different approaches: detailed analysis of IR spectra from the APOGEE database and examination of persistence, observing extremely metal-poor stars using the Plaskett telescope at the DAO, three analyses of various applications of machine learning in astronomy, and efficient transmission of light through optical fibres.
Through the first study, the technical effects of persistence in the APOGEE's IR spectra are examined, and a new technique for removing the persistence is introduced. Most of the globular cluster Pal 1's spectra in the APOGEE database are affected by persistence. Therefore, the Pal 1 spectra are corrected for the persistence and their stellar abundances are determined independently from the APOGEE's pipeline, ASPCAP. Our results for the known members of Pal 1 were in a close agreement with the results from Sakari et al. (2011). Comparison between the results from the corrected and the original spectra suggest that the persistence could have a critical effect on the results.
The second study of this thesis focused on observations of extremely metal-poor (EMP) stars from the Pristine survey. Through the DAO-Pristine project, we narrowed down the initial list of the Pristine survey by observing over 50 targets during 25 observing nights. The Ca II triplet absorption lines of the observed targets were examined and used for estimating the metallicity of the objects. Twelve candidate EMP stars with weak Ca II triplet lines are chosen from the observed targets. These candidate EMP stars will be observed with larger telescopes for more accurate determination of their metallicity.
This thesis also presents the result of a threefold analysis for using machine learning techniques in astronomy. The supervised machine learning methods are used for determination of the stellar parameters of stars using their raw spectra, and unsupervised machine learning methods are used for classification of supernovae Type Ia from their calibrated spectra. The supervised analysis of the IR and optical spectra suggested that the StarNet neural network (Fabbro et al. 2017) can predict the stellar parameters of the APOGEE database and synthetic spectra, efficiently and accurately. The effect of persistence in the StarNet's results are examined, and we showed that the persistence does not have a critical effect on the overall performance of the StarNet. In addition, multiple unsupervised machine learning techniques such as K-mean and Self Organizing Maps (SOMs) are used for classification of the supernovae Type Ia spectra. The preliminary results suggest that a minimum of three subclasses of supernovae Type Ia can be found from our data, which are consistent with the previous studies.
Finally, this thesis presents our final results for an optical system we designed for the MSE project. At UVic, we have used the standard collimated beam method, or "ring test," to measure the Focal Ratio Degradation (FRD) of MSE-like fibres. The FRD of the system is determined from the ratio of the Full Width Half Maximum (FWHM) to the radius of the ring. Early ring test results from a sample of MSE-like fibres show an FRD of 3.7%, which meets the MSE science requirement (i.e. FRD < 5% at f/2). Also, we have automated the ring test for fast, repeatable, and efficient measurements of an individual fibre in multi-fibre bundles. Our future tests will include automated non-static fibres in preparation for the MSE build phases. / Graduate
| Identifer | oai:union.ndltd.org:uvic.ca/oai:dspace.library.uvic.ca:1828/9613 |
| Date | 05 July 2018 |
| Creators | Jahandar, Farbod |
| Contributors | Venn, Kim |
| Source Sets | University of Victoria |
| Language | English, English |
| Detected Language | English |
| Type | Thesis |
| Format | application/pdf |
| Rights | Available to the World Wide Web |
Page generated in 0.0024 seconds