基于主成分分析的测光红移方法

Photometric redshift estimation based on PCA

通过对最新的合成星系光谱库GALAEV进行主成分分析(principle component analysis,PCA),发现该光谱库所包含的信息可以充分压缩为3个本征谱,用其可精确地重构真实星系的谱能量分布.结果显示,用PCA得到的3个本征谱可作为模板估计星系的测光红移.我们使用BATC15色测光系统,用合成与真实星系测试了用这一套模板决定的测光红移精度,结果显示对于测光误差约为σm=0.05mag时,测光红移测量误差大约为zσ=0.013,错误率约为0.12%;当红移低于z=1.20时,测量的误差为zσ=0.012,而错误率为0.与传统方法相比,基于PCA的测光红移方法需要的模板谱数量少、运算速度快、测量的准确度较高.利用该方法可以通过对各种测光系统及其组合进行测试,选用尽可能少的中带滤光片,结合宽带测光,在保证科学目标要求的测光红移精度前提下,可最大限度地减少望远镜的观测时间.

PCA （principle component analysis） was employed to a newly released spectral library of synthetic galaxies, GALAEV. It was found that the library could be sufficiently compressed to the first 3 eigen-spectra. The spectral energy distribution （SED） of observed galaxies could be well reconstructed by these 3 eigen-spectra, indicating that they were good templates to estimate photometric redshift of galaxies. Both synthetic and real galaxies were used to estimate the accuracy of derived redshifts if the BATC 15-color photometric system was adopted. Extensive tests show that, for a given redshift range of z-0-2 and a photometric uncertainty of t0.05 mag, the 1-σ error of the estimated photometric redshift is 0. 013, and the catastrophic percentage of the photometric redshift determination is about 0.12%. For a redshift range of z - 0-1. 2, the uncertainty of photometric redshift decreases to σz = 0. 012 and the catastrophic percentage vanishes. Using the present method, the accuracy in photometric redshift determination is higher than conventional techniques, while the number of galaxy templates is, to the greatest extent, reduced. Using the technique presented, we can perform extensive tests and select as few as possible intermediate band filters with appropriate filter profiles. Incorporating properly selected broad-band filters, the new photometric system will ensure the accuracy of the derived photometric redshift, and in the mean time reduce the observation time.