ATTAINING DOPPLER PRECISION OF 3 M S-1
ATTAINING DOPPLER PRECISION OF 3 M S-1
复制标题
DOI：
10.1086/133755
复制
发表时间：
1996-06
影响因子：
3.5
通讯作者：
R. P. Butler;G. Marcy;E. G. Williams;C. McCarthy;P. Dosanjh;S. Vogt
中科院分区：
文献类型：
--
作者：
R. P. Butler;G. Marcy;E. G. Williams;C. McCarthy;P. Dosanjh;S. Vogt
目前的光谱技术产生的多普勒频移误差为10至50 m s-1，几乎不足以检测由类彗星和低质量行星引起的反射速度。我们描述了一种技术，产生相对径向速度误差为3 m s-1。该技术利用分辨率为R= 62，000的快速阶梯光栅摄谱仪和在每次曝光中获取整个可见光和近红外光谱的大幅面CCD。星光通过放置在光谱仪入口狭缝处的碘吸收池发送。由此产生的叠加碘线提供了一个基准波长尺度，以测量径向速度变化。碘线的形状传达光谱仪的PSF，以说明光谱仪光学器件和照明在所有时间尺度上的变化。我们通过将恒星光谱与碘光谱相乘并将结果与光谱仪PSF卷积来构建每个观测光谱的模型。模型的自由参数包括波长尺度、光谱仪点扩展函数和恒星多普勒频移。所有的模型参数推导出新的每次曝光和合成的CCD子像素的网格上完成，使用样条函数作为插值预测。我们目前的多普勒测试的太阳，鲸鱼座τ星，和107 PSC，观察与利克和凯克阶梯。所有表现出明显的误差约3米秒-1，维持在几分钟到一年的时间尺度。这种精度与理论上预测的主要来自光子统计的误差一致。
Current spectroscopic techniques yield Doppler-shift errors of 10 to 50 m s-1, barely adequate to detect reflex velocities caused by Jupiter-like and lower-mass planets. We describe a technique which yields relative radial velocity errors of 3 m s-1. This technique makes use of a fast echelle spectrograph at resolution of R=62,000 and a large format CCD which acquires the entire visible and near IR spectrum in each exposure. Starlight is sent through an iodine absorption cell placed at the spectrometer entrance slit. The resulting superimposed iodine lines provide a fiducial wavelength scale against which to measure radial velocity shifts. The shapes of iodine lines convey the PSF of the spectrometer to account for changes in spectrometer optics and illumination on all times scales. We construct a model of each observed spectrum by multiplying a stellar spectrum with an iodine spectrum and convolving the result with the spectrometer PSF. The free parameters of the model include the wavelength scale, spectrometer PSF, and stellar Doppler shift. All model parameters are derived anew for each exposure and the synthesis is done on a grid of CCD sub-pixels, using spline functions as interpolation predictors. We present Doppler tests of the Sun, Tau Ceti, and 107 Psc, observed with the Lick and Keck echelles. All exhibit apparent errors of about 3 m s-1, maintained on time scales of minutes to a year. This precision agrees with the theoretically predicted errors that stem primarily from photon statistics.