In this article I discuss the design of échelle spectrographs for the measurement of radial velocities. I consider the attainment of very high precision, and whether we can do still better than the current generation of instruments, and what problems have been resolved and what problems still remain to be solved in spectrograph design. The article makes special reference to the design principles followed in the Marseille-Geneva instrument ELODIE, for which the goal has been the measurement of stellar radial velocities of high precision using an R4 échelle grating.

The use of fibers to feed spectrographs is a very efficient way to increase the precision of radial velocity measurements. It has already proved to be successful with the very first detection by the ELODIE fiber-fed spectrograph of the planet orbiting the star 51 Peg. The basic key properties of fibers in the very high radial velocity precision measurements context are described in this review. The ELODIE spectrograph is used to illustrate the thorium simultaneous calibration technique. The use and the effects of a double-scrambler to increase the stability of the slit illumination are also discussed.

The fiber-fed echelle spectrograph EMILIE designed for the measurement of stellar radial-velocity changes is presently being laboratory-tested. Using a 204 × 408 mm grating and a 1k × 1k CCD detector, it samples about 50% of the 410−620 nm wavelength range with spectral resolution R ∼ 150 000. This spectrograph is coupled to the telescope via a single 50μm fiber, which accepts 2.7 arcsec from the sky. An automatic guiding system stabilizes the image at the entrance of the fiber. The fiber transmits alternately to the spectrograph the stellar beam and the reference beam. The same pixels of the CCD detector are used alternately for the two spectra, which requires an excellent short term stability. Hence, later on, the spectrograph will be located within a vacuum tank, with active thermal control. The instrument will be set up at the focus of the 1.52-m coudé telescope at the Observatoire de Haute-Provence (OHP) and will undertake programs of asteroseismology and detection of extrasolar-planets.

A new fiber-fed échelle spectrograph (Hercules) is being designed for the 1-m telescope at Mt John University Observatory. The goals are to have a wavelength capability of 380 to 880 nm, covered in a single exposure on a 50-mm square CCD, to have a choice of resolving powers of 35000 or 70000 and to have no moving parts. High precision radialvelocity observations are a major but not the only goal. Designs with both R2 (blaze angle 63.4 deg) and R4 (blaze angle 76 deg) échelle gratings are being considered, in either case with a dimension of 408 mm perpendicular to the grooves.

Stellar spectra have been obtained using a multichannel Fourier Transform Spectrometer (FTS) which incorporates components of the Navy Prototype Optical Interferometer. It is well known that a FTS can provide superior wavelength stability as compared to traditional spectrometers. Unfortunately the FTS traditionally suffers from exceptionally poor sensitivity, which until now has limited its uses to sources with high fluxes and/or those with narrow band emission (e.g. the Sun, nebulae, and laboratory samples). We present stellar observations using a new FTS design which overcomes this sensitivity limitation by using a conventional multichannel spectrometer in conjunction with the FTS system. The signal-to-noise ratio of spectra from our test-bed observations are consistent with the theoretical prediction and show that for N channels the sensitivity scales like N, while the signal-to-noise ratio scales like . With this type of an instrument on a 3-m telescope and 9 000 channels we expect to be able to detect and measure such exciting astrophysical phenomenon as gravitational redshifts from single, main sequence stars and extrasolar planets of terrestrial mass.

The apparent radial velocity of a star, as deduced from wavelength shifts, comprises not merely its true velocity, but also components arising from dynamics in the star’s atmosphere, gravitational redshift, and other effects. For the Sun, such phenomena can be segregated since the relative Sun-Earth motion is known from planetary system dynamics. This is now becoming possible also for other stars, whose true radial motions are determined through space astrometry. A study of the differences between accurate astrometric velocities (from Hipparcos), and precise spectroscopic values (from ELODIE) is in progress. Data for cool stars in the Hyades indicate a tendency of relative blueshifts among earlier main-sequence F-type stars, and in giants. This is theoretically expected: an increased convective blueshift due to the more vigorous convection in F-stars, and a decreased gravitational redshift in giants.

Although the use of telluric lines as wavelength fiducials to measure radial velocities does not achieve as high a precision as other fiducial-imposition techniques, this very convenient technique can be used concurrently with other observing programs to increase the temporal sampling of target stars. We have been carrying out a program to monitor the line-profile variations of early-type non-radial pulsators at the Hα and He I 667.8-nm region. With the rather modest reciprocal dispersion of 1nm/mm and the use of a 4096-element CCD, the telluric lines in the 630-nm region are also available in the observed spectra. We decided to use these telluric lines as wavelength fiducials to monitor bright, latetype stars for radial-velocity variations. As an experiment, we have also decided to reduce the spectra using available simple IRAF tasks to see how high a velocity precision can be achieved with only minor tweaking. The precision certainly would not rival other precise techniques, but the convenience in both the observing and reduction procedure may enable more target stars to be monitored by more observers. Moreover, interesting results can still be obtained with a mere 100 m/s precision. The result for a few late-type stars which also have prior HF velocities will be presented.

The measurement of accurate radial velocities for A- or B-type dwarfs poses a separate category of problems: the small number of suitable lines, the wide wavelength range over which such lines are distributed, thermal broadening, rotational broadening, line distortions caused by rotating spots, spectrum variations due to obvious duplicity, and low-level velocity variations due to undetected companion spectra. Moreover, the fact that A- and B-type spectra fall into very distinct groups, each with its own sub-set of the above problems, means that it may be both difficult and unsatisfactory to specify velocity standards, since intrinsic uncertainties caused by differences in spectral type may exceed the measuring errors. The investigation summarized here into the nature and magnitude of some of these intrinsic errors employs wide spans of high-resolution, high-dispersion spectra of Sirius and Vega as ‘natural’ templates. Attention is also drawn to systematic errors which may arise (a) when modelling a cross-correlation ‘dip’ and (b) whenever a spectrum or a cross-correlation dip is measured in the unavoidable presence of another such spectrum or dip.

We are developing methods to reach high absolute accuracy in spectroscopic radial velocities for stars of different spectral types. The basic idea is to remove the effects from convection and surface gravity that cause large systematic errors, in order to improve the absolute accuracy of radial velocities by one order of magnitude. This paper briefly describes observations and methodology in computing the radial velocities.

We briefly describe the instrumental effects which affect the accuracy of the radial velocity measurements. We have implemented several methods to correct for the instability effects and improve the accuracy of the measurements. These include modifications of the observational strategy and a time-dependent wavelength solution as well as a discussion of the error of the offset from cross-correlation. These methods are applied to observations obtained with the high resolution échelle spectrograph SOFIN mounted at the Cassegrain focus of the alt-azimuth 2.56-m Nordic Optical Telescope, La Palma, Canary Islands.

We present our measurements of radial velocities of two stars suspected to have a substellar companion by using observations made with a cassegrain échelle spectrograph. The stability issues and details of the data reduction are discussed in Ilyin & Duemmler (paper I, these proceedings). The results obtained here show that relatively high accuracy of radial velocity determinations is also attainable for cassegrain spectrographs.

Accuracy levels of metres per second require the concept of ‘radial velocity’ to be examined, in particular with respect to relativistic velocity effects and spectroscopic measurements made inside gravitational fields. Already in a classical (non-relativistic) framework the line-of-sight velocity component is an ambiguous concept. In the relativistic context, the observed wavelength shifts depend e.g. on the transverse velocity of the star and the gravitational potential at the source. We argue that the observational quantity resulting from high-precision radial-velocity measurements is not a physical velocity but a spectroscopic radial-velocity measure, which only for historic and practical reasons is expressed in velocity units. This radial-velocity measure may be defined as cz, where c is the speed of light and z is the observed relative wavelength shift reduced to the solar system barycentre. To first order, cz equals the line-of-sight velocity, but its precise interpretation is model dependent.

We discuss non-spectroscopic (astrometric) ways to determine radial velocities and their potentials in future astrometric missions. Radial-velocity accuracies are presented, based on Hipparcos parallax and proper motion data for several open clusters.

The cross-correlation function (CCF) has become the standard tool for extraction of radial-velocity and broadening information from high resolution spectra. It permits integration of information which is common to many spectral lines into one function which is easy to calculate, visualize and interpret. However, the CCF is not the best tool for many applications where it should be replaced by the proper broadening function (BF). Typical applications requiring use of BFs rather than CCFs involve finding locations of star spots, studies of projected shapes of highly distorted stars such as contact binaries (as no assumptions can be made about BF symmetry or even continuity) and [Fe/H] metallicity determinations (good baselines and avoidance of negative lobes are essential). It is stressed that the CCFs are not broadening functions. This note concentrates on the advantages of determining BFs through the process of linear inversion, preferably accomplished using the singular value decomposition (SVD). Some basic examples of numerical operations are given in the IDL programming language.

Precise absolute radial velocities have been measured for several hundred late-type stars at Mt John University Observatory with the 1-m telescope, fiber-fed échelle spectrograph, and Thomson CCD. Six échelle orders in the green (5000−5600 Å) are used. Many delicate steps have been undertaken in order to maintain exactly the same conditions, both in recording and reducing the spectra, over a period of 18 months. A cross-correlation technique with theoretical spectra computed by R. L. Kurucz has been chosen to determine the absolute radial velocities. Blue sky spectra have been used to monitor systematic zero-point fluctuations from one observing run to another. An additional correlation between the measured velocities and average number of A/D units in the continuum has been discovered and used for fine adjustments, significantly improving the results. A random uncertainty of 10–20 ms−1 has been achieved for stellar spectra having intrinsically constant radial velocities.

Two observing runs have been undertaken at the Dominion Astrophysical Observatory (DAO) using the radial velocity spectrometer (RVS). About 800 observations have been performed with both the K and F masks. All cross-correlation profiles have been divided by the corresponding flux monitor records and full gaussian functions have been fitted in order to determine the profile centers. A number of IAU standard radial velocity stars have been observed, with a typical internal precision for a given star of about 100 m s−1, which is a factor of two or three better than the usually adopted value for the RVS. Zero-point offsets between these observations and the standard values were also investigated.