ASTR 511 (O'Connell) Lecture Notes


2dF on AAT

Two Degree Field Fiber-Fed Spectrograph
on the AAT (J. Pogson)

In Lecture 8 we discussed the design of optical spectrographs. Here, we show some examples of state-of-the-art spectrographs, which come in a wide variety of forms. Classical spectrograph designs employed circular or long-slit entrance apertures, intended for a single target or a 1-D slice through a single extended object. But this is inefficient in the case of fields containing many comparable targets, and modern designs usually feature multiple entrance apertures to make full use of the spectrograph's focal plane.

Multi-object spectrographs are among the most sophisticated and expensive instrumentation used by astronomers today, costing upwards of $10M on a large ground-based telescope. Their primary application has been to low-resolution spectroscopy of large samples of faint galaxies. At the other end of the design regime, the very high resolution, single-object spectrographs have been essential to the discovery and exploration of exoplanets.


Fiber-fed spectrographs use bundles of optical fibers to transfer light from arbitrary positions in the focal plane to the input of the spectrograph. With care to minimize attenuation, fiber lengths can be up to 10's of meters -- in order to feed a bench spectrograph external to the telescope, for instance. Fiber systems typically offer small (2-4 arcsec) entrance apertures, with a single fiber assigned to each target. There is an "exclusion zone" around each fiber, within which another fiber cannot be placed. Fibers must be repositioned with high precision for each new field. This is usually done by mechanical robots. In most designs, individual fiber apertures are clamped magnetically on a flat focal-plane plate. The Sloan Digital Sky Survey (SDSS) uses specially-drilled "plug plates," into which each fiber is inserted by hand. Output of the fiber unit is usually a linear (slit-like) array at the spectrograph input, yielding a fixed position on the focal plane for each spectrum.


Hectospec Focal Surface

  • SDSS-I Spectrograph (2.5-m)

      Cassegrain mounting

      Two moderate resolution (R ~ 2000) spectrographs. Separate red and blue light cameras and detectors in each; incoming beam is split by a diagonal dichroic mirror (red light transmitted, blue reflected). 2048x2048 CCD detectors. Wavelength coverage: 3800-9200 Å.

      640 fibers (320 for each spectrograph), manually plugged into precision drilled aperture plate covering a 3 degree FOV. Minimum fiber separation is 55 arcsec. Fiber diameter on sky: 3 arcsec.

SDSS Fibers

20-fiber cartridge, SDSS

  • SDSS-III APOGEE Spectrograph (2.5-m)

      UVa-designed, bench-mounted, R ~ 22,500, infrared (1.5-1.7 µ) spectrograph; 3-segment mosaic VPH grating; 3 2048x2048 HgCdTe detectors. Intended to determine temperatures, gravities, and 15 elemental abundances in a large sample (over 150,000) of Galactic stars with minimal interference from extinction from interstellar dust (which is small in the IR).

      Infrared operation requires a cryogenically-cooled, evacuated enclosure; significantly complicates design.

      300 40-m long fibers run from plugplates at the SDSS telescope Cassegrain focus to a separate building holding the spectrograph.


APOGEE Spectrograph


Aperture-plate spectrographs use small apertures in a focal-plane mask at the spectrograph entrance to transmit light from selected targets. Usually, computer-controlled devices (mechanical cutters, lasers) are used to cut apertures in a thin, shaped, metallic mask. Early designs used photographic masks. The apertures can be of arbitrary shape and length within overall constraints set by the spectrograph focal plane, but they normally have small widths in the direction of spectral dispersion. The distribution of spectra in the focal plane depends on the distribution of targets in the field.

The main operational problem is to avoid overlap of spectra and to maximize use of the detector area in a given field; this requires special optimizing software. In principle, aperture plate designs should have better throughput, better sky background subtraction, and better flux calibration than fiber designs. Fiber designs can accommodate more targets, however, because the output format on the detector is fixed and optimally packed.

IMACS Prism Slit Mask


An integral field unit (IFU) produces distinct spectra for many contiguous elements in a given compact field. Powerful for the study of extended objects like globular clusters or nearby galaxies. Relative aperture positions and sizes are fixed and generally cover a square/retangular area. IFU's have been designed using fiber bundles, lenslet arrays, and configurable microaperture or micromirror arrays.


A "grism" is a prism with a transmission grating mounted on its entrance face. This combination produces a series of spectra of each source in the focal plane, with the first and second orders of most interest. There is also a zeroth-order image of each source, which provides a means to determine the order positions and wavelength scales for that source. Grisms are used only for low resolution spectroscopy.

Grisms can be deployed in the same way as a filter, without a collimator, ahead of the focal plane of an imaging camera. Grism spectrographs are normally operated "slitlessly," without focal-plane apertures, so they image the entire field. This captures spectra of any object in the field, which is their primary advantage. The main operation/data-reduction complication is the overlap of spectra from different sources or even different parts of the same extended source. To mitigate overlap, spectra are typically taken with the spectrograph oriented in multiple (say up to 5) different position angles on the sky. Even where there is no object overlap, the superposed spectrum of the background sky limits signal-to-noise. Data reduction pipelines are complex (e.g. see Pirzkal 2017).

  • HST Wide Field Camera 3 IR Grism Spectroscopy

      WFC3 carries two IR grisms, with wavelength ranges of 8000-11500 Å and 10750-17000 Å, respectively. Corresponding resolutions are 210 and 130. The imaging field is about 130 arcsec square. The grism mode is mainly used to obtain spectral energy distributions of high redshift sources, where the information-rich UV/optical spectrum is shifted to the near-IR. A sample data frame with high redshift emission line sources identified is shown at the right. An important advantage of IR grisms used on space telescopes is that they avoid the tremendously bright IR night sky emission lines in the Earth's atmosphere. JWST will also carry grisms.

WFC3 IR Grism Frame


The most conspicuous use of high precision spectroscopy has been in the detection of extra-solar planets through stellar reflex Doppler shifts, where velocity differences of order 5 m/s must be measured. Requires both high spectral resolution and great mechanical/optical stability. Suitable designs employing digital detectors have been around since the late 1970's (e.g. Campbell et al. 1988) but were not energetically exploited until the surprising detection of a Jupiter mass planet in a sub-AU orbit made in 1995 by Mayor & Queloz at Haute-Provence Observatory.

  • Marcy-Butler Technique

      Cross-dispersed echelle spectrometer, R ~ 60000

      Employs a gaseous iodine cell at the entrance slit to impress a calibration absorption spectrum on each stellar spectrum taken (see sample at right). The dense molecular spectrum yields ~10 wavelength standard lines per Å

      The calibration signal passes through the optics in exactly the same way as stellar light and simultaneously with it

      Users must perform a cross-correlation analysis on a large number of spectral segments of the star+iodine spectrum covering ~ 800 Å to determine the wavelength shift of a target star

      SNR ~ 200 in flux yields a velocity precision ~ 3 m/s. Since world-class athletes can achieve ~ 10 m/s, we can now detect stars moving at a human pace.

Sample Hi-Res Iodine Cell Spectrum

To date, most exoplanets have been first identified by the transit eclipse method, especially from the Kepler mission in space. However, high resolution spectroscopy remains essential for verifying the identifications using the Doppler technique and for determining the physical characteristics of the planets and their parent stars.


Previous Lecture Lecture Index Next Lecture

Last modified September 2017 by rwo

Grism data frame from Pat McCarthy. Text copyright © 2001-2017 Robert W. O'Connell. All rights reserved. These notes are intended for the private, noncommercial use of students enrolled in Astronomy 511 at the University of Virginia.