CU team develops SMART software for analyzing space telescope's data

By Bill Steele

For an astronomer, one spectrum is worth a thousand pictures.

A picture shows where things are in the sky. A spectrum shows what they're made of, how big they are, how and where they're moving and sometimes what's between here and there.

Gathered outside of the Space Sciences Building, Sept. 4, are, from left: Don Barry, manager for scientific computing in radio physics and space research, and SMART project team members Sarah Higdon, James Higdon, Daniel Devost and Peter Hall. Frank DiMeo/University Photography

Ironically, the Infrared Spectrograph (IRS) on the Space Infrared Telescope Facility, launched Aug. 25 from Cape Canaveral, Fla., sends back pictures of spectra. To convert these into a form that astronomers can use, a team of Cornell researchers has developed a suite of computer programs collectively known as SMART (Spectroscopic Modeling Analysis and Reduction Tool). SMART has been developed over the past two years by Sarah Higdon, leader of the SMART project, and Daniel Devost, James Higdon and Peter Hall. All are postdoctoral research associates on the Cornell SIRTF/IRS science team, led by Cornell astronomy Professor James Houck. Initially, SMART grew out of the needs of their own astronomical research, but they have expanded their work into a set of tools available to the entire team. Some of it, they carefully point out, builds on earlier work, including tools developed for the Infrared Space Observatory launched by the European Space Agency in 1995 and image-display software developed by Betty Stobie of Arizona University for the Hubble Space Telescope.

SMART offers a couple of hundred different functions to analyze data sent back by the telescope, but the most important is the conversion of spectra from raw images into simple line graphs, called spectrograms, that show the wavelength distribution of light from a star or other target.

If you shine a narrow beam of white light through the right kind of prism, you'll see a rainbow -- each of the many wavelengths of light that make up "white" has been deflected at a slightly different angle, and the eye sees the different wavelengths as different "colors." Although the infrared light detected by SIRTF is invisible to the human eye, it also consists of a wide range of wavelengths that can be spread out to form an invisible rainbow called a spectrum. Instead of a prism, the infrared spectrometer on SIRTF uses a diffraction grating -- many very fine parallel lines etched on a plate -- to spread out the incoming light.

When an atom or molecule is excited by heat or collisions with electrons, it emits light at a few specific wavelengths, determined by its atomic structure. These emissions will show up in the spectrum as bright lines that reveal which elements or compounds are present in a star or glowing cloud of gas. Likewise, dark lines may appear showing that light has been absorbed by some atoms or molecules along the way, perhaps as it passed through a cloud of interstellar gas. These lines will show up on a graph as sharp peaks and valleys.

A second diffraction grating breaks up the IRS spectrum into a series of short, overlapping sections that are then projected across the chip so they can be clearly seen. It's a bit like the original Western Union telegrams: the message was printed on a long, narrow paper tape, then an operator cut the tape into short strips and pasted them one below another on a sheet of paper to form the message delivered to a customer. Part of SMART's job is to join these pieces back into a continuous spectrum.

The image of several short bands is sent back to earth as a series of 16,234 numbers, each representing the intensity of light falling on one pixel of the chip. The numbers are very precise, Devost said. The infrared-sensitive detector literally counts the photons falling on each pixel. "Each photon kicks out an electron, and the circuit counts the electrons," he explained. In the computer running SMART, each image is represented by an "array," a collection of the 16,234 numbers.

The first job for SMART's image processing is to find the bright bands that represent the spectrum and separate them from the background. Because of limitations in the SIRTF optics, the stripes are not perfectly straight. Tests performed on the ground gave the computer programmers a rough outline of where they should be, but there will always be minor variations. The software starts with a guess based on the ground calibrations then uses image-processing techniques to zero in.

There's a slight overlap at the ends of the bands, so the next step is to eliminate the duplication and stitch them together into one long string. In doing all this, the software must take into account possible variations. "We have to correct for space 'weather' [cosmic rays] and small changes in the instrument performance," Sarah Higdon said. In most cases the telescope will take many snapshots of the same target over a period of a minute or two, and the software averages many snapshots to get the final high-definition result.

That result is a series of numbers that represent the intensity of light at each wavelength. The numbers can be plotted on a graph or used for further processing.

The SMART software package includes "template-fitting" programs that can compare a spectrum with known spectra to identify the elements in a source. Other tools can measure the "red shift" that shows the distance and intensity of an object.

"Many galaxies in the distant, early universe are buried in dust and hidden from optical telescopes, but SIRTF can peer in," Sarah Higdon said. "The history of star formation is encoded in their infrared spectra, and SMART will help us tell this story."