ScienceProblem: GalaxySpectralAnalysis

PrimaryActor:

Research astronomer

---

ScienceGoal:

To learn about the properties (age, metallicity, star formation histories) of the stellar populations of a representative sample of galaxies, through analysis of their optical spectra.

---

DataSets:

The spectroscopic catalogue (~900,000 galaxy spectra) being produced by the Sloan Digital Sky Survey(SDSS).

---

ProblemDescription:

The optical spectra of galaxies contain a wealth of information about that stars that have been in the galaxy (and its precursors) - not only the current generation that contributes flux directly to the spectra, but also the previous generations, which make their presence felt indirectly, via the metallicities of the younger stars.

Methods now exist (e.g. Heavens et al. 2000) for extracting this information (in a suitable parametric form) from the optical spectrum of a galaxy very quickly, using data compression techniques, which reduce the dimensionality of the problem radically. The coming generation of large spectroscopic galaxy survey - notably the SDSS - provide an opportunity to run such analysis codes on very large samples of galaxies,thereby obtaining a picture of the evolution of the galaxy population as a whole.

---

CurrentSolution:

There is no current solution, as such, because this problem only really arises with the new, very large spectroscopic datasets. The method developed by Heavens et al. (now called MOPED) can be run now on limited numbers of spectra, but the problem will really arise when SDSS delivers a spectroscopic dataset that is too large to be downloaded, in which case MOPED will have to be run at a data centre holding a copy of the catalogue.

---

VOSolution:

The VO solution to the problem of analysing a large spectroscopic catalogue is to allow the uploading of the MOPED code (and the library of model galaxy spectra required to run it) onto a computer at a data centre serving the spectroscopic catalogue.

The spectra are analysed independently, so this problem is ideally solved by having a PC farm attached to the archive server, which would enable a number of spectra to be analysed in parallel. The whole procedure would run something like this:

1. user uploads MOPED code and library of model galaxy spectra onto one node of a PC farm attached to the archive server.
2. that node checks that the code compiles OK, and then exports it to the other nodes in the PC farm.
3. one-by-one, the spectra from the database are extracted and sent to a node on the cluster for analysis using the MOPED code, with results from each run being written to a file or database.

A possible extension to this features the inclusion of near-infrared magnitudes for those galaxies in the Large Area Survey (LAS) of UKIDSS, which might help break age/metallicity degeneracies which cannot be removed by the use of optical data alone. In that case Step 3 would be split, as follows:

3a. for each spectra in the spectroscopic database, see whether it is located in the region of overlap with the UKIDSS LAS.

3b. if so, then try and find a reliable association with a source in the UKIDSS database, and, if found, extract its JHK magnitudes, and pass them, along with the spectrum, to a node of the PC farm.

3c. if not, then extract upper limits at the galaxy's position in the JHK images from the UKIDSS database, and pass them, together with the galaxy spectrum, to a node of the PC farm.

---

KeyReferences:

Alan Heavens (IfA, Edinburgh), private communication

Heavens A.F., Jimenez R., Lahav O., 2000, MNRAS, 317, 965

---

---

GoodStyle: Please add comments below. This area should be used for refinement of the above document. If you want to ask questions or start a dialogue with the author, please use (or create) a topic in the Science Problems Forum. For other ScienceProblems, refer to the ScienceProblemList.
Author: Once the refinements here and comments in the forum die down, perhaps you could rewrite the problem, incorporating the comments and refinements.

---

A good example of user code analsysis of large data sets in a grid space - use as AstroGrid driver

-- NicholasWalton - 17 Apr 2002

---

-- BobMann - 15 Feb 2002