An X-ray spectral classification algorithm with application to young stellar clusters

S. M. Hojnacki, J. H. Kastner, G. Micela, E. D. Feigelson, S. M. LaLonde

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

A large volume of low signal-to-noise, multidimensional data is available from the CCD imaging spectrometers aboard the Chandra X-Ray Observatory and the X-Ray Multimirror Mission (XMM-Newton). To make progress analyzing this data, it is essential to develop methods to sort, classify, and characterize the vast library of X-ray spectra in a nonparametric fashion (complementary to current parametric model fits). We have developed a spectral classification algorithm that handles large volumes of data and operates independently of the requirement of spectral model fits. We use proven multivariate statistical techniques including principal component analysis and an ensemble classifier consisting of agglomerative hierarchical clustering and A-means clustering applied for the first time for spectral classification. The algorithm positions the sources in a multidimensional spectral sequence and then groups the ordered sources into clusters based on their spectra. These clusters appear more distinct for sources with harder observed spectra. The apparent diversity of source spectra is reduced to a three-dimensional locus in principal component space, with spectral outliers falling outside this locus. The algorithm was applied to a sample of 444 strong sources selected from the 1616 X-ray emitting sources detected in deep Chandra imaging spectroscopy of the Orion Nebula Cluster. Classes form sequences in NH, AV, and accretion activity indicators, demonstrating that the algorithm efficiently sorts the X-ray sources into a physically meaningful sequence. The algorithm also isolates important classes of very deeply embedded, active young stellar objects, and yields trends between X-ray spectral parameters and stellar parameters for the lowest mass, pre-main-sequence stars.

Original languageEnglish (US)
Pages (from-to)585-598
Number of pages14
JournalAstrophysical Journal
Volume659
Issue number1 I
DOIs
StatePublished - Apr 10 2007

Fingerprint

x rays
loci
Orion nebula
pre-main sequence stars
outlier
imaging spectrometers
XMM-Newton telescope
principal components analysis
classifiers
young
falling
principal component analysis
spectrometer
newton
observatory
spectroscopy
accretion
charge coupled devices
observatories
trends

All Science Journal Classification (ASJC) codes

  • Astronomy and Astrophysics
  • Space and Planetary Science

Cite this

Hojnacki, S. M. ; Kastner, J. H. ; Micela, G. ; Feigelson, E. D. ; LaLonde, S. M. / An X-ray spectral classification algorithm with application to young stellar clusters. In: Astrophysical Journal. 2007 ; Vol. 659, No. 1 I. pp. 585-598.
@article{a5fc5cebaa8a400b9e04d69ffc0fad45,
title = "An X-ray spectral classification algorithm with application to young stellar clusters",
abstract = "A large volume of low signal-to-noise, multidimensional data is available from the CCD imaging spectrometers aboard the Chandra X-Ray Observatory and the X-Ray Multimirror Mission (XMM-Newton). To make progress analyzing this data, it is essential to develop methods to sort, classify, and characterize the vast library of X-ray spectra in a nonparametric fashion (complementary to current parametric model fits). We have developed a spectral classification algorithm that handles large volumes of data and operates independently of the requirement of spectral model fits. We use proven multivariate statistical techniques including principal component analysis and an ensemble classifier consisting of agglomerative hierarchical clustering and A-means clustering applied for the first time for spectral classification. The algorithm positions the sources in a multidimensional spectral sequence and then groups the ordered sources into clusters based on their spectra. These clusters appear more distinct for sources with harder observed spectra. The apparent diversity of source spectra is reduced to a three-dimensional locus in principal component space, with spectral outliers falling outside this locus. The algorithm was applied to a sample of 444 strong sources selected from the 1616 X-ray emitting sources detected in deep Chandra imaging spectroscopy of the Orion Nebula Cluster. Classes form sequences in NH, AV, and accretion activity indicators, demonstrating that the algorithm efficiently sorts the X-ray sources into a physically meaningful sequence. The algorithm also isolates important classes of very deeply embedded, active young stellar objects, and yields trends between X-ray spectral parameters and stellar parameters for the lowest mass, pre-main-sequence stars.",
author = "Hojnacki, {S. M.} and Kastner, {J. H.} and G. Micela and Feigelson, {E. D.} and LaLonde, {S. M.}",
year = "2007",
month = "4",
day = "10",
doi = "10.1086/512232",
language = "English (US)",
volume = "659",
pages = "585--598",
journal = "Astrophysical Journal",
issn = "0004-637X",
publisher = "IOP Publishing Ltd.",
number = "1 I",

}

An X-ray spectral classification algorithm with application to young stellar clusters. / Hojnacki, S. M.; Kastner, J. H.; Micela, G.; Feigelson, E. D.; LaLonde, S. M.

In: Astrophysical Journal, Vol. 659, No. 1 I, 10.04.2007, p. 585-598.

Research output: Contribution to journalArticle

TY - JOUR

T1 - An X-ray spectral classification algorithm with application to young stellar clusters

AU - Hojnacki, S. M.

AU - Kastner, J. H.

AU - Micela, G.

AU - Feigelson, E. D.

AU - LaLonde, S. M.

PY - 2007/4/10

Y1 - 2007/4/10

N2 - A large volume of low signal-to-noise, multidimensional data is available from the CCD imaging spectrometers aboard the Chandra X-Ray Observatory and the X-Ray Multimirror Mission (XMM-Newton). To make progress analyzing this data, it is essential to develop methods to sort, classify, and characterize the vast library of X-ray spectra in a nonparametric fashion (complementary to current parametric model fits). We have developed a spectral classification algorithm that handles large volumes of data and operates independently of the requirement of spectral model fits. We use proven multivariate statistical techniques including principal component analysis and an ensemble classifier consisting of agglomerative hierarchical clustering and A-means clustering applied for the first time for spectral classification. The algorithm positions the sources in a multidimensional spectral sequence and then groups the ordered sources into clusters based on their spectra. These clusters appear more distinct for sources with harder observed spectra. The apparent diversity of source spectra is reduced to a three-dimensional locus in principal component space, with spectral outliers falling outside this locus. The algorithm was applied to a sample of 444 strong sources selected from the 1616 X-ray emitting sources detected in deep Chandra imaging spectroscopy of the Orion Nebula Cluster. Classes form sequences in NH, AV, and accretion activity indicators, demonstrating that the algorithm efficiently sorts the X-ray sources into a physically meaningful sequence. The algorithm also isolates important classes of very deeply embedded, active young stellar objects, and yields trends between X-ray spectral parameters and stellar parameters for the lowest mass, pre-main-sequence stars.

AB - A large volume of low signal-to-noise, multidimensional data is available from the CCD imaging spectrometers aboard the Chandra X-Ray Observatory and the X-Ray Multimirror Mission (XMM-Newton). To make progress analyzing this data, it is essential to develop methods to sort, classify, and characterize the vast library of X-ray spectra in a nonparametric fashion (complementary to current parametric model fits). We have developed a spectral classification algorithm that handles large volumes of data and operates independently of the requirement of spectral model fits. We use proven multivariate statistical techniques including principal component analysis and an ensemble classifier consisting of agglomerative hierarchical clustering and A-means clustering applied for the first time for spectral classification. The algorithm positions the sources in a multidimensional spectral sequence and then groups the ordered sources into clusters based on their spectra. These clusters appear more distinct for sources with harder observed spectra. The apparent diversity of source spectra is reduced to a three-dimensional locus in principal component space, with spectral outliers falling outside this locus. The algorithm was applied to a sample of 444 strong sources selected from the 1616 X-ray emitting sources detected in deep Chandra imaging spectroscopy of the Orion Nebula Cluster. Classes form sequences in NH, AV, and accretion activity indicators, demonstrating that the algorithm efficiently sorts the X-ray sources into a physically meaningful sequence. The algorithm also isolates important classes of very deeply embedded, active young stellar objects, and yields trends between X-ray spectral parameters and stellar parameters for the lowest mass, pre-main-sequence stars.

UR - http://www.scopus.com/inward/record.url?scp=34248182550&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34248182550&partnerID=8YFLogxK

U2 - 10.1086/512232

DO - 10.1086/512232

M3 - Article

AN - SCOPUS:34248182550

VL - 659

SP - 585

EP - 598

JO - Astrophysical Journal

JF - Astrophysical Journal

SN - 0004-637X

IS - 1 I

ER -