SRGz: классификация точечных рентгеновских источников еРОЗИТА в области 1%DESI и калибровка фотометрических красных смещений

Cover Page

Cite item

Full Text

Open Access Open Access
Restricted Access Access granted
Restricted Access Subscription Access

Abstract

Рассматривается популяция точечных рентгеновских источников двухлетнего обзора СРГ/еРОЗИТА в области 1%-го спектроскопического обзора DESI в восточной галактической полусфере (eROSITA-1%DESI-East). Рассматриваемые данные сочетают в себе большую площадь обзора (91.4 кв. градусов) и рекордно высокую полноту (90–95%) спектроскопии оптических компаньонов рентгеновских источников с потоком FX,0.5–2 ≥ 1.5×10–14 эрг с–1 см–2. Мы сравниваем результаты фотометрических (SRGz) и спектральных/астрометрических измерений (DESI EDR, SDSS, HELP, GAIA) классов и красных смещений объектов в зависимости от их рентгеновского потока. Нами отмечается высокая точность фотометрических красных смещений, полученных моделями SRGz для рентгеновских источников в двухлетнем обзоре неба еРОЗИТА (в области покрытия фотометрических обзоров DESI Legacy Imaging Surveys/Pan-STARRS1/SDSS): стандартное отклонение σNMAD ≈ 4% и доля выбросов на уровне n>0.15 = 7–8.5% (для оптических компаньонов рентгеновских источников с FX,0.5–2 ≥ 1.5 × 10–14 эрг с–1 см–2). Прогнозы photo-z рентгеновских источников в SRGz имеют негауссовый характер; качество калибровки PDF(z) важно для точной оценки доверительных интервалов прогноза красных смещений. Предложен новый метод постобработки вероятностных прогнозов photo-z, основанный на двухтемпературной коррекции распределения PDF(z). Подход позволяет значительно улучшить калибровку вероятностных прогнозов и доверительных интервалов фотометрических красных смещений рентгеновских источников еРОЗИТА.

Full Text

Restricted Access

About the authors

А. В. Мещеряков

Институт космических исследований РАН; Московский государственный университет им. М. В. Ломоносва

Author for correspondence.
Email: mesch@cosmos.ru
Russian Federation, Москва; Москва

Г. А. Хорунжев

Институт космических исследований РАН

Email: mesch@cosmos.ru
Russian Federation, Москва

С. А. Воскресенская1

Институт космических исследований РАН; НИУ Высшая школа экономики

Email: mesch@cosmos.ru
Russian Federation, Москва; Москва

П. С. Медведев

Институт космических исследований РАН

Email: mesch@cosmos.ru
Russian Federation, Москва

М. Р. Гильфанов

Институт космических исследований РАН; Институт им. Макса Планка

Email: mesch@cosmos.ru
Russian Federation, Москва; Германия, Гаршинг

Р. А. Сюняев

Институт космических исследований РАН; Институт им. Макса Планка

Email: mesch@cosmos.ru
Russian Federation, Москва; Германия, Гаршинг

References

  1. DESI коллаборация и др. (DESI Collaboration, G. Adame, J. Aguilar, S. Ahlen, S. Alam, G. Aldering, et al.), arXiv e-prints, p. arXiv:2306.06308 (2023).
  2. Алмейда и др. (A. Almeida, S.F. Anderson, M. Argudo-Fernández, C. Badenes, K. Barger, J.K. Barrera-Ballesteros, et al.), Astrophys. J. Suppl. Ser. 267, 44 (2023).
  3. Ананна и др. (T.T. Ananna, M. Salvato, S. LaMassa, C.M. Urry, N. Cappelluti, C. Cardamone, et al.), Astrophys. J. 850, 66 (2017).
  4. Брейман (L. Breiman), Machine Learn. 45, 5 (2001).
  5. Брешиа и др. (M. Brescia, M. Salvato, S. Cavuoti, T.T. Ananna, G. Riccio, S.M. LaMassa, et al.), MNRAS 489, 663 (2019).
  6. Буренин Р.А., Письма в Астрон. журн. 48, 167 (2022) [R.A. Burenin, Astron. Lett. 48, 153 (2022)].
  7. Ванг (C. Wang), arXiv e-prints, p. arXiv:2308.01222 (2023).
  8. Ву, Шен (Q. Wu and Y. Shen), Astrophys. J. Suppl. Ser. 263, 42 (2022).
  9. Гомес и др. (Z. Gomes, M.J. Jarvis, I.A. Almosallam, and S.J. Roberts), MNRAS 475, 331 (2018).
  10. Д’Инсанто, Полстерер (A. D’Isanto and K.L. Polsterer), Astron. Astrophys. 609, A111 (2018).
  11. Дей и др. (A. Dey, D.J. Schlegel, D. Lang, R. Blum, K. Burleigh, X. Fan, et al.), Astron. J. 157, 168 (2019).
  12. Динардо и др. (J. Dinardo, N. Fortin, and T. Lemieux), Econometrica 64, 1001 (1996).
  13. Дэвид (A.P. Dawid), J. Royal Statistic. Soc.: Ser. A (General) 147, 278 (1984).
  14. Карраско Кинд, Бруннер (M. Carrasco Kind and R.J. Brunner), MNRAS432, 1483 (2013).
  15. Коллаборация Euclid и др. (Euclid Collaboration, G. Desprez, S. Paltani, J. Coupon, I. Almosallam, A. Alvarez-Ayllon, et al.), Astron. Astrophys. 644, A31 (2020).
  16. Лин и Джеон (Y. Lin and Y. Jeon), J. Am. Statistic. Associat. 101, 578 (2006).
  17. Люк и др. (B.W. Lyke, A.N. Higley, J.N. McLane, D.P. Schurhammer, A.D. Myers, A.J. Ross, et al.), Astrophys. J. Suppl. Ser. 250, 8 (2020).
  18. Майнсхаузен (N. Meinshausen), J. Machine Learn. Res. 7, 983 (2006).
  19. Медведев П.С., Гильфанов М.Р., Сазонов С.Ю. и др., Письма в Астрон. журн. 49, 35 (2023) [P.S. Medvedev, M.R. Gilfanov, S.Y. Sazonov, R.A. Sunyaev, and G.A. Khorunzhev, Astron. Lett. 48, 735 (2022)].
  20. Мейнзер и др. (A. Mainzer, J. Bauer, T. Grav, J. Masiero, R.M. Cutri, J. Dailey, et al.), Astrophys. J. 731, 53 (2011).
  21. Мещеряков А.В., Глазкова В.В., Герасимов С.В., Машечкин И.В., Письма в Астрон. журн. 44, 801 (2018) [A.V. Meshcheryakov, V.V. Glazkova, S.V. Gerasimov, and I.V. Mashechkin, Astron. Lett. 44, 735 (2018)].
  22. Мещеряков А.В., Борисов В.Д., Хорунжев Г.А. и др., Письма в Астрон. журн. 49, 441 (2023) [A.V. Meshcheryakov, V.D. Borisov, G.A. Khorunzhev, P.A. Medvedev, M.R. Gilfanov, M.I. Belvedersky, et al., Astron. Lett. 49, 359 (2023)].
  23. Мозафари и др. (A. Sadat Mozafari, H. Siqueira Gomes, W. Leao, S. Janny, and C. Gagn´e), arXiv e-prints, p. arXiv:1810.11586 (2018).
  24. Ньюман, Груен (J.A. Newman and D. Gruen), Ann. Rev. Astron. Astrophys. 60, 363 (2022).
  25. Павлинский и др. (M. Pavlinsky, A. Tkachenko, V. Levin, N. Alexandrovich, V. Arefiev, V. Babyshkin, et al.), Astron. Astrophys. 650, A42 (2021).
  26. Платт (J. Platt), Adv. Large Margin Classif. 10 (2000).
  27. Предель и др. (P. Predehl, R. Andritschke, V. Arefiev, V. Babyshkin, O. Batanov, W. Becker, et al.), Astron. Astrophys. 647, A1 (2021).
  28. Райт и др. (E.L. Wright, P.R.M. Eisenhardt, K. Mainzer, M.E. Ressler, R.M. Cutri, T. Jarrett, et al.), Astron. J. 140, 1868 (2010).
  29. Сюняев и др. (R. Sunyaev, V. Arefiev, V. Babyshkin, A. Bogomolov, K. Borisov, M. Buntov, et al.), Astron. Astrophys. 656, A132 (2021).
  30. Танг и др. (C. Tang, D. Garreau, and U. von Luxburg), in S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett (Es.), Adv. in Neural Information Processing Systems, Vol. 31, Curran Associates, Inc. (2018).
  31. Чамберс и др. (K.C. Chambers, E.A. Magnier, N. Metcalfe, H.A. Flewelling, M.E. Huber, C.Z. Waters, et al.), arXiv e-prints, p. arXiv:1612.05560 (2016).
  32. Ширли и др. (R. Shirley, K. Duncan, M.C. Campos Varillas, P.D. Hurley, K. Malek, Y. Roehlly, et al.), MNRAS 507, 129 (2021).
  33. Шмидт и др. (S.J. Schmidt, A.I. Malz, J.Y.H. Soo, I.A. Almosallam, M. Brescia, S. Cavuoti, et al.), MNRAS 499, 1587 (2020).
  34. Шоссидон и др. (E. Chaussidon, C. Y`eche, N. Palanque-Delabrouille, D.M. Alexander, J. Yang, S. Ahlen, et al.), Astrophys. J. 944, 107 (2023).

Supplementary files

Supplementary Files
Action
1. JATS XML
Download
2. Fig. 1. Distribution of eROSITA point X-ray objects in the eRo-1DESI-East survey area by X-ray flux in the 0.5–2 keV range. Distributions for two sites near the north ecliptic pole (eRo-SV3-NEP) and for all other survey sites (eRo-SV3–9) are shown separately. The vertical dashed lines mark the threshold X-ray fluxes for the bright, medium, and faint survey samples: (“bright”) FX,0.5–2 ≥ 4 × 10–14 erg s–1 cm–2, (“medium”) 1.5 × 10–14 ≤ FX,0.5–2 < 4 × 10–14 erg s–1 cm–2, (“faint”) 6 × 10–15 ≤ FX,0.5–2 < < 1.5 × 10–14 erg s–1 cm–2.

Download (238KB)
Indexing metadata
3. Fig. 2. Photometric classification confusion matrices (upper panel) and Precision–Recall plots (lower panel) for three eRosita-1%DESI-East samples: “bright” (left), “medium” (center), and “faint” (right). The Precision–Recall plots show the classification curves of X-ray stars (STAR), X-ray objects with the optical spectrum of a quasar (QSO), and the spectrum of a galaxy (GALAXY). The legend to the curves provides the values ​​of the average accuracy achieved in selecting objects of each class.

Download (378KB)
Indexing metadata
4. Fig. 3. Upper panel: scatterplot of photometric redshifts for three samples of optical companions of X-ray sources. Lower panel: dependence of the normalized deviation of photometric redshift (∆znorm) on the stellar magnitude in the r filter. Circles and diamonds in the plots are objects with spectroscopic measurements from DESI EDR and other surveys, respectively. The region of catastrophic outliers in photo-z measurements is located outside the black dashed lines. The plots show how the accuracy of photo-z (in terms of the spread of ∆znorm and the appearance of outliers) changes depending on the X-ray and optical fluxes of the object.

Download (635KB)
Indexing metadata
5. Fig. 4. Scatterplots of photometric redshifts for eRosita-1%DESI-East X-ray sources for different photometric and spectroscopic classes of extragalactic X-ray sources (discussed in Section 4.1). The plots show only objects with the DESI LIS spectral classification (QSO is the upper panel of the plots, GALAXY is the lower panel of the plots). Photometric quasars are shown on the left, photometric galaxies are shown on the right. In all plots, objects in different ranges of the photo-z estimate reliability parameter (zConf) are shown with different symbols and colors: zConf > 0.6 (crimson triangles), 0.4 ≤ zConf ≤ 0.6 (red circles), 0.3 < zConf ≤ 0.4 (blue pluses), zConf ≤ 0.3 (light blue crosses). Beyond the black dashed lines is the region of catastrophic outliers (∆znorm > 0.15) in photo-z measurements.

Download (707KB)
Indexing metadata
6. Fig. 5. Distributions of deviations ∆z = zsp – zph of SRGz point predictions of photometric redshifts of X-ray sources, normalized to the estimate of the standard deviation σNMAD, for samples of eRosita-1%DESI-East sources in different X-ray flux intervals: “bright” (left), “medium” (center), and “faint” (right). Dashed lines show normal distributions with zero mean and unit variance. It follows from the graphs that the measurements of photometric redshifts of X-ray sources in SRGz have a significantly non-Gaussian character. Both the excess of objects in the wings of the distribution and their significantly asymmetric shape are noteworthy.

Download (214KB)
Indexing metadata
7. Fig. 6. Examples of probabilistic PDF(z) predictions made by the SRGz models for objects in the “bright” (upper panel), “medium” (middle panel), and “faint” (lower panel) samples. The empirical distribution PS(z) is shown as a blue histogram, and PKDE(z) is a dashed blue line in the plots. The calibrated empirical distributions PS(z) are shown as red histograms, and the calibrated kernel density estimates PKDE(z) are shown as a solid red line in the plots. The parameters of the two-temperature calibration model are labeled in the plot titles. The vertical dotted and dashed lines denote the point prediction of the photometric redshift (zph) and the spectral redshift (zsp) of the object, respectively. In the upper plot, the photometric redshift of the object agrees very well with the spectroscopic measurement result. The other two sources show discrepancies between the zph and zsp measurements. The largest discrepancy between the photometric and spectral measurements is demonstrated by the object in the middle panel. Also clearly visible here is the multimodal nature of the photo-z predictions of X-ray quasars for the selected system of broadband filters: PDF(z) has several peaks in this case, one (the most probable) of which follows the photo-z prediction, while the other contains the spectral measurement of the redshift.

Download (563KB)
Indexing metadata
8. Fig. 7. PIT diagrams (top) and QQ diagrams (bottom) illustrate the calibration of the probabilistic predictions of the SRGz photometric redshifts for three samples of eRosita-1%DESI-East X-ray sources: “bright” (left), “medium” (center), and “faint” (right). The probabilistic predictions with perfect calibration correspond to the horizontal straight line in the PIT diagram and the diagonal of the quantile plot (shown as black dashed lines in the plots). The PIT diagram for the probabilistic predictions of SRGz before and after calibration is shown as a filled blue histogram and a solid red line in the upper panel of the plots, respectively. The corresponding QQ diagrams of the SRGz predictions are shown as blue dotted lines (before calibration) and solid red lines (after calibration) in the lower panel of the figure. The values ​​of the parameters of the PDF(z) calibration method are given above the top panel of the graphs.

Download (391KB)
Indexing metadata

Copyright (c) 2024 Russian Academy of Sciences