http://lib.ulg.ac.be http://matheo.ulg.ac.be

The flaring activity of pre-main sequence stars in very young open clusters

Auteur : Nelissen, Marie Promoteur(s) : Rauw, Gregor Faculté : Faculté des Sciences Diplôme : Master en sciences spatiales, à finalité approfondie Année académique : 2016-2017 URI/URL : http://hdl.handle.net/2268.2/2504

Avertissement à l'attention des usagers :

Tous les documents placés en accès ouvert sur le site le site MatheO sont protégés par le droit d'auteur. Conformément aux principes énoncés par la "Budapest Open Access Initiative"(BOAI, 2002), l'utilisateur du site peut lire, télécharger, copier, transmettre, imprimer, chercher ou faire un lien vers le texte intégral de ces documents, les disséquer pour les indexer, s'en servir de données pour un logiciel, ou s'en servir à toute autre fin légale (ou prévue par la réglementation relative au droit d'auteur). Toute utilisation du document à des fins commerciales est strictement interdite.

Par ailleurs, l'utilisateur s'engage à respecter les droits moraux de l'auteur, principalement le droit à l'intégrité de l'oeuvre et le droit de paternité et ce dans toute utilisation que l'utilisateur entreprend. Ainsi, à titre d'exemple, lorsqu'il reproduira un document par extrait ou dans son intégralité, l'utilisateur citera de manière complète les sources telles que mentionnées ci-dessus. Toute utilisation non explicitement autorisée ci-avant (telle que par exemple, la modification du document ou son résumé) nécessite l'autorisation préalable et expresse des auteurs ou de leurs ayants droit. The flaring activity of pre-main sequence stars in very young open clusters

Nelissen Marie Second year of the Space Sciences Master’s degree at the University of Liege` Master’s thesis director: Rauw Gregor Academic year 2016-2017

Contents

1 Introduction 3 1.1 Pre-main sequence stars...... 3 1.2 The open cluster NGC 6530...... 6 1.3 The XMM-Newton satellite...... 7 1.4 Summary...... 8

2 Processing 10 2.1 Lists of events...... 10 2.2 EPIC cameras...... 11 2.3 SAS processing...... 13

3 Source detection 17 3.1 Detection on combined images...... 17 3.2 Detection verification...... 21 3.3 Detection on individual observations...... 21 3.4 Three colors image...... 22

4 Correlations 24 4.1 Catalogs...... 24 4.2 Correlation radius...... 25

5 Rotation rates 29

6 Hertzsprung-Russell diagram 31 6.1 Color-magnitude diagram...... 31 6.2 Hertzsprung-Russell diagram...... 32

7 Inter-pointing variability 36 7.1 Correlation...... 36 7.2 Chi squared...... 37 7.3 Probability...... 37

8 Lightcurves 40 8.1 Extraction...... 40 8.2 Analysis...... 41

1 8.3 Intra-pointing variability...... 45 8.4 Comparison with other studies...... 48 8.4.1 Loop sizes...... 48 8.4.2 Flaring frequency...... 48

9 Spectra 51 9.1 Extraction...... 51 9.2 Spectral fits...... 51 9.3 Comments on flaring sources...... 56

10 Conclusions 57

2 Chapter 1

Introduction

1.1 Pre-main sequence stars

For stars of mass similar to our Sun, their formation begins with a molecular cloud. When a portion of this cloud collapses - due to its own gravity1 - into several pieces, these pieces then form protostars surrounded by accretion disks. The establishment of a hydrostatic equilibrium in the core of the future star corresponds to the transition from the protostar phase to the pre-main sequence phase. Initially, the core of the pre-main sequence star (or PMS) is entirely hidden by the portion of molecular cloud from which it was formed. The gas surrounding the celestial body then dissipates or is accreted by the PMS object and the star is revealed (Dupret 2015-2016; Schulz 2012).

Figure 1.1: Star formation (”Stars, Supernovas and Neutron Stars - Black Holes and Wormholes - The Physics of the Universe”, 2017).

1Molecular clouds have a high mass and therefore a significant auto-gravity. This means that they are in an unstable equilibrium. Forces due to the interstellar magnetic field, internal pressure gradients and centrifugal effects due to rotation work to maintain this equilibrium. Instability factors are a transit in a high-density zone or a shock wave due to a supernova explosion.

3 1.1. PRE-MAIN SEQUENCE STARS CHAPTER 1. INTRODUCTION

Before reaching the main sequence, pre-main sequence stars move along the Hayashi tracks, corre- sponding to models of entirely convective stars in hydrostatic equilibrium (but not necessarily in thermal equilibrium, unlike main sequence stars) of constant mass and chemical composition. They are represented in the Hertzsprung-Russell diagram as quasi vertical curves, each track corresponding to a specific stellar mass (see Fig. 1.2)(Dupret 2015-2016).

Figure 1.2: The Hayashi tracks (Dupret 2015-2016).

Protostars and PMS can be classified into different categories, based on their IR excess2 and Hα emis- sion3 (see Fig. 1.3)(Feigelson & Montmerle 1999; Rauw 2016-2017; Schulz 2012):

The Class 0 sources or infalling protostars are young protostars deeply embedded in their nascent molec- ular cloud (which is collapsing toward the central regions, accreting onto the protostar and forming a disk). They are invisible in the optical domain and rarely detected in the X-ray domain due to heavy circumstellar extinction.

The Class I sources or evolved protostars correspond to protostars surrounded by a thick disk. They present a strong IR excess, are visible in the optical domain and emit weakly in the X-ray domain.

Class II sources or classical T Tauri stars (or cTTs) are PMS surrounded by a thick disk that is believed to be slowly dissipating. They present a weak IR excess, are visible in the optical domain, are strong Hα and X-ray emitters.

Class III sources or weak-lined T Tauri stars (or wTTs) are PMS surrounded by a very thin or non- existent disk (and mostly do not accrete matter from it). They present no IR excess, are visible in the optical domain, present weak (if any) Hα emission and very strong X-ray emission.

2This refers to the level of IR excess compared to the emission of a stellar photosphere. It is believed that thermal emission from the dust present in circumstellar envelopes is responsible for the IR excess. 3The Hα emission line is produced when the matter that migrates from a circumstellar disk toward the surface of the star (within 3 or 4 stellar radii) is forced by the stellar magnetosphere to flow into accretion streams and columns.

4 1.1. PRE-MAIN SEQUENCE STARS CHAPTER 1. INTRODUCTION

Figure 1.3: Types of protostars and PMS (Feigelson & Montmerle 1999).

T Tauri stars probably emit X-rays thanks to a scaled-up version of solar-like magnetic activity. Indeed, they are thought to contain a dynamo, called the α – Ω dynamo, that is generated through the interplay of convection and differential rotation at the interface between the radiative core and the convective envelope. This dynamo in turn produces an external magnetic field. These stars then generate fairly strong4 and fre- quently variable X-ray emissions (Feigelson & Montmerle 1999; Rauw 2016-2017; Schulz 2012).

While both ends of magnetic loops can be connected to the stellar photospheres, T Tauri stars may also emit X-rays because of magnetic interactions with their circumstellar disks5 (Feigelson & Montmerle 1999; Rauw 2016-2017; Schulz 2012).

Younger T Tauri stars, on the other hand, are fully convective (and the α – Ω dynamo cannot work). This means that they probably contain a dynamo linked to convective motion (Rauw 2016-2017; Schulz 2012).

The PMS X-ray emissions frequently take the form of flares, i.e. sudden increases (in minutes or hours) of the X-ray luminosity6 (by up to a factor 100) followed by a decline (of several hours) ruled by the size of the loops (Gudel¨ & Naze´ 2009; Schulz 2012).

4The X-ray luminosity of PMS is typically 100 to 1 000 times the X-ray luminosity of the Sun. 5One could also consider the possibility of magnetic field lines with both feet in the disks. In this case, a dynamo would be generated thanks to convective motions and differential rotation. 6Flares emit not only in the X-ray domain, but all over the electromagnetic spectrum. However, their contrast is usually higher at higher photon energies.

5 1.2. THE OPEN CLUSTER NGC 6530 CHAPTER 1. INTRODUCTION

Flares are produced by the reconnections of unstable magnetic field lines. In the reconnection zone, particles are heated and accelerated. When they reach the stellar surface (after travelling along the ”new” field lines), they heat the plasma to temperatures of the order of 107 K. The induced pressure forces the hot plasma into loops and flares appear (Feigelson & Montmerle 1999;G udel¨ & Naze´ 2009; Schulz 2012).

The observation of PMS flaring activity can provide information about whether or not these PMS are surrounded by a circumstellar disk. Indeed, if the PMS is encircled by a disk of matter, the magnetic loops connect to the disk and do not display the same size as in the absence of a disk, impacting the flare decay time, which can be obtained thanks to the lightcurve of the event. In addition, spectral fits can notably be used to obtain information about the plasma temperature and about the interstellar column density (Schulz 2012).

1.2 The open cluster NGC 6530

In this work, we studied pre-main sequence objects contained in the very young open cluster NGC 6530. This cluster is located within the Lagoon Nebula (also known as M8), in the Sagittarius-Carina Arm. The nebula was discovered before 1654 by Giovanni Battista Hodierna, an Italian astronomer. It is classified as an HII region and is presently undergoing star formation. The OB stars of NGC 6530 play a key role in the ionization of M8 and it is believed that this ionization created a cavity, enabling us to see inside the nebula (Rauw et al. 2002; Tothill et al. 2008; ”Messier 8: Lagoon Nebula — Messier Objects” 2016).

Figure 1.4: Lagoon Nebula (ESO).

6 1.3. THE XMM-NEWTON SATELLITE CHAPTER 1. INTRODUCTION

NGC 6530, which was formed from the material of the Lagoon Nebula, is located about 1.3 kpc away. It is centered on 18:04:24, -24:21:12 (J2000.0), on the eastern part of the nebula and covers 14’ on the sky. NGC 6530 is only a few Myr old. It is an interesting open cluster because it contains massive stars and a population of lower mass stars (Tothill et al. 2008; ”Messier 8: Lagoon Nebula — Messier Objects” 2016).

Figure 1.5: Maps of the Lagoon Nebula (Tothill et al. 2008).

The Hourglass Nebula, consisting of gas and dust, is contained within the Lagoon Nebula as well. It is located in its bright central region. Since this region is currently experiencing a period of active star forma- tion, the Hourglass Nebula is illuminated by hot young stars and notably by 9 Sagittarii, an O-star binary. The central region of M8 is also illuminated by Herschel 36, a bright O-type star (”Messier 8: Lagoon Neb- ula — Messier Objects”, 2016).

Prior to the present study, NGC 6530 has been studied in the X-ray domain notably by Damiani et al. (2004) using Chandra7 data and by Rauw et al.(2002) using XMM-Newton data.

1.3 The XMM-Newton satellite

The X-ray images of NGC 6530 we analyzed were taken with XMM-Newton, a mission from the Euro- pean Space Agency dedicated to finding and studying X-ray sources. The X-ray Multi-Mirror Mission was launched in 1999 and is still in operation to this day (”XMM-Newton > XMM-Newton SOC Home Page”, 2016).

The XMM-Newton spacecraft carries three European Photon Imaging Cameras (or EPIC cameras) (see Sect. 2.2), two Reflection Grating Spectrometers (or RGS spectrometers) and one Optical Monitor (or OM). The EPIC cameras and the RGS spectrometers are located on the Focal Plane Platform (or FPP), the OM and the mirrors8 are on the Mirror Support Platform (or MSP). The 6.80 m Telescope Tube maintains the

7The Chandra X-ray Observatory is a NASA space telescope dedicated to X-ray astrophysics (”Chandra :: About Chandra”, 2017). 8XMM-Newton carries three mirror assemblies. Each assembly is made of 58 gold-coated Wolter I mirrors that are embedded inside each other.

7 1.4. SUMMARY CHAPTER 1. INTRODUCTION

Figure 1.6: Artist’s view of the XMM-Newton spacecraft (Image courtesy of C. Carreau and ESA) (”XMM-Newton > XMM-Newton Image Gallery”, 2017). Figure 1.7: XMM-Newton’s components (”Tech- nical Details - Spacecraft - Cosmos”, 2017).

FPP and the MSP at their relative position (see Fig. 1.7)(”Technical Details - Spacecraft - Cosmos” 2017, ”XMM-Newton Users Handbook” 2016).

XMM-Newton’s instruments allow the satellite to perform, among other things, X-ray imaging, opti- cal and UV imaging, X-ray photometry and relatively high resolution X-ray spectroscopy (”XMM-Newton Users Handbook”, 2016).

1.4 Summary

We first processed the XMM-Newton data consisting of four observations of NGC 6530, taken in March 2001, March 2013, September 2013 and March 2014 (see Chapter2). In fact, we ”recycled” data that were taken in order to study the massive star population of the cluster, and in particular the long-period O-star binary 9 Sagittarii9.

We then detected the X-ray sources of the cluster (see Chapter3). In order to identify the optical coun- terpart of the X-ray stars, to obtain their known properties and notably their rotation rates (see Chapter5), we correlated their positions with catalogs of infrared, optical and X-ray observations (see Chapter4). We also positioned our sources in a Hertzsprung-Russell diagram (see Chapter6).

9The 9 years period of this binary explains the temporal gap between the observations.

8 1.4. SUMMARY CHAPTER 1. INTRODUCTION

We determined which sources were variable (see Chapter7). Next, by inspecting the lightcurves and spectra that we extracted for the brightest variable sources (see Chapters8&9), we searched for the ones presenting flares. Then we studied the characteristics of these flares in order to constrain the size of the associated magnetic loop and to look for possible interactions with a circumstellar disk. We also used the spectral fits to deduce some properties. Finally, we compared our results to the literature and considered their implications (see Chapter 10).

9 Chapter 2

Processing

2.1 Lists of events

The raw data from XMM-Newton consist of a lists of events. Each event corresponds to the recording of the impact of a photon on a detector. The detectors on board XMM-Newton are Charge Coupled Devices (or CCDs), matrices of pixels. When a photon arrives on the semi-conductor that forms the CCD, it generates a given number of free electrons (and the same number of holes). This number is proportional to the energy of the incoming photon. Consequently, by measuring the number of free electrons produced, one can deduce the energy of the photon that arrived on the CCD (Rauw 2016-2017).

The lists of events contains: • the position of the pixel affected by the arrival of the photon (i.e. x, y coordinates on the detector); • a measure of the raw number of free electrons generated; • the time of the impact of the photon; • a grade, which describes the distribution of free electrons in the 8 adjacent pixels around the pixel where the arrival of the photon was recorded.

Figure 2.1: Schematic of the grade 0 (Rauw 2016-2017).

When a photon arrives on the matrices of pixels, a cloud of charged particles is formed. The shape of the cloud, its pattern (see Sect. 2.3), indicates the nature of the event that produced the free electrons. A critical threshold can be defined, corresponding to the minimum number of free electrons generated upon the arrival of a photon on a pixel. Since we know the number of free electrons that should be generated in the eight

10 2.2. EPIC CAMERAS CHAPTER 2. PROCESSING neighboring pixels, by looking which pixels are above the threshold, one can differentiate between a photon and a cosmic ray. Consequently, we define the grade as an indicator of the probability that the recorded event is due to a photon (Rauw 2016-2017).

In most X-ray missions, the events are screened a first time during on-board processing and only those having a specific grade (that are therefore likely corresponding to the impact of an X-ray photon on the CCDs) are transmitted to the ground (Rauw 2016-2017).

2.2 EPIC cameras

Our observations were obtained with the EPIC cameras. These three X-ray CCD cameras are the Metal Oxide Semi-conductor cameras (called MOS1 and MOS2, they use MOS CCDs) and the pn camera (that uses pn-CCDs) (”XMM-Newton > Technical Details - EPIC”, 2016).

Figure 2.2: CCDs of the MOS cameras (”XMM- Figure 2.3: CCDs of the pn camera (”XMM- Newton > Technical Details - EPIC”, 2016). Newton > Technical Details - EPIC”, 2016).

The MOS1 and MOS2 cameras have 7 identical CCDs (MOS2 is turned 90◦ compared to MOS1). The pn camera has 12 identical rectangular CCDs. (During an observation, the central source is positioned on the central CCD of the MOSes and in the corner close to the center of the mosaic of the pn.) (”XMM-Newton Users Handbook”, 2016)

The EPIC cameras use optical blocking filters. This is due to the fact that their CCDs are sensitive to UV, visible and near-IR photons (as well as X-ray photons) (”XMM-Newton Users Handbook”, 2016; Rauw 2016-2017). The filters are selected according to the visible magnitude and spectral type of the target to be observed. In our case, a thick filter was used in order to eliminate the UV and visible light of 9 Sgr (that has an apparent magnitude of 5), to avoid any contamination of the X-ray signal.

It needs to be noted that there are slight differences between the first observation (made in March 2001) and the three other ones (made between March 2013 and March 2014). The right ascension and the dec- lination of the center of the field of view, as well as the durations of the exposures are different. Another

11 2.2. EPIC CAMERAS CHAPTER 2. PROCESSING

Figure 2.4: Sketch of the field of view of the MOS cameras and the pn camera. The CCDs of MOS2 are rotated by 90◦ on the sky compared to those of MOS1. (”XMM-Newton Users Handbook”, 2016)

Table 2.1: Our four observations

Observation Julian Date MOS mode pn mode pn integration time (middle of the observation) (s) #1 2451977.108 Full Frame Extended Full Frame 19495 #2 2456360.297 Large Window Full Frame 20039 #3 2456539.484 Large Window Full Frame 23739 #4 2456721.922 Large Window Full Frame 24539

difference concerns the fact that two peripheral CCDs of MOS1 were lost in 2005 and 2012 following the impact of a micrometeoroid on the instrument. These CCDs are thus missing in the 2013 and 2014 data. Additionally, in the case of the 2001 observation, the instruments MOS1 and MOS2 are in the Full Frame mode, whereas in the other cases, their central CCDs are in the Large Window mode. The latter mode was chosen to avoid pile-up for the main target (9 Sgr).

In the (Extended) Full Frame mode1, all the CCDs are read. In the Large Window mode, only the central part of the central CCD and the eight neighboring CCDs are read, but with different read-out times (”XMM- Newton Users Handbook”, 2016).

The temporal resolution, or reading time, of the pn-CCDs is 199.1 ms in the Extended Full Frame mode (the mode used for the first observation) and is 73.4 ms in the Full Frame mode (the mode used for the

1The integration time is longer for the extended mode.

12 2.3. SAS PROCESSING CHAPTER 2. PROCESSING

Figure 2.5: Left: Full Frame mode; right: Large Window mode (”XMM-Newton Users Handbook”, 2016).

last observations). For the MOS-CCDs in Full Frame mode, it is of 2.6 s. For the MOS-CCDs in Large Window mode, it is of 0.9 s for the central part of the central CCD and of 2.6 s for its neighboring CCDs (”XMM-Newton Users Handbook”, 2016).

The Large Window mode is used to decrease the integration time and consequently the risk of photon pile-up. Indeed, in order to avoid the superposition of two photons on the same pixels, the integration time must be short. Otherwise, there is an error on the estimation of the energy of the photon (since we suppose that only one photon has arrived, when in reality two - or more - came). The phenomenon of arrival of more than one photon before the read-out time is called pile-up (”XMM-Newton Users Handbook”, 2016; Rauw 2016-2017). Fortunately, this problem is not common for XMM, especially not for weaker sources. Here, it was used because 9 Sgr is a binary system that shows a level of X-ray emission modulated by the separation between the stars. The first observation was done close to apoastron, but the others were done around the periastron, when the X-ray flux was expected to be largest. (The Large Window mode was therefore used due to a fear of an increase in brightness).

2.3 SAS processing

The data recorded on board XMM-Newton are transmitted to the ground and are subsequently processed with the pipeline reduction chain. The results are found in two subdirectories: the Observation Data Files (ODF) and Pipeline Processing files (or Pipeline Processing (Sub)system, PPS) (Snowden et al. 2016). The files related to the four observations that we analyzed are raw data taken from the ODF.

We used a software called the Science Analysis Software (or Science Analysis System, SAS)2 to process our data (”XMM-Newton Users Handbook”, 2016).

2The SAS was specifically written for XMM-Newton. It can be used to screen the data, to build images, lightcurves and spectra.

13 2.3. SAS PROCESSING CHAPTER 2. PROCESSING

First, we used the command cifbuild (CIF referring to the Calibration Index File) to create a list of the calibration files required for each observation (according to the date and the instrumental configuration) by searching in the Current Calibration Files (or CCF) that are stored on the same computer as our data. The ”current” part of the name refers to the fact that the calibration evolves with time and depends on the aging of the detector. The CCF contains information about the date until which a particular calibration is valid (”XMM-Newton Users Handbook”, 2016; ”Users Guide to the XMM-Newton Science Analysis System”, 2016).

There is a complication that arises from the lecture mode of the CCD, which is a sequential lecture mode, i.e. in order to be read at the end of a line (where the reading register is located), the charges generated in one pixel have to pass through the neighboring pixels. This can lead to losses, that are quantified by the Charge Transfer Inefficiency (or CTI) or by the Charge Transfer Efficiency (or CTE)3. Consequently, to know the true number of photons that arrived on the detector, one needs to apply a correction (that depends on the date of the observations, because of the radiation damage, and on the location of the pixel that encounter the photon, since it dictates the number of transfers across pixels) (Rauw 2016-2017). This correction can be found inside the CCF.

Next, we used the command odfingest to create an inventory of the information contained in the ODF (including information about which instrument was active, the exposure time, the number of exposures, which CCDs were used and in what mode). This inventory is called the SAS summary file (”XMM-Newton Users Handbook”, 2016; ”Users Guide to the XMM-Newton Science Analysis System”, 2016).

Then, we used the SAS reduction tasks emproc (for the files related to the MOS instruments) and epproc (for the ones related to the pn instrument). Since the list of events is not organized inside the ODF, we ran these pipeline processing commands in order to rearrange the events, making them suitable for manipulation (”XMM-Newton Users Handbook”, 2016; ”Users Guide to the XMM-Newton Science Analysis System”, 2016).

Lastly, with the command evselect, we filtered the events according to their grade (to dissociate events due to photons from those due to cosmic rays) (see Sect. 2.1) and criteria of proximity to defective columns of the CCDs (to eliminate those columns) (”Users Guide to the XMM-Newton Science Analysis System”, 2016).

The grade is associated to a pattern (see Fig. 2.6). The pattern expresses the dispersion of the free elec- trons generated upon the impact of a photon on the pixels of a CCD (and it is consequently linked to the number of pixels above the detection threshold and to their position) (”Users Guide to the XMM-Newton Science Analysis System”, 2016).

3In practice, the parameters on board the satellite are determined by a radioactive source (with a known intensity) that illuminates the pixels. By measuring the number of recorded electrons and by knowing the real number of electrons produced, we can deduce the loss.

14 2.3. SAS PROCESSING CHAPTER 2. PROCESSING

Figure 2.6: List of MOS patterns and the associated grade (”Patterns”, 2017).

15 2.3. SAS PROCESSING CHAPTER 2. PROCESSING

In Fig. 2.6, the red pixel is the central one (and the central pixel is the one with the highest charge), the green pixels are the ones above the threshold, the white pixels are below the threshold and the crosses repre- sent pixels whose charge is indifferent. The events that are most likely due to an X-ray photon are the ones closest to the grade 0 (i.e. the ones with a compact patten, the highest charge in the center and surrounding pixels below the threshold) (”Patterns”, 2017).

For the MOS instruments, 32 patterns have been defined. We selected events with grade in the 0 to 12 range (because these grades encompass most of the events due to X-ray light). The grade 0 corresponds to a single pixel event, grades 1 to 4 correspond to double pixel events and grades 5 to 12 correspond to triple and quadruple pixel events (”Users Guide to the XMM-Newton Science Analysis System”, 2016; Snowden et al. 2016).

For the pn instrument, 13 patterns have been defined and we selected events with grade in the 0 to 4 range. We also set the flag (representing the quality of the event) to 0 (which corresponds to the most severe screening criteria) (”Users Guide to the XMM-Newton Science Analysis System”, 2016; Snowden et al. 2016).

16 Chapter 3

Source detection

3.1 Detection on combined images

Before performing the source detection, we used evselect to extract lightcurves for each instrument over its entire field of view and each observation for events of energy higher than 10 keV and of grade 0 in order to see if the curves presented strong flare-like variations that would be linked to low energy protons. Due to the Wolter I - type structure of XMM-Newton’s mirrors (see Fig. 3.1), such protons can follow the same path as photons. Since at these energies protons produce a strong signal, while astrophysical sources are dim, if we had noticed a flare, it would have been due to low energy protons. Discarding the time intervals corresponding to these flares from our data is important to reduce the background and obtain clean images and spectra. We observed no flare for the MOS and pn cameras, allowing us to keep the entire duration of the exposure.

Figure 3.1: Schematic of XMM-Newton’s mirrors (”Technical Details - Mirrors - Cosmos”, 2017).

17 3.1. DETECTION ON COMBINED IMAGES CHAPTER 3. SOURCE DETECTION

To detect sources1, we first extracted images for the three instruments separately, selecting a specific energy range because of the straylight that affects our observations.

The straylight results from the geometry of XMM-Newton’s mirrors. For a source in the field of view, its light is first reflected under grazing incidence on a parabolic mirror, then on a hyperbolic mirror, and is focused on the detector. When a bright source is located outside of the field of view (within about 1 - 2◦ from the center of the field), its photons can reach the detector unfocused (they undergo a single reflection, on the hyperbolic mirror). Consequently, the image of this off-axis source does not appear as a point-like source, but as a fingerprint-like shape (we are in fact seeing the shell structure of the mirrors) that appears extended over an area away from the actual location of the source in the sky (Rauw 2016-2017). In the present case, the straylight that affects our data is due tot the photons from the low-mass X-ray binary Sgr X-3 located at 1◦ from the center of the field of view.

Figure 3.2: Image, in the 0.5 - 8 keV band, of Figure 3.3: Image, in the 0.4 - 1.4 keV band, of NGC 6530 with the straylight. NGC 6530 without the straylight.

The straylight gets mixed with the true sources that we want to study. This is a problem for their detec- tion: the software will detect thousands of fake sources on the arcs of the straylight. In order to make the straylight disappear (or at least to attenuate it), we chose to perform the source detection on a soft X-ray image (with the energy ranging between 0.4 and 1.4 keV), because in our case the source that is responsible for the straylight is an X-binary that is very heavily absorbed and presents therefore a very hard spectrum with essentially no photons below 1.4 keV.

There is another source of false detection: the out of time (or OoT) events. They originate, in our case, from the fact that 9 Sgr, situated in the center of the field of view, produces photons constantly, including

1We tried different ways of detecting sources. The result of those attempts was to use the technique described here.

18 3.1. DETECTION ON COMBINED IMAGES CHAPTER 3. SOURCE DETECTION also during the read-out time when the electrons travel from one pixel to the other towards the read-out node. Since there are no shutters in front of the instruments, the light that arrives during that time impacts pixels that are transferring charges generated by photons emanating from other places in the sky. The electrons created by out of time light are therefore attributed erroneously to locations in the sky that correspond to pixels positioned between the bright source and the read-out node. Consequently, a trail appears along the reading direction (Rauw 2016-2017).

Figure 3.4: Image with out of time photons (in red).

To get rid of the OoT events present on the pn images, we ran the epchain command. Based on a prelim- inary list of sources detected in the field, this command produced a simulated OoT events list, so that images and spectra built from this list can be subtracted from images and spectra extracted from the actual event list (”XMM-Newton Users Handbook”, 2016).

We wanted to combine, for each instrument, the four observations into one image. In order to unify the four data sets into a single event list, we used the merge command. We merged the first observation with the second, the third with the fourth and finally we merged the two intermediate products together. We also merged the OoT events lists.

Then, for each instrument, we added the exposure maps2 of the four observation with the farith com- mand. Exposure maps are images of the sky showing the effective exposure time as a function of the location. The sensitivity of the instrument is not uniform over the entire field of view (because of the vignetting of the

2The exposure maps were used to create detection masks (with the emask command). Then the exposure maps were used along with the detection masks and background maps (produced by the esplinemap command) by the source detection commands.

19 3.1. DETECTION ON COMBINED IMAGES CHAPTER 3. SOURCE DETECTION mirrors, gaps between the detector chips, dead pixels, ...). The exposure map provides, for each position in the field, an effective exposure time that accounts for these effects. This allows to assign the correct count rate to a source regardless of its position inside the field of view (Rauw 2016-2017).

Next, using the evselect command, we extracted an image (with the energy ranging between 0.4 and 1.4 keV) for each merged event list. At this point, we had three combined images, one for the MOS1 cameras, one for the MOS2 and one for the pn. We also built an image of the merged OoT events list, multiplied it by 0.0633 and subtracted it from the pn images.

We were then able to move on to the actual source detection. We worked with the eboxl4 command that scanned the images and took note of potential sources. Next, the eboxm5 command searched for potential diffuse emission and how it might vary inside the field of view (taking into account the information obtained during the first local scan). Eventually, the emldetect command fitted PSFs on the data collected by eboxl and eboxm (”Specifications for individual SSC data products”, 2015).

We detected 550 stars with a significance threshold of at least 20 (this means that we imposed a proba- bility of less than e−20 that random Poissonian fluctuations could be mistaken for sources).

In addition, a mosaic of the three images (MOS1, MOS2 and pn) was created with the emosaic com- mand. This mosaic notably allows to reduce the impact of the gaps between the CCDs and of the dead pixels on detectors (for the MOS1). In the end, we obtained a single image for all four observations and all three instruments.

Figure 3.5: Mosaic.

3For the Full Frame mode, 6.3 % of events are OoT. 4The name refers to the fact that the command acts like a box that ”slides” over the image in search of sources, estimating the background in the box region. It is a local mode of examination. 5The name indicates a ”map” mode. The command hides the sources detected by eboxl to determine a global background.

20 3.2. DETECTION VERIFICATION CHAPTER 3. SOURCE DETECTION

3.2 Detection verification

We took note of the position of each source detected by the SAS on the mosaic, then, using SAOImage DS9, a visualization tool for flexible image transfer system (or fits) files, we checked each one by visually inspecting the images (”SAOImage DS9”, 2017). We decided to ”quarantine” the ones that we believed were proba- bly not real sources. We also added 6 stars (that had not been detected by the SAS but seemed real) to our list.

Figure 3.6: Detected sources displayed on the mosaic. (The different colors correspond to our subjective perception of the reality of the sources, the red color being used for ”questionable” detections.)

We found 352 probable sources and 204 ”questionable” detections.

3.3 Detection on individual observations

The first source detection we carried out was on the four observations combined into one image. Next, we decided to identify stars on individual observations.

For each observation, we first used evselect to extract images in the energy band that we previously worked with on the combined images (i.e. between 0.4 and 1.4 keV). We extracted images for the MOS1, MOS2 and pn instruments.

On the twelve images (three images for each of the four observations) that we obtained, we chose to detect stars by imposing the positions we determined earlier on the combined images (see Sect. 3.2). In order to do so, we created detection masks, with the emask command, using the exposure maps. Then we used esplinemap to construct background maps (where the positions of the sources detected on the combined

21 3.4. THREE COLORS IMAGE CHAPTER 3. SOURCE DETECTION images were taken into account). Finally, we ran the emldetect command to produce the new source lists (”Users Guide to the XMM-Newton Science Analysis System”, 2016).

If we compare the list of sources detected on the combined images with the lists of sources from the individual observations, we find that we lost stars on the way. This is due to the fact that the orientation of the CCDs is not exactly the same from one observation to another and sources identified on the combined images can be out of the field of view of individual observations. Furthermore, the combined image has a four-times larger exposure time than the individual images, thus allowing to achieve deeper detections.

3.4 Three colors image

For each observation and each instrument, we instructed the evselect task to extract images in three consec- utive energy bands: a soft band (ranging from 0.4 to 0.8 keV), a medium band (ranging from 0.8 to 1.2 keV) and a hard band (ranging from 1.2 to 1.4 keV)6. Additionally, we built the associated exposure maps. We built images of the OoT events as well, multiplied them by 0.063 and subtracted them from the pn images (for each observation, in every energy band).

Subsequently, we added the soft-band exposure maps for the MOS1, MOS2 and pn cameras. We did the same for the medium- and hard-band exposure maps, yielding, for one observation, three exposure maps, one for each energy range. Thereafter, we added the soft-band images for the MOS1, MOS2 and pn instru- ments, and similarly for the medium- and hard-band images.

Next, we added the four images (one for every observation), in the three energy ranges. We also added the four exposure maps (one for each observation), in the three energy ranges. Then, we normalized the three exposure maps we obtained by dividing them by the corresponding total integration time at the center of the field of view7. Afterwards, we divided each image by the matching normalized exposure map8. In the end, we produced three images: one in the soft band, one in the medium band and one in the hard band. Lastly, we combined the three images into a three colors image (see Fig. 3.7) using the dmimg2jpg command of the CIAO (Chandra Interactive Analysis of Observations) software package (”Ahelp: dmimg2jpg - CIAO 4.9”, 2017).

6At first, we chose the ranges 0.5-1, 1-2 and 2-8 keV, but due to the presence of the straylight even in the medium band, we had to adjust the energy bands. 7This was done in order not to have images with very small values for individual pixels. Indeed, we normalized the exposure maps in order to later divide the images by a number close to 1 and not by a large number. 8Even though we divided by the exposure map at first, since it increased the noise in the external regions in a way that was difficult to control, we ultimately chose not to do this division.

22 3.4. THREE COLORS IMAGE CHAPTER 3. SOURCE DETECTION

Figure 3.7: Three colors image (without the exposure map correction).

23 Chapter 4

Correlations

4.1 Catalogs

We correlated the positions of the X-ray sources we found with various catalogs in order to find their known properties. We looked inside the 2MASS (from Cutri et al.(2003)) (presenting IR sources), Damiani et al. (2004) (presenting X-ray sources), Henderson & Stassun(2012), Kalari et al.(2015), Kumar & Anandarao (2010) (presenting IR sources), Prisinzano et al.(2005), Prisinzano et al.(2007), Prisinzano et al.(2012) and Sung et al.(2000) (presenting visible sources) catalogs. For this purpose, we used the ”VizieR” interface to a library of astronomical catalogs to compare our list of sources with the different catalogs (”VizieR Service”, 2017).

The 2MASS, for ”Two Micron All Sky Survey”, is a project that scanned the complete sky in three near- infrared bands in order to find and characterize sources (Cutri et al. 2003).

In their article, Damiani et al.(2004) analyzed observations (of 60 ks) of NGC 6530 made with Chandra. They detected 884 sources and identified more than 90% of them as low-mass PMS.

Henderson & Stassun(2012) studied I-band photometric time series of about 50 000 sources in the HII region of the Lagoon Nebula. They identified the rotation periods (between 0.5 and 10 days - with sources surrounded by disks rotating more slowly in general) of 290 low-mass stars present in NGC 6530.

The accretion rates of 235 cTTs candidates were observationally estimated by Kalari et al.(2015) in M8 using Hα and ugri photometry. They were found to increase with the square of the stellar mass.

Kumar & Anandarao(2010) studied the Lagoon Nebula using Spitzer IRAC 1 photometry. They detected 64 class 0 and I objects, and 168 cTTs.

NGC 6530 was also studied by Prisinzano et al. In their 2005 article, they produced astrometry and BVI photometry of the open cluster. In 2007, they analyzed 332 candidate members in order to examine their

1The Spitzer Space Telescope is a NASA space telescope dedicated to IR detection. The Infrared Array Camera (or IRAC) is one of its instrument (”Mission Overview - NASA Spitzer Space Telescope”, 2017& ”The Infrared Array Camera (IRAC) - NASA Spitzer Space Telescope”, 2017).

24 4.2. CORRELATION RADIUS CHAPTER 4. CORRELATIONS properties and found that, on average, sources with disks rotate slower, in agreement with the conclusions of Henderson & Stassun(2012). In 2012, they studied 88 confirmed members in order to obtain their stellar parameters.

UBVRI and Hα photometry of 887 stars in the region of NGC 6530 as well as a list of 37 PMS and 9 PMS candidates were supplied by Sung et al.(2000). They additionally found the mean color excess E(B-V) of the stars in the cluster to be equal to 0.35 mag.

4.2 Correlation radius

To determine the optimal correlation radius, we looked at the number of correlations (for the probable sources) for a correlation radius going from 1” to 10” with a 1” step. For the 2MASS, Damiani et al.(2004) and Sung et al.(2000) catalogs, we obtained:

Correlation radius Number of correlations Number of correlations Number of correlations (in ”) 2MASS Damiani Sung 1 81 65 39 2 199 148 91 3 272 216 130 4 314 242 152 5 332 255 163 6 343 264 173 7 348 267 178 8 349 269 183 9 350 269 187 10 352 269 193

We plotted the number of correlations as a function of the correlation radius (see Fig. 4.1, Fig. 4.2& Fig. 4.3). Then we fitted the curve using two gaussian expressions representing the probability of finding a source either because there is a real correlation or by chance (due to the density of sources in the field):

−r2 Φ(r)= A[1 - exp( )] + (N − A)[1 - exp(−πr2B)] 2σ2 (4.1) 2 2 = A[1 - exp(−k1r )] + (N − A)[1 - exp (−k2r )] where A corresponds to the number of true correlations, σ is the uncertainty on the positions of our sources, N is equal to the number of sources and B is the number of sources in the catalog per unit area (Jeffries et al. 1997).

We used the Levenberg-Marquardt method to determine the optimal parameters (Press et al. 1997):

2 Catalog A k1 σ k2 B χ 2MASS 199.2 0.4062 1.1095 0.08379 0.02667 211.88 Damiani 249.1 0.2293 1.4767 0.00278 0.00088 323.48 Sung 151.7 0.2193 1.5100 0.00249 0.00079 150.52

25 4.2. CORRELATION RADIUS CHAPTER 4. CORRELATIONS

Figure 4.1: The number of correlations as a function of the correlation radius for the 2MASS catalog. (The black histogram is our curve, the blue line is its fit, the green line corresponds to the real correlations and the red line refers to the spurious correlations.)

26 4.2. CORRELATION RADIUS CHAPTER 4. CORRELATIONS

Figure 4.2: Number of correlations as a function Figure 4.3: Number of correlations as a function of the correlation radius for the Damiani et al. of the correlation radius for the Sung et al.(2000) (2004) catalog. catalog.

Table 4.1: Correlation radii for the different catalogs. The percentage of spurious correlations is obtained by dividing the number of spurious correlations (obtained from our fit) by the total number of correlations (i.e. the number of spurious and true correlations obtained from our fit). The percentage of true correlations is obtained by dividing the number of true correlations (obtained from our fit) by A (the asymptotic value of the number of true correlations).

Catalog Correlation Number of Number of Percentage of Percentage of radius (in ”) true correlations spurious correlations spurious correlations true correlations 2MASS 1 66.5 12.9 16.2 33.3 2 160.0 45.3 22.1 80.3 2.5 183.5 64.6 26.0 92.1 3 194.0 83.6 30.1 97.4 4 198.9 115.4 36.7 99.8 Damiani 1 51.0 0.3 0.6 20.5 2 149.4 1.1 0.8 60.0 3 217.4 2.6 1.2 87.3 4 242.7 4.5 1.8 97.4 5 248.3 7.0 2.7 99.7 Sung 1 29.8 0.5 1.6 19.6 2 88.5 2.0 2.2 58.4 3 130.5 4.5 3.3 86.0 4 147.1 7.9 5.1 97.0 5 151.0 12.1 7.4 99.6

27 4.2. CORRELATION RADIUS CHAPTER 4. CORRELATIONS

In order to get a high percentage of true correlations, while limiting the number of spurious correla- tions and the number of double matches2, we chose a correlation radius of 2.5” for the 2MASS and of 4” for the other catalogs. These correlation radii are similar to the size of the core of the XMM-Newton PSF (∼ 5”).

We used these correlation radii to correlate our probable and ”questionable” sources with the nine cat- alogs we selected. In this way, we could build the Hertzsprung-Russell diagram of our X-ray sources (see Chapter6) and more generally study the impact of rotation on the level of X-ray emission (see Chapter5).

2This means that we wanted to be as close as possible to the bend of the green lines in Fig. 4.1, Fig. 4.2 & Fig. 4.3. To quantify this, we evaluated two percentages representing different notions (see Table 4.1).

28 Chapter 5

Rotation rates

Using the cross-correlation of our probable X-ray sources with the catalog of Henderson & Stassun(2012), we established the histogram of the rotation periods of the 92 objects in common. We then compared the result with the histogram of the rotation periods of the full sample of 290 objects in the catalogue of Hen- derson & Stassun(2012). The result of this comparison is illustrated in Fig. 5.1.

Figure 5.1: Frequency of rotation periods among our sample and the full sample of NGC 6530 stars of Henderson & Stassun(2012). Left: the blue histogram corresponds to the X-ray selected stars from our sample which have a counterpart in the Henderson & Stassun(2012) catalogue (92 objects), whereas the red histogram yields the results of the complete sample of Henderson & Stassun(2012) (290 objects). Right: the blue histogram yields the 35 X-ray sources with a Henderson & Stassun(2012) counterpart and displaying significant inter-pointing variability.

From this figure, we see that, on average, the rotation rates of the X-ray selected stars do not appear to be larger than those of the full sample studied by Henderson & Stassun(2012). We repeated the same exer- cise restricting ourselves to the X-ray sources that display significant inter-pointing variability (see Sect. 7.3)

29 CHAPTER 5. ROTATION RATES which yields 35 correlations with the Henderson & Stassun(2012) catalogue. Given the low number of ob- jects, again, there is no significant difference between the two distributions. In summary, the X-ray emitting stars in NGC 6530 do not appear to be rotating faster than other stars in the sample of Henderson & Stassun (2012).

Figure 5.2: Left: EPIC-pn count rate (average value over all four observations) as a function of rotational period for the stars in common with the catalogue of Henderson & Stassun(2012). Right: same, but this time grouping the stars into bins according to their rotational velocities. The error bars on the bins correspond to the standard deviations of the count rates for stars in the bin.

The left panel of Fig. 5.2 illustrates the mean EPIC-pn count rate as a function of the rotation period. We clearly see that there exists a huge dispersion among the data points, although a trend appears for the highest count rates, they seem to increase with decreasing rotational period. However, as illustrated on the right panel of that figure, if we average the EPIC-pn count rates per bin of the rotation period, we find no obvious dependency on rotation period.

Apart from the fact that the highest fluxes for a given rotation period increase with decreasing rotation period, the stars in our sample thus do not display any clear signature of the α – Ω dynamo.

30 Chapter 6

Hertzsprung-Russell diagram

6.1 Color-magnitude diagram

Before constructing a Hertzsprung-Russell diagram, we first built a color-magnitude diagram using the re- sults obtained from the cross-correlation with the catalog of Sung et al.(2000) (see Fig. 6.1).

Figure 6.1: Color-magnitude diagram for the 163 probable and 17 ”questionable” sources having a counter- part in Sung’s catalog.

31 6.2. HERTZSPRUNG-RUSSELL DIAGRAM CHAPTER 6. HERTZSPRUNG-RUSSELL DIAGRAM

We see no obvious difference between the locus of the known PMS objects (classical and weak-line T Tauri stars) and stars not showing any Hα emission in the Sung et al.(2000) photometry. Hence the X-ray selected stars in NGC 6530 consist of a mix of non-Hα emitting stars (the majority of the objects) and Hα selected PMS stars.

6.2 Hertzsprung-Russell diagram

To assemble the Hertzsprung-Russell diagram, we had to convert the intrinsic V-I index into an effective temperature and convert the magnitude into a bolometric magnitude. We used data from the correlations with the Henderson & Stassun(2012), Prisinzano et al.(2005), Prisinzano et al.(2012) and Sung et al. (2000) catalogs. We further used the A5 table from Kenyon & Hartmann(1995). This table considers main- sequence stars of spectral type later than B0 and provides, among others, their effective temperatures, the bolometric corrections for the V filter and intrinsic colors.

For each star, we calculated the intrinsic (V − I)0 index. The observed (V − I)obs index is equal to the sum of the intrinsic index and the color excess E(V − I):

(V − I)obs = (V − I)0 + E(V − I) = (V − I)0 + 1.6 × E(B − V ) = (V − I)0 + 0.56 (6.1) where the relation E(V − I) = 1.6 × E(B − V ) stems from the reddening law that we have adopted1 and where we assumed a color excess E(B − V ) = 0.35 (Sung et al. 2000). This leads to

(V − I)0 = (V − I)obs − 0.56 (6.2) Assuming a constant reddening for all our stars in the field of view is only a first order approximation, as differential reddening is certainly present in the Lagoon Nebula. However, this should be sufficient to derive the average properties of the population of X-ray emitting stars in NGC 6530.

For the probable sources which had a counterpart in the catalog of Sung et al.(2000), we calculated the effective temperature, the bolometric magnitude and also checked for the presence or absence of Hα emission. In order to do so, we first interpolated in the A5 table, to establish the effective temperatures and bolometric corrections corresponding to the star’s intrinsic colors.

Then, for each source, we combined this information with some data taken from the catalog of Sung et al.(2000), notably including the ID of the source, its flag 2, its V magnitude, as well as the V − I and B − V indices3.

The absolute magnitudes in the V band were obtained from

MV = V − RV × E(B − V ) − DM = V − 3.1 × 0.35 − 10.48 = V − 11.565 (6.3)

1 AV We followed Sung et al.(2000) and adopted R = E(B−V ) = 3.1 for the reddening law. 2Each source that appears in one of the tables of Sung et al.(2000) carries a flag with the number of this table. Table 3 contains PMS sources presenting strong Hα emission, table 4 contains PMS sources presenting weak Hα emission and table 5 contains PMS sources identified with the Hubble Space Telescope. 3Our list of X-ray sources contains several massive stars that are not covered by the data of the Kenyon & Hartmann(1995) table. Their color indices fall outside the range of values in the A5 table and in our automatic procedure, we set their effective temperature and the bolometric magnitude to 0 to identify them as being irrelevant for the present study of low-mass PMS stars.

32 6.2. HERTZSPRUNG-RUSSELL DIAGRAM CHAPTER 6. HERTZSPRUNG-RUSSELL DIAGRAM

where we have again adopted RV = 3.1 and E(B − V ) = 0.35. The distance modulus DM = 10.48 (corresponding to a distance of 1250 pc) was adopted from Prisinzano et al.(2005). Using the bolometric correction obtained via interpolation in the A5 table, we then calculated the bolometric magnitude.

In addition, we identified sources flagged with the number ”3” (i.e. strong Hα emitters) as classical T Tauri stars and sources flagged with the number ”4” (i.e. the weak Hα emitters) as weak-line T Tauri stars.

The same procedures were applied to the lists of correlations of the probable sources with the other cat- alogs, with a few exceptions (because they do not all contain the same information). In the Prisinzano et al. (2005) catalog, there is no information regarding Hα emission.

Finally, we combined the information from the various catalogs and used a supermongo4 routine to plot the bolometric magnitude as a fonction of the logarithm of the effective temperatures for those 292 X-ray sources that have an optical counterpart consistent with a late-type star (”SM”, 2017). To further inter- pret the Hertzsprung-Russell diagram in terms of PMS evolution, we used the PMS evolutionary tracks and isochrones provided by Siess et al.(2000). The results are displayed in Fig. 6.2.

The same procedures were repeated to build a Hertzsprung-Russell diagram for the ”questionable” sources. Only 48 of them were found to have an optical counterpart in the catalogs considered here (see Fig. 6.3).

We can see that the majority of X-ray sources in the Hertzsprung-Russell diagrams are located in be- tween the isochrones corresponding to ages of 1.5 and 10 Myr according to the Siess et al.(2000) isochrones. By comparison, Rauw et al.(2002) found ages ranging from 4 to 20 Myr, Damiani et al.(2004) obtained an age range between 0.5 and 1.5 Myr and Prisinzano et al.(2005) found ages between 1-2 Myr and 6-7 Myr for the cluster members.

Classical T Tauri stars seem to be the exception rather than the rule among the X-ray selected objects in NGC 6530: for the vast majority of the X-ray sources, there is no observational indication of Hα emission. This result is in agreement with the studies of Rauw et al.(2002) and Damiani et al.(2004). By comparison, for NGC 6231, Sana et al.(2007) found 302 wTTs and 70 cTTs. Rauw et al.(2003) found that NGC 6383 contains probably only wTTs. Stelzer & Neuhauser¨ (2001) obtained similar results for the Taurus-Auriga- Perseus complex.

The majority of the X-ray sources fall in a mass range between 0.5 and 1.0 M . This result could be somewhat biased. At the lower-mass end, our X-ray observations could lack the sensitivity to detect X-ray emission from even lower mass stars. By comparison, Rauw et al.(2002) found masses ranging from 1.0 to 2.0 M , Damiani et al.(2004) found masses between 0.5 and 1.5 M and Prisinzano et al.(2005) found masses ranging from 0.6 to 4.0 M for NGC 6530. Our results are thus in agreement with previous stud- ies, though we stress that the mass ranges and ages inferred from the Hertzsprung-Russell diagram strongly depend upon the adopted distance modulus, reddening, and the PMS evolutionary tracks used to build the isochrones. For NGC 6231, Sana et al.(2007) found a mass range between 0.3 and 3.0 M for its PMS.

4Supermongo is a plotting software.

33 6.2. HERTZSPRUNG-RUSSELL DIAGRAM CHAPTER 6. HERTZSPRUNG-RUSSELL DIAGRAM

Figure 6.2: Hertzsprung-Russell diagram of the probables sources. The isochrones from Siess et al.(2000) are shown in green. They correspond to ages of 0.5, 1.5, 4.0, 10.0 and 20.0 Myr. The masses of the PMS evolutionary tracks are 0.3, 0.4, 0.5, 0.7, 1.0, 1.5, 2.0, 2.5, 3.0, 4.0, 5.0, 6.0 and 7.0 M .

34 6.2. HERTZSPRUNG-RUSSELL DIAGRAM CHAPTER 6. HERTZSPRUNG-RUSSELL DIAGRAM

Figure 6.3: Hertzsprung-Russell diagram of the ”questionable” sources.

35 Chapter 7

Inter-pointing variability

7.1 Correlation

Using the four files containing the lists of sources detected on individual observations, we searched for the variable sources. The files included the identification number of the source (specific to the given observation, see below), the right ascension and declination, the total count rate, the error on it, the count rates for each EPIC-instruments and the errors on them.

Since the source lists were specific to each observation, the identification number produced by the SAS routine (which is directly linked to the brightness of a star - the fainter it is, the bigger its identification number will be) of a given source did not correspond to the identification number of this particular source in another list. (If a star is variable and is brighter during a given observation, it will be detected earlier for this observation than for the other exposures where the star was fainter.) This means that we had to correlate the coordinates of a given source for a specific observation with the coordinates of this source for the other three datasets and with the complete list of sources (built from the merged images, comprising the probable and ”questionable” sources). For this purpose, we adopted a correlation radius of 5”12 , corresponding to the width of the PSF core of XMM-Newton.

As a result we obtained a file with four lines per source of the complete list, one line per observation. The lines specify the identification number of the source in the complete catalog, the identification number of the source in the different observations, the right ascension and declination, the count rates for each EPIC- instrument and the errors on them. (When a source was not detected in a particular observation, there was a line of zeros.) We noted that not all of the sources are detected on all of the observations and/or with all of the EPIC cameras. This is either because of genuine source variability, field rotation, the dead detectors of the MOS1 camera (that were damaged after the first observation) or the Large Window mode used for the MOS detectors on the last three observations.

1We tried using 4” as a correlation radius, but in that case several of the brightest sources were not found in the correlation. 2We also implemented a test in order to see if a source possessed several counterparts in the correlation radius. This situation occurs for three objets: the sources number 78 (from the third observation), number 138 (from the fourth observation) and number 266 (from the first observation). In these cases, we selected the counterpart with the highest count rate.

36 7.2. CHI SQUARED CHAPTER 7. INTER-POINTING VARIABILITY

7.2 Chi squared

From the count rates, we evaluated the weighted (accounting for the uncertainty on the count rate) mean 2 count rate for each EPIC instrument and computed the χ about these means:

P 2 j CRj (1/σj ) < CR >= P 2 (7.1) j(1/σj )

2 X (CRj− < CR >) χ2 = (7.2) σ2 j j Note that the j indices of the sums are restricted to those observations where the source was detected with the corresponding instrument.

7.3 Probability of inter-pointing variability

The χ2 values were then used to evaluate the probability of the null hypothesis. In our case, the null hypoth- esis is that the source has a constant count rate between the four observations.

The probability of inter-pointing variability can be obtained thanks to the Q function, the complement of the incomplete gamma function P (where P ≡ 1 - Q). Q is the probability of validity of the null hypothesis (and is based on the value of the χ2). To calculate it, we used the algorithm provided by Press et al.(1997).

In order to decide which sources should be studied in greater detail, we considered that a source was • definitively variable if Q < 1%; • probably variable if 1% < Q < 5 %.

In this way, we found that, for 120 sources, the null hypothesis has a probability of less than 1% for at least one instrument, i.e. we could reject the assumption that they are constant at the more than 99% confi- dence level.

About 21% of the full list of sources are variable. Out of the 352 probable sources, 117 (i.e. one third) are variable. If we restrict ourselves to the probable sources that are detected in more than one pointing (i.e. 278), the fraction of inter-pointing variables raises to 42% (i.e. 117/278). Out of the 204 ”questionable” sources, only 3 are variable. Again, eliminating the single detections from the list of ”questionable” sources, we obtain 3 variable objects among the 158 ”questionable” sources which have more than one detection. All these sources are listed in Table 7.1.

The first of these sources is the O-star binary 9 Sgr that was studied by Rauw et al.(2016) and we will not repeat its study here, as our focus is rather on the low-mass star population of NGC 6530. A fair fraction of the 74 probable and, to a lesser extent, of the 46 ”questionable” sources that have only a single detection may also be variable. Therefore, our fractions of inter-pointing variables are most likely lower lim- its. We thus conclude that inter-pointing variability is quite common among the X-ray sources of our sample.

37 7.3. PROBABILITY CHAPTER 7. INTER-POINTING VARIABILITY

Table 7.1: List of sources having a probability of the null hypothesis of a constant inter-pointing count rate of less than 1%. The columns yield the χ2 values for the count rates of the various EPIC instruments.

Source # pn MOS1 MOS2 Source # pn MOS1 MOS2 1 46.8 18.2 12.8 84 – 6.2 10.9 2 48.1 – – 86 19.1 10.6 17.6 3 9.7 – 4.9 90 7.3 – – 4 312.0 86.1 113.5 93 7.6 – 6.5 5 – – 5.3 95 7.9 – – 6 18.4 – – 97 4.1 – – 8 7.5 – – 98 34.9 20.1 12.2 10 50.8 – – 101 5.9 – 5.1 11 17.4 – – 104 18.5 – 9.1 14 – – 4.6 107 3.9 – – 16 11.8 7.1 6.5 110 3.8 – – 20 6.9 4.3 – 111 6.7 – – 22 4.2 – – 117 20.9 – 6.8 23 88.1 – – 119 83.4 – – 26 198.7 62.2 64.5 120 11.1 9.1 – 27 104.8 47.3 39.9 122 4.9 – – 29 5.7 – – 126 6.9 – – 30 10.6 – – 127 5.3 10.1 8.9 33 12.1 22.2 – 132 9.2 9.4 – 35 14.5 – – 135 19.9 – – 36 22.0 16.6 7.4 136 9.2 – – 42 41.7 5.8 13.1 138 110.5 – – 45 148.0 – – 142 3.9 13.6 32.2 47 – 6.5 – 144 6.4 5.6 – 48 20.1 – 9.7 147 15.0 – – 49 10.5 – 13.0 148 7.8 – – 55 – 5.5 – 151 6.5 – – 57 7.7 – – 152 21.2 – 11.2 61 8.0 – – 154 4.8 – – 62 6.5 – – 157 – 17.3 27.4 69 4.2 – – 158 29.7 10.4 – 72 5.5 – – 159 11.6 – – 75 25.6 7.8 13.4 160 – 5.1 5.6 76 11.2 – – 161 69.5 23.1 7.7 78 – 10.9 9.0 163 4.1 – – 80 8.1 – 8.5 164 9.1 5.9 – 83 7.8 – – 165 – 6.1 10.2

38 7.3. PROBABILITY CHAPTER 7. INTER-POINTING VARIABILITY

Table 7.1: Continued. Source # pn MOS1 MOS2 Source # pn MOS1 MOS2 166 – 16.8 – 279 24.4 7.9 10.2 167 6.1 9.6 4.7 283 5.9 – 4.7 169 3.8 – – 287 – – 5.3 179 13.2 – – 289 – 7.2 5.0 180 4.8 – – 291 – – 5.1 181 5.1 – – 298 6.9 – – 182 4.7 – – 299 5.2 – – 199 11.1 7.0 6.3 304 8.2 5.5 9.3 200 51.8 – – 307 13.8 7.8 5.0 204 7.8 – – 319 17.6 6.2 6.1 205 12.4 – 5.9 320 30.1 8.3 – 219 6.2 – – 328 4.5 – – 240 14.0 – – 329 32.2 11.7 12.9 244 8.7 – – 330 51.0 11.6 32.8 248 10.9 – – 335 – – 9.6 251 – – 8.6 340 – 5.3 – 253 4.4 – – 342 9.7 – – 257 12.1 8.8 5.4 343 – 5.2 – 261 – – 4.2 350 17.8 – – 264 22.7 82.2 86.4 351 5.9 – – 268 8.3 6.8 – 420 10.9 – – 270 8.4 5.5 7.4 437 29.7 – – 272 5.5 6.8 6.1 485 – 37.1 65.8

39 Chapter 8

Lightcurves

8.1 Extraction

For the sources and observations where we suspected variability based on the inter-pointing analysis (see Sect. 7.3), we traced their lightcurves, inspected them and noted if a flare seemed to be present. When it was the case, we fitted the exponential decay of those lightcurves to constrain the properties of the magnetic loops generating the X-ray emission.

In order to obtain lightcurves, we first had to define the position for the source extraction and a back- ground (over a source-free region)1 for each object using SAOImage DS9 (”SAOImage DS9”, 2017). We could not choose rings surrounding sources as backgrounds because of the surface density of the stars. We therefore selected circular backgrounds on the side of each source, in places where there were no other stars. Most of the time, we took radii of 200 ”physical” pixels (i.e. 10”) for the sources (to avoid an overlap with neighboring sources while still collecting the majority of the counts) and radii of 350 ”physical” pixels (i.e. 17.5”) for the backgrounds (to obtain better statistics).

Then, we used the evselect command from the SAS to extract the lightcurves of the selected sources (for the MOS1, MOS2 and pn instruments, for a given set of observations and for specific energy bands (here we chose the 0.5 - 10 keV, 0.4 - 1.4 keV and 1.4 - 8 keV bands) - while subtracting their background). The lightcurves were extracted for time bins of 1 ks.

We first chose to extract the lightcurves for three energy bands because we had planned to analyse the curves in each band. Unfortunately, because of the low statistics of the lightcurves, we ultimately chose to focus our attention on the total energy band (i.e. the 0.5 - 10 keV band) for our analysis.

In this way, we obtained EPIC lightcurves for a total of 75 sources, leading to 171 combinations of sources and observations.

1For a given source, we sometimes had to slightly change its coordinates and the coordinates of its background from one observation to another (in particular for the third observation because it has a shift in its coordinates).

40 8.2. ANALYSIS CHAPTER 8. LIGHTCURVES

8.2 Analysis

The background-corrected lightcurves in the 0.5 - 10 keV band were then inspected by eye (using super- mongo) to search for genuine flare events. Clear indications of such events were found for only a handful of sources. Many sources do display variability, but without obvious flaring behavior. Figure 8.1 provides some examples.

41 Figure 8.1: Examples of lightcurves. Top row: two sources that show no significant intra-pointing variability (see Table 8.2). Top left: source #8 pn curve during the four observations. Top right: same for source #11. The middle and bottom row illustrate lightcurves that display highly significant variability. Middle left: pn lightcurve of source #200 for the two observations where the source was detected. Middle right: EPIC lightcurves of source #10 during the fourth observation. Bottom left: same for source # 4 during the fourth observation. Bottom right: same for source #45 during the first observation. 8.2. ANALYSIS CHAPTER 8. LIGHTCURVES

In the case of source # 45, the lightcurve is reminiscent of a rotational modulation of an X-ray active region. Yet, the rotational period of 1.91 day (i.e. 165 ks) derived by Henderson & Stassun(2012) for the op- tical counterpart of this source seems too long to explain the variation (occurring on a timescale of ∼ 15 ks) in this way.

By far the clearest example of a flare is provided by source #26 during the first observation (see Fig. 8.2). Other examples are source #9 during observation 3 (only the decay phase is observed) and #36 during ob- servation 4. The case of source #23 is more ambiguous.

Figure 8.2: Lightcurves of the best candidate flaring sources. Top row: source #23 (left) and #26 (right) during the first observation. Bottom row: sources #9 during the third observation (left) and # 36 during the fourth observation (right).

43 8.2. ANALYSIS CHAPTER 8. LIGHTCURVES

We fitted the decay phases of the flaring sources in the pn lightcurves with an exponential

CR = CR0 + A exp [−(t − t0)/τ] (8.1) where CR0 is the count rate outside the flare, t0 is the time of the flare maximum, A its amplitude and τ is the decay time. The resulting decay times are given in Table 8.1.

Serio et al.(1991) modeled the magnetic loops associated with solar flares 2 as hydrodynamic processes in rigid tubes. They provided a relation, expressed in cgs units, linking the decay time τ to the loop half-length L (defined as the distance between one of the flare footpoints and its apex) and to the initial temperature at the top of the loop T0: 3.7 × 10−4 × L τ = √ (8.2) T0 This equation can be approximated by: p l ∼ 0.129 × τ × kTmax (8.3) where l is the the loop half-length (expressed in solar radii), τ is expressed in ks and kT is expressed in keV (Sana et al. 2007).

Table 8.1: Inferred properties of the plasma loops in the flaring sources. The values of kT were obtained thanks to spectral fitting (for the spectra fitted with double-temperature thermal plasma models, the hard component was used here) (see Sect. 9.2). l corresponds to the size of a half-loop (defined as the distance between one of the flare footpoints and its apex). The stellar radii and rotation periods are taken from Henderson & Stassun(2012), whereas the H α equivalent width comes from Kalari et al.(2015).

Source τ kT l R∗ Prot EW(Hα) (ks) (keV) (R )(R ) (days) #9 9.3 5.89 2.91 −90.3 #23 2.4 5.87 0.75 1.64 0.70 – #26 4.6 7.45 1.62 2.61 2.10 – #36 2.2 4.62 0.61 −20.5

The longest decay time and thus the largest half-loop size are observed for source #9 which has a rather large Hα equivalent width, suggesting that this is a classical T Tauri star surrounded by an accreting struc- ture. Whereas the source #36 has the smallest half-loop size, even though it presents (weaker) Hα emission. This makes a categorization more hazardous.

The sources #23 and #26 have half-loop sizes smaller than their stellar radii (and they present no Hα emission), implying that the magnetic loops are anchored at the stellar surfaces and that the sources might be weak-line T Tauri stars.

2Their results can be used for stellar flares as well.

44 8.3. INTRA-POINTING VARIABILITY CHAPTER 8. LIGHTCURVES

Unfortunately, the low number of flaring sources that we have found does not allow us to carry out a statistical study of the properties of the optical counterparts of these sources.

8.3 Intra-pointing variability

In addition to the visual inspection of the lightcurves, we applied a χ2 test to quantify the significance level of intra-pointing variability. For this purpose, we used a fortran code designed by Naze´ et al.(2013). This code tests several null hypotheses: a constant count rate, as well as linear or quadratic trends with time. The test is applied to the source lightcurve corrected for the background, the background lightcurve and the uncorrected source and background lightcurves.

The results for the test of the constant rate null hypothesis are presented in Table 8.2. We note that the different EPIC instruments do not always lead to the same conclusions. In some cases this can be due to the fact that the source lies near a gap between detectors or is located near the edge of the field of view. Since the pn camera receives in general more photons than the MOS instruments, the pn lightcurve usually has better statistics and should be given more weight than the EPIC-MOS data. We shall thus consider that sources with a detection of intra-pointing variability in one of the MOS cameras only are dubious.

Following the application of the test, we re-inspected all the lightcurves for which at least one of the instruments led to rejection of the null hypothesis at the < 1% level. We found that a number of false vari- ability alerts are associated with lightcurves that feature temporal bins with net (i.e. background-corrected) count rates below zero. This kind of situation can arise for weak sources which are barely detected above the background. Discarding these objects, we found that definite intra-pointing variability is present in sources #4 (obs. 4), #9 (obs. 3), #10 (obs. 4), #23 (obs. 1), #26 (obs. 1), #36 (obs. 4), #45 (obs. 1), and #200 (obs. 2). Hence 11, of the 75 sources for which we have extracted lightcurves display definite short-term (intra- pointing) variability. An additional 7 objects are candidates for such variability (#69, obs. 4; #75, obs. 1; #127, obs. 4; #132, obs. 4; #147, obs. 3; #204, obs. 4; #205, obs. 3). In total, we thus find that between 15 and 24% of the sources for which we have extracted lightcurves display short-term variations.

It should be noted that the source #26, which presented a flare during the first observation, was not detected as variable in the other three observations (where it displayed a much smaller count rate of about 0.005 counts per seconds compared to the count rate of 0.120 counts per second of the first observation). This means that this source did not present any detectable variability apart from this strong flare. This result shows that even sources which do not exhibit any significant short-term variability most of the time can undergo a very strong flare from time to time.

45 8.3. INTRA-POINTING VARIABILITY CHAPTER 8. LIGHTCURVES

Table 8.2: Results of the test for statistically significant intra-pointing variability. For each source for which at least one lightcurve was extracted, the results of the test of the null hypothesis of a constant count rate are expressed as percent of probability that the source was constant during a specific observation and as observed with a given EPIC instrument.

Source Obs. 1 Obs. 2 Obs. 3 Obs. 4 M1 M2 pn M1 M2 pn M1 M2 pn M1 M2 pn 2 99.9 30.1 12.8 45.9 3 2.0 46.1 98.6 3.3 93.4 6.6 13.8 18.7 0.2 0.0 34.3 43.2 4 23.0 49.8 56.3 22.7 21.5 2.5 96.8 1.1 8.5 0.0 0.0 0.0 6 20.9 61.2 12.1 46.0 8 68.8 9.6 36.1 11.2 9 0.0 10 2.5 0.0 0.0 11 58.6 53.2 29.2 12.1 16 0.3 5.3 90.1 20 0.0 1.3 86.6 23 0.3 0.0 0.0 26 0.0 0.0 0.0 27 31.8 29.7 26.0 33 41.1 33 54.2 36 0.0 0.0 0.0 42 12.5 0.2 81.0 41.1 81.0 0.3 36.0 45 0.0 0.0 0.0 48 88.4 0.3 13.7 0.0 44.2 66.4 10.2 18.1 43.1 0.0 32.5 73.5 49 53.4 57 7.6 55.1 38.2 61 16.8 69 48.6 77.3 8.8 0.0 72 84.3 75 0.0 11.6 1.6 76 11.5 80 0.7 86 84.6 0.0 81.2 37.6 8.5 69.9 98 66.5 87.9 55.7 104 35.2 111 54.5 117 4.4 96.9 79.8 17.5 119 34.7 127 62.6 7.1 0.6

46 8.3. INTRA-POINTING VARIABILITY CHAPTER 8. LIGHTCURVES

Table 8.2: Continued

Source Obs. 1 Obs. 2 Obs. 3 Obs. 4 M1 M2 pn M1 M2 pn M1 M2 pn M1 M2 pn 132 93.1 26.7 1.1 32.1 29.6 74.2 8.9 18.2 94.2 0.0 63.5 0.0 135 66.3 136 3.7 92.7 0.0 88.2 142 0.4 0.2 16.0 89.0 0.8 1.5 144 2.4 0.1 3.5 147 43.8 0.5 152 64.5 10.7 66.9 70.8 53.4 34.8 158 57.2 159 66.8 11.4 56.4 43.9 161 64.9 2.9 7.6 163 39.1 28.6 44.9 20.4 164 6.6 8.0 0.5 2.3 13.7 65.1 0.1 85.5 4.3 165 32.9 64.9 53.7 167 0.0 18.8 9.5 3.3 55.2 29.6 12.6 2.0 29.1 4.7 19.0 90.5 179 97.9 1.8 34.3 181 53.9 37.3 7.6 34.8 182 0.0 76.5 10.9 199 65.4 34.3 29.9 200 0.0 2.6 204 15.4 2.8 42.9 0.6 205 5.8 72.6 0.1 71.0 219 91.5 91.5 19.2 22.6 251 30.3 21.5 13.6 44.1 257 0.0 0.0 34.3 264 12.0 49.6 0.6 268 6.2 60.4 12.8 95.8 270 92.1 0.1 1.1 11.0 272 9.7 7.4 75.0 57.3 279 1.0 11.1 60.0 0.4 85.2 43.7 0.0 0.0 1.3 84.9 0.1 19.8 283 7.9 31.3 55.8 55.9 44.5 32.4 13.1 69.6 52.6 8.2 0.6 23.0 287 99.3 1.8 87.0 299 29.8 36.5 87.0 31.6 307 14.7 32.3 84.4 7.2 58.4 25.4 0.0 0.2 97.1 0.2 0.0 45.5 319 23.8 95.8 69.1 320 13.9 0.0 36.9 0.4 60.0 1.1 0.4 32.8 61.2 329 96.2 96.8 96.2 84.7 330 36.1 19.7 53.2 342 35.6 98.4 20.7 350 33.9 90.8 82.8 437 11.5 0.5 65.5 485 1.8 2.1 0.0 1.9 0.4 0.4

47 8.4. COMPARISON WITH OTHER STUDIES CHAPTER 8. LIGHTCURVES

8.4 Comparison with other studies

In this section, we compare the results we obtained thanks to the analysis of our lightcurves, namely con- cerning the magnetic loop sizes and the frequency of flares, to similar studies of other clusters.

8.4.1 Loop sizes Favata et al.(2005) used the flare model of Reale et al.(1997) to determine the size of half-loops of about 30 flaring events detected in the Orion Nebula Cluster. They found loops as short as a fraction of solar radii, as well as loops as long as 0.2 AU (i.e. 43 R ), which suggest a connection between the star and an accretion disk. In contrast, Getman et al.(2008) applied a modified version of this model on flaring events in the same cluster and found only evidence of flares having both footpoints anchored in the stellar chromosphere. It is interesting to note that Aarnio et al.(2010) claimed that 58% of the stars in this cluster displaying high- contrast flares are not surrounded by an accretion structure.

McCleary & Wolk(2011) analyzed 30 high-contrast flares (using the flare model of Reale et al.(1997) and a slightly different technique than Favata et al.(2005)) from different clusters observed with Chandra. They found loop lengths ranging from 0.03 to 3.0 R . They categorized their sources into class I (which con- tains two objects), class II (which consists of twelve objects) and class III (which contains twelve sources) objects.

Sana et al.(2007) studied the PMS stars of the young open cluster NGC 6231 by means of six XMM- Newton observations. They found loop half-lengths ranging from 0.3 to 2.4 R (and most of them smaller than the stellar radii, implying loops anchored in the stellar chromosphere) for their brightest flaring sources.

We found loop sizes ranging from 0.61 to 2.91 R , similar to the ones presented in the studies of Mc- Cleary & Wolk(2011) and Sana et al.(2007). We hypothesized that, out of 4 flaring sources, one might be a cTTs. This result is similar to the percentage given by Aarnio et al.(2010), although our statistical significance is quite low. It has to be stressed also that the duration of our observations could introduce a bias against the detection of flares with decay times in excess of ∼ 30 ksec. The same remark applies to the observations of Sana et al.(2007), Rauw(2011) and Rauw & Naz e´(2016).

8.4.2 Flaring frequency Sana et al.(2007) examined NGC 6231. They studied about 1.5 times our number of sources, for time peri- ods similar to ours and found 10 genuine flares.

Rauw(2011) studied the core of the Cygnus OB2 association (being an association and not a cluster, its density is smaller than the one of NGC 6530, the conditions of stellar formation are different and it does not present the same properties as our cluster). This association was observed at six occasions and Rauw(2011) found about 170 low-mass sources (which are mostly PMS), 8 of which presented flares.

Rauw & Naze´(2016), while studying the young open cluster IC1805 (which is similar to NGC 6530, though less densely populated), found a single flare out of about 170 PMS.

48 8.4. COMPARISON WITH OTHER STUDIES CHAPTER 8. LIGHTCURVES

We can use the approach of McCleary & Wolk(2011) to quantify the frequency with which strong flares occur in the stars of NGC 6530. For this purpose, we first multiply the number of X-ray sources likely associated with PMS stars by the total exposure time to estimate the ”on-star” time. If we restrict ourselves to the 352 confirmed sources and account for the fact that about 10 of these objects are associated with early- type stars (see Rauw et al. 2002), we come up with a total ”on-star” time of 342 × 87.8 ks = 30 Ms. Since we observe 3 clear flare events (leaving aside the case of source #23), this corresponds to one flare every 10 Ms. On average, a PMS star in NGC 6530 produces a strong flare every 116 days. This frequency is a factor 10 higher than the estimate of McCleary & Wolk(2011). Let us therefore repeat the same calculations for other clusters that were studied in a similar way to our study:

Table 8.3: Estimated flaring frequency for PMS stars in very young open clusters observed with XMM- Newton. N∗ stands for the number of X-ray selected PMS stars. texp and ton−star indicate the total integra- tion time and the ”on-star” time, respectively. Pflare finally yields the average time between two flares.

Cluster Age d N∗ texp ton−star Nflare Pflare Reference (Myr) (kpc) (s) (s) (s) NGC 6231 1 – 12 1.6 446 170.6 103 76.1 106 10 7.6 106 Sana et al. (2007) Cyg OB2 3 – 7 1.4 174 138.9 103 24.2 106 8 3.0 106 Rauw (2011) IC 1805 3.5 2.4 170 48.7 103 8.3 106 1 8.3 106 Rauw & Naze´ (2016) NGC 6530 0.3 – 10 1.3 342 87.8 103 30.0 106 3 107 this study

Whilst there clearly exists a dispersion among the various clusters as far as the flaring frequency is con- cerned, all clusters obviously yield frequencies that are significantly above the determination of McCleary & Wolk(2011).

The time-integrated fluxes of sources #26 and #36 during their flares are about 2.14 10−8 and 7.5 10−9 erg cm−2. Adopting a distance of 1.3 kpc, these values correspond to total flare energies of 4.3 1036 and 1.5 1036 erg, i.e. in the same range of energies as the flares studied by McCleary & Wolk(2011). Hence the difference in flaring frequency between our study and the work of McCleary & Wolk(2011) cannot be attributed to a difference in the total energy of the events that were studied. Part of this difference might come from the superior angular resolution of Chandra compared to XMM-Newton. This allows Chandra to detect more (and actually fainter) X-ray sources. Therefore the total ”on-star” time of a given observation is larger, thus leading to a lower flaring frequency.

Finally, we can compare the frequency of our flares to the flaring activity of a star that has been studied quite closely: our Sun.

Statistics show that the average occurence of flares at solar minimum is of about once per day, whereas at solar maximum it is of about 20 per day. It needs to be noted, however, that the flaring frequency and intensity is very irregular (Crosby et al. 1993; ”Questions and Answers” 2017).

Moreover, the rate of solar flares also depends on the flare intensity. Li et al.(2016) gave a power-law function: dN = Ax−δdx (8.4)

49 8.4. COMPARISON WITH OTHER STUDIES CHAPTER 8. LIGHTCURVES where dN represents the number of events, x (which is positive) refers to the magnitude of the event, A is a constant positive coefficient and δ is the constant power-law index. They specified that X-ray peak flux of solar flaring events has a power-law frequency distribution3 (i.e. the bigger flares are more spaced in time, whereas the smaller flares are numerous) with a power-law index ranging from 1.6 to 2.1 (depending on the study).

Figure 8.3: Exemple of an histogram (from GOES (or Geostationary Operational Environmental Satellite) data) of the number of solar flares as a function of the flux fitted by a power-law model (Li et al. 2016).

The X-ray emission of solar flares can last from about 10 minutes to several hours (Shakhovskaya & Akhtemov 2013). By comparison, the decay times of the flaring sources of NGC 6530 that we detected ranged from about 2 ks to about 9 ks (i.e. from about 30 minutes to 2.5 hours).

For the NGC 6530 cluster, we found 3 clear flaring sources out of 352 probable sources, while this cluster was observed at four occasions, for about 20 ks (i.e. about 5.5 hours). The small number of flares that were detected yields a flaring frequency well below that of the Sun at solar minimum. However, in this comparison, one needs to keep in mind the power-law distribution of solar flares: since PMS are much farther away than the Sun, only their strongest flares will be detected, and they are expected to be the rarest.

The fluxes of sources #26 and #36 during the initial phases of the flares (up to 1 × τ) are about 4.7 10−12 and 3.4 10−12 erg cm−2 s−1, i.e. 4.7 10−15 and 3.4 10−15 W m−2. If these objects were located at 1 AU (in- stead of 1.3 kpc), the GOES satellite would have measured a flux 7.2 1016 times higher, i.e. their equivalent flux at 1 AU would be around 300 W m−2. We thus see that these flares are far more energetic than the most energetic solar flares measured by GOES and shown in Fig. 8.3. Since the frequency of solar flares drops with increasing energy, the vast majority of the events that make up the flaring activity of the Sun are in fact low energy events that are well below the detection threshold for distant stars.

3They indicated that, at the lower end of the magnitude spectrum, the distribution deviates from a power-law distribution (but this might be imputed to observational bias).

50 Chapter 9

Spectra

9.1 Extraction

For the sources that have a sufficiently large count rate in order for their EPIC spectra of a given observation to contain at least several hundred counts (i.e. 28 sources, leading to 51 combinations of sources and obser- vations), we extracted the spectra of the source and of the background (over a source-free region)1 using the evselect command from the SAS. We could not choose rings surrounding sources as backgrounds because of the proximity of other sources. We therefore selected circular backgrounds on the side of each source, in places where there were no other stars. Most of the time, we took radii of 200 ”physical” pixels (i.e. 10”) for the sources (to avoid an overlap with neighboring sources while still collecting the majority of the counts) and radii of 350 ”physical” pixels (i.e. 17.5”) for the backgrounds (to obtain better statistics).

For each spectrum, we further generated an appropriate response matrix file (rmf)2, allowing the energy calibration of the data, and an ancillary response file (arf)3, which provides the flux calibration of the spectra. The background corrected, energy- and flux-calibrated spectra were binned to have at least 25 counts per bin and to oversample the spectrum by at most a factor 5. We followed those steps for the MOS1, MOS2 and pn instruments.

9.2 Spectral fits

The spectra were then fitted using the xspec software (version 12.9.0i). For spectra with less than 15 spectral bins, we fitted the data assuming an absorbed single-temperature thermal plasma model. For spectra with more bins, we tested the presence of a second temperature component.

1For a given source, we sometimes had to slightly change its coordinates and the coordinates of its background from one observation to another (in particular for the third observation because it has a shift in its coordinates). 2Since the detection of photons is a quantum process, each time a photon possessing a specific energy arrives on a detector, the number of free electrons generated is not exactly the same. The rmf expresses the probability that a photon of a given energy generates a given number of free electrons (Rauw 2016-2017). 3The arf is the effective area of the instrument as a function of the energy (for a specific position on the detector), it measures the number of photons collected per square centimeters for a given energy (expressed in keV) (Rauw 2016-2017).

51 9.2. SPECTRAL FITS CHAPTER 9. SPECTRA

Source #9, obs. 3 Source #23, obs. 1

0.01

5×10−3 0.01 −1 −1

keV keV −3 −1 −1 2×10

10−3

−3 10 5×10−4 normalized counts s normalized counts s

2×10−4 4 2 2 2 χ χ

2 ∆ ∆ 0 0

−2 −2

−4 −4 sign(data−model) × sign(data−model) × 1 10 1 10 Energy (keV) Energy (keV)

Source #26, obs. 1 Source #36, obs. 4

0.01

5×10−3 −1 −1

keV 0.01 −1

keV −3 −1 2×10

10−3

10−3 5×10−4 normalized counts s normalized counts s

2×10−4 4 2 χ

2 ∆ χ

∆ 2 0 0

−2 −5 sign(data−model) × sign(data−model) × −4 1 10 1 10 Energy (keV) Energy (keV)

Figure 9.1: Spectral fits (assuming emission from plasma at two different temperatures) of four of the sources that display flares (#9, 26 and 36) or flare-like variations (#23). Except for the upper left panel, where the black line corresponds to the EPIC-pn spectrum, the colors correspond to MOS1 (black), MOS2 (red) and pn (green).

The results of the various fits are shown in Table 9.1 and some examples of spectral fits are shown in Fig. 9.1. Not all of the fits are of acceptable quality. For sources with a relatively low count rate, residual straylight contamination probably affects the quality of the spectral fits.

Considering values between 0.3 and 0.35 mag for the color excess E(B-V) (see Tothill et al. 2008), and the value of 5.8 × 1021 atoms cm−2 mag−1 for the ratio between the neutral hydrogen column density and E(B-V) from Bohlin et al.(1978), we calculated the mean value of the interstellar column density between 21 −2 22 −2 21 −2 NH = 0.3 × 5.8 × 10 atoms cm = 0.174 × 10 atoms cm and NH = 0.35 × 5.8 × 10 atoms cm = 0.203 × 1022 atoms cm−2.

We expected our NH values from the spectral fits to be of the same magnitude as the mean value derived above, however, our values show a very large dispersion (see Fig. 9.2) (some values are larger than 1022, some are around 1021, one is as small as 4 × 1020atoms cm−2) and most are larger than the mean value.

52 9.2. SPECTRAL FITS CHAPTER 9. SPECTRA

Table 9.1: Results of the spectral fits of the brightest sources assuming an absorbed 1-T or 2-T ther- mal plasma model (tbabs*apec or tbabs*(apec+apec)) (we assumed solar abundances (Anders −14 R 10 ne nH dV & Grevesse 1989).). The normalization of the apec models corresponds to d2 where d is the −3 distance of the source (in cm), ne and nH are the electron and hydrogen densities of the source (in cm ). The fluxes fX correspond to the observed fluxes in the 0.5 - 10.0 keV energy band.

2 Source # Obs. # Instr. NH kT1 norm1 kT2 norm2 χν d.o.f. fX (1022 cm−2) (keV) (keV) (10−14 erg cm−2 s−1) 2 1 EPIC 1.08 ± 0.22 0.12 ± 0.07 1.92 10−2 1.38 ± 0.22 1.59 10−4 1.58 30 9.5 2 2 pn 0.91 ± 0.17 0.23 ± 0.05 1.27 10−3 2.56 ± 0.30 2.07 10−4 2.12 18 20.1 2 3 pn 0.25 ± 0.06 2.60 ± 0.34 5.23 10−5 1.64 7 5.9 2 4 pn 0.11 ± 0.04 2.92 ± 0.52 4.02 10−5 1.26 6 5.4 3 1 EPIC 1.02 ± 0.08 0.80 ± 0.12 2.10 10−4 2.20 11 7.3 3 2 EPIC 1.03 ± 0.11 0.56 ± 0.14 2.00 10−4 2.79 6 4.0 3 3 EPIC 0.90 ± 0.22 0.49 ± 0.34 1.34 10−4 0.97 3 2.7 3 4 EPIC 0.80 ± 0.08 0.73 ± 0.09 1.23 10−3 4.00 12 5.2 4 1 EPIC 1.12 ± 0.21 0.14 ± 0.02 1.40 10−2 1.11 ± 0.65 4.05 10−5 0.56 11 4.1 4 2 EPIC 0.87 ± 0.10 0.14 ± 0.01 1.07 10−2 1.61 ± 0.09 2.20 10−4 1.85 47 17.0 4 3 EPIC 0.93 ± 0.08 0.14 ± 0.01 1.20 10−2 1.70 ± 0.10 1.58 10−4 1.54 41 13.0 9 3 pn 0.41 ± 0.03 5.89 ± 0.65 4.10 10−4 0.92 41 60.7 11 2 pn 0.61 ± 0.16 0.19 ± 0.03 5.71 10−4 1.99 ± 0.18 2.31 10−4 2.32 22 20.6 11 3 pn 0.79 ± 0.12 0.20 ± 0.04 1.86 10−3 2.00 ± 0.14 2.35 10−4 1.19 25 21.2 11 4 pn 0.31 ± 0.10 0.38 ± 0.11 4.20 10−5 2.58 ± 0.23 2.05 10−4 1.16 27 24.5 23 1 EPIC 0.21 ± 0.06 1.06 ± 0.16 6.69 10−6 5.87 ± 1.83 7.44 10−5 1.34 17 13.0 26 1 EPIC 0.27 ± 0.02 1.23 ± 0.21 3.86 10−5 7.45 ± 1.59 6.28 10−4 1.10 72 110.0 27 2 EPIC 0.27 ± 0.04 0.95 ± 0.12 2.91 10−5 3.01 ± 0.31 2.61 10−4 1.08 39 35.0 36 4 EPIC 1.24 ± 0.23 0.12 ± 0.03 2.80 10−2 4.62 ± 1.03 2.76 10−4 1.27 29 30.7 45 1 EPIC 0.73 ± 0.18 0.34 ± 0.12 2.13 10−4 3.63 ± 0.56 2.44 10−4 1.08 44 29.0 75 1 EPIC 0.28 ± 0.07 0.71 ± 0.13 1.41 10−5 2.48 ± 0.31 9.58 10−5 1.26 31 12.0 86 1 EPIC 0.53 ± 0.29 0.25 ± 0.10 1.58 10−4 2.07 ± 0.31 1.15 10−4 1.17 17 12.0 86 4 EPIC 0.36 ± 0.33 0.28 ± 0.15 2.75 10−5 2.56 ± 0.79 6.83 10−5 1.16 15 8.0 98 2 EPIC 0.73 ± 0.16 0.23 ± 0.04 4.16 10−4 2.27 ± 0.25 1.88 10−4 1.07 36 17.0 119 4 pn 0.81 ± 0.24 0.22 ± 0.05 5.85 10−4 2.44 ± 0.36 1.81 10−4 0.55 14 17.0 127 4 EPIC 1.04 ± 0.17 0.14 ± 0.02 1.25 10−2 1.72 ± 0.31 1.14 10−4 0.79 15 8.8 135 2 EPIC 0.84 ± 0.19 0.19 ± 0.03 1.38 10−3 2.00 ± 0.21 1.34 10−4 1.81 11 11.0 152 1 EPIC 1.26 ± 0.37 0.11 ± 0.04 1.08 10−1 2.00 ± 0.83 2.75 10−4 1.79 16 18.7 167 1 M1 & pn 0.80 ± 0.10 0.80 ± 0.11 1.35 10−4 2.94 6 6.3 167 2 EPIC 0.51 ± 0.14 0.98 ± 0.10 5.71 10−5 1.72 7 4.9 167 3 EPIC 0.87 ± 0.15 0.66 ± 0.34 1.28 10−4 2.31 6 4.3 167 4 EPIC 0.96 ± 0.13 0.58 ± 0.18 1.37 10−4 1.79 5 3.3 179 1,2,4 pn 0.59 ± 0.19 0.23 ± 0.04 2.08 10−4 2.01 ± 0.24 4.48 10−5 3.13 21 5.3 200 2,3 pn 1.08 ± 0.09 0.17 ± 0.01 6.23 10−3 2.42 ± 0.39 1.04 10−4 2.08 25 10.8 264 1 MOS 0.16 ± 0.04 3.16 ± 0.47 1.31 10−4 0.90 10 17.3 +0.31 0.28 1.61 ± 0.54 3.34 10−5 279 1 EPIC −0.28 1.87 5 3.4 279 2 EPIC 0.08 ± 0.04 2.85 ± 0.75 4.46 10−5 0.84 7 6.1 279 3 EPIC 0.08 ± 0.02 3.07 ± 0.34 9.77 10−5 1.49 21 13.8 279 4 EPIC 0.05 ± 0.03 2.28 ± 0.41 3.84 10−5 0.92 10 5.2 283 1 EPIC 1.03 ± 0.11 0.65 ± 0.18 2.82 10−4 0.60 7 7.2 +0.09 0.07 1.84 ± 0.54 4.12 10−5 283 2 MOS −0.07 0.50 3 5.5 283 3 EPIC 0.14 ± 0.03 2.56 ± 0.30 7.32 10−5 2.12 17 9.2 283 4 EPIC 0.04 ± 0.02 2.62 ± 0.30 4.48 10−5 2.51 19 6.4 307 1 EPIC 0.83 ± 0.11 0.62 ± 0.18 9.61 10−5 1.64 7 3.2 307 2 EPIC 1.12 ± 0.15 0.22 ± 0.06 3.40 10−3 2.82 ± 1.17 1.04 10−4 1.55 15 13.0 307 3 EPIC 0.89 ± 0.16 0.96 ± 0.15 1.23 10−4 1.42 8 6.0 307 4 EPIC 0.88 ± 0.09 0.73 ± 0.10 1.13 10−4 3.03 11 4.2 330 1 EPIC 1.02 ± 0.10 0.68 ± 0.13 2.15 10−4 1.72 10 6.0 437 3 EPIC 0.58 ± 0.05 0.19 ± 0.01 2.63 10−3 1.98 ± 0.16 9.37 10−5 2.82 53 19.0 485 1 EPIC 0.93 ± 0.10 0.69 ± 0.13 2.43 10−4 1.76 8 7.8 485 4 MOS 0.82 ± 0.53 0.96 ± 0.91 7.51 10−5 0.13 3 3.7

53 9.2. SPECTRAL FITS CHAPTER 9. SPECTRA

8

6

4

2

0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 NH

Figure 9.2: Histogram of the different values of NH.

This surprising result could come from the ”usual” degeneracy between NH and kT for low quality CCD spectra. However, another effect, linked to the properties of the Lagoon Nebula, could contribute to our high mean value. Indeed, M8 is an HII region where the gas is partially ionized by 9 Sgr and its radiation destroys dust particles. Therefore, while we calculated the mean value of NH based on the average gas-to-dust ratio of the interstellar medium, the local absorption contains less dust than the galactic average. Furthermore, the Lagoon Nebula is not an homogeneous environment as some areas contain more matter and this can explain the dispersion found in our values.

The lowest values of NH are associated with kT values that are quite large (e.g. sources 279 and 283). Conversely, very large values of NH go along with very soft plasma components. This situation could stem from the degeneracy between NH and kT.

By looking at Table 9.2 or at Fig. 9.3, we can see that for the spectra fitted with a single-temperature ther- mal plasma model, kT oscillates between a relatively low and a slightly higher temperature; whereas for the spectra fitted with double-temperature thermal plasma model, there is always one temperature that is very low and the other one that is higher. (We can also note that kT1 is lower than 1 keV (with two exceptions).)

54 9.2. SPECTRAL FITS CHAPTER 9. SPECTRA

Table 9.2: Values of kT for the different models.

Bins kT kT1 kT2 for the 1-T thermal plasma model for the 2-T thermal plasma model for the 2-T thermal plasma model (keV) (keV) (keV) (keV) 0.0 - 0.5 1 21 0 0.5 - 1.0 14 2 0 1.0 - 1.5 0 2 2 1.5 - 2.0 2 0 8 2.0 - 2.5 1 0 6 2.5 - 3.0 5 0 4 3.0 - 3.5 2 0 1 3.5 - 4.0 0 0 1 4.0 - 4.5 0 0 0 4.5 - 5.0 0 0 1 5.0 - 5.5 0 0 0 5.5 - 6.0 1 0 1 6.0 - 6.5 0 0 0 6.5 - 7.0 0 0 0 7.0 - 7.5 0 0 1

20

15

10

5

0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 kT kT1 kT2

Figure 9.3: Histogram of the different values of kT.

55 9.3. COMMENTS ON FLARING SOURCES CHAPTER 9. SPECTRA

9.3 Comments on flaring sources

The flaring sources (#9 during the third observation, source #23 and #26 during the first observation, # 36 during the fourth observation) do not present peculiar spectral characteristics (compared to the other sources whose spectra were fitted). We note however that these sources have rather hard spectra as revealed by the rather high values of kT2 (and its associated normalization). This is certainly not unexpected as the X-ray emission usually hardens during a flare event. The spectra of source #9 is the only one that was fitted with a single-temperature thermal plasma model4.

By comparison, Sana et al.(2007) found that most of NGC 6231flaring sources spectra were best fitted with a single-temperature thermal plasma model.

4This source is also presumed to be the only cTTs. However, drawing any conclusion would be hazardous because of the small number of sources displaying flares.

56 Chapter 10

Conclusions

We studied XMM-Newton observations of NGC 6530. We found that the rotation rates of the X-ray se- lected PMS in this cluster were not more important than the ones presented for the full sample of cluster members studied by Henderson & Stassun(2012). We obtained ages between 1.5 and 10 Myr, and masses ranging from 0.5 to 1.0 M , similar to the ones found in the literature. We found that classical T Tauri stars were outnumbered by the weak-line T Tauri stars, a result that is again in agreement with previous studies of NGC 6530 and of other clusters. Out of 352 probables sources, 117 are variable and 4 present flares (only one of which is presumably a cTTs). This number of flares might seem low, but can be compared to similar studies of other young open clusters and to the most energetic flares on the Sun. We found NH values exceeding the values expected for the mean E(B-V). This, along with the fact that lower values of NH are associated with higher values of kT (and vice versa) could stem from the degeneracy between NH and kT.

How could we obtain better results?

It would be interesting to observe the cluster for a significantly longer duration (i.e. for instance, observe it for 100 ks instead of 20 ks). This would increase the sensibility to flares because it would allow the de- tection of flares whose decay time is longer than 20 ks. This, in turn, would reveal the true number of flares (including the ones presenting a very long decay time) and therefore eliminate the bias coming from our observations durations.

Additionally, performing numerous observations would facilitate the detection of variability.

Finally, once the Athena X-ray Observatory will be launched into orbit, we will be able to re-observe NGC 6530 with the Athena Wide Field Imager and obtain far better statistics for the spectra - which would allow the removal of the ambiguity for the NH and kT - and for the lightcurves - which would enable the detection of flares of lower amplitude (”Athena X-ray observatory”, 2017).

57 Thanks

I would like to thank Gregor Rauw, my master’s thesis director, for his availability, his time and his guidance. I would like to thank Yael¨ Naze´ as well. I would also like to thank my family, in particular my grand-father Kekˆ eˆ (1936.10.22 - 2017.03.22) and my grand-mother Mem´ e´ (b. 1938.03.07), for their support. Finally, I would like to thank my friends for their encouragement, in particular Hanako for her help, Zazou for her support and Namnam for his presence.

58 Bibliography

Aarnio, Alicia N., Stassun, Keivan G., & Matt, Sean P. 2010, in A Search for Star-Disk Interaction among the Strongest X-ray Flaring Stars in the Orion Nebula Cluster, The Astrophysical Journal, Volume 717, Issue 1, pp. 93-106 Anders, Edward, & Grevesse, Nicolas 1989, in Abundances of the elements: Meteoritic and solar, Geochim- ica et Cosmochimica Acta, Volume 53, Issue 1, pp. 197-214

Bohlin, R. C., Savage, B. D., & Drake, J. F. 1978, in A survey of interstellar H I from Lalpha absorption measurements. II., Astrophysical Journal, Vol. 224, pp. 132-142 Crosby, N., Aschwanden, M., & Dennis, B. 1993, in Frequency distributions of solar X-ray flare parameters, Advances in Space Research, Volume 13, Issue 9, p. 179-182 Cutri, R. M., Skrutskie, M. F., van Dyk, S., Beichman, C. A., Carpenter, J. M., Chester, T., Cambresy, L., Evans, T., Fowler, J., Gizis, J., Howard, E., Huchra, J., Jarrett, T., Kopan, E. L., Kirkpatrick, J. D., Light, R. M., Marsh, K. A., McCallon, H., Schneider, S., Stiening, R. Sykes, M., Weinberg, M., Wheaton, W. A., Wheelock, S., & Zacarias, N. 2003, in 2MASS All Sky Catalog of point sources Damiani, F., Flaccomio, E., Micela, G., Sciortino, S., Harnden, F. R., Jr., & Murray, S. S. 2004, in A deep Chandra X-ray observation of the rich young cluster NGC6530. I. The X-ray source catalog and the cluster population, The Astrophysical Journal, Volume 608, Issue 2, pp. 781-796 Dupret, M. A. 2015-2016, in Structure et evolution´ des etoiles´ , ULg, Liege` Favata, F., Flaccomio, E., Reale, F., Micela, G., Sciortino, S., Shang, H., Stassun, K. G., & Feigelson, E. D. 2005, in Bright X-Ray Flares in Orion Young Stars from COUP: Evidence for Star-Disk Magnetic Fields?, The Astrophysical Journal Supplement Series, Volume 160, Issue 2, pp. 469-502 Feigelson, E. D., & Montmerle, T. 1999, in High-Energy Processes in Young Stellar Objects, Annual Review of Astronomy and Astrophysics, Vol. 37, pp. 363-408 Getman, Konstantin V., Feigelson, Eric D., Broos, Patrick S., Micela, Giuseppina, & Garmire, Gordon P. 2008, in X-Ray Flares in Orion Young Stars. I. Flare Characteristics, The Astrophysical Journal, Volume 688, Issue 1, article id. 418-436 Getman, Konstantin V., Feigelson, Eric D., Micela, Giusi, Jardine, Moira M., Gregory, Scott G., & Garmire, Gordon P. 2008, in X-Ray Flares in Orion Young Stars. II. Flares, Magnetospheres, and Protoplanetary Disks, The Astrophysical Journal, Volume 688, Issue 1, article id. 437-455

59 BIBLIOGRAPHY BIBLIOGRAPHY

Gudel,¨ Manuel, & Naze,´ Yael¨ 2009, in X-ray spectroscopy of stars, The Astronomy and Astrophysics Re- view, Volume 17, Issue 3, pp.309-408 Henderson, Calen B., & Stassun, Keivan G. 2012, in Time-series Photometry of Stars in and around the Lagoon Nebula. I. Rotation Periods of 290 Low-mass Pre-main-sequence Stars in NGC6530, The Astro- physical Journal, Volume 747, Issue 1, article id. 51

Jeffries, R. D., Thurston, I M. R., & Pye, J. P. 1997, in An X-ray survey of the young open cluster NGC2516, Mon. Not. R. Astron. Soc. 287, 350-380 Kalari, V. M., Vink, J. S., Drew, J. E., Barentsen, G., Drake, J. J., Eisloffel,¨ J., Mart´ın, E. L., Parker, Q. A., Unruh, Y. C., Walton, N. A., & Wright, N. J. 2015, in Classical T Tauri stars with VPHAS+ - I. Hα and u-band accretion rates in the Lagoon Nebula M8, Monthly Notices of the Royal Astronomical Society, Volume 453, Issue 1, pp. 1026-1046 Kenyon, Scott J., & Hartmann, Lee 1995, in Pre-Main-Sequence Evolution in the Taurus-Auriga Molecular Cloud, Astrophysical Journal Supplement v.101, p.117 Kumar, Dewangan Lokesh, & Anandarao, B. G. 2010, in A study of the massive star-forming region M8 using images from the Spitzer Infrared Array Camera, Monthly Notices of the Royal Astronomical Society, Volume 407, Issue 2, pp. 1170-1181 Li, You-ping, Feng, Li, Zhang, Ping, Liu, Siming, & Gan, Weiqun 2016, in On the Power-Law Distribu- tions of X-ray Fluxes from Solar Flares Observed with GOES, Research in Astronomy and Astrophysics, Volume 16, Issue 10, article id. 161

McCleary, J. E., & Wolk, S. J. 2011, in A Survey of High-contrast Stellar Flares Observed by Chandra, The Astronomical Journal, Volume 141, Issue 6, article id. 201, 17 pp. Naze,´ Y., Oskinova, L.M., & Gosset, E. 2013, in A Detailed X-Ray Investigation of ζ Puppis. II. The Vari- ability on Short and Long Timescales, The Astrophysical Journal, Volume 763, Issue 2, article id. 143, 21 pp.

Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. 1997, in Numerical Recipes in Fortran 77: The Art of Scientific Computing, Cambridge University Press Prisinzano, L., Damiani, F., Micela, G., & Sciortino, S. 2005, in The star formation region NGC6530: Distance, ages and initial mass function, Astronomy and Astrophysics, v.430, pp. 941-957

Prisinzano, L., Damiani, F., Micela, G., & Pillitteri, I. 2007, in VLT/Flames observations of the star forming region NGC6530, Astronomy and Astrophysics, Volume 462, Issue 1, January IV 2007, pp. 123-138 Prisinzano, L., Micela, G., Sciortino, S., Affer, L., & Damiani, F. 2012, in Spectral classification and HR diagram of pre-main sequence stars in NGC6530, Astronomy and Astrophysics, Volume 546, id.A9

Rauw, G., Naze,´ Y., Gosset, E., Stevens, I.R., Blomme, R., Corcoran, M. F., Pittard, J. M., & Runacres, M. C. 2002, in An XMM-Newton observation of the Lagoon Nebula and the very young open cluster NGC6530, A&A 395, 499-513 Rauw, G., De Becker, M., Gosset, E., Pittard, J. M., & Stevens, I. R. 2003, in An XMM-Newton observation of the very young open cluster NGC 6383, Astronomy and Astrophysics, v.407, p.925-934

60 BIBLIOGRAPHY BIBLIOGRAPHY

Rauw, G. 2011, in A multi-epoch XMM-Newton campaign on the core of the massive Cygnus OB2 associa- tion, Astronomy & Astrophysics, Volume 536, id.A31, 15 pp. Rauw, G., Blomme, R., Naze,´ Y., Spano, M., Mahy, L., Gosset, E., Volpi, D., van Winckel, H., Raskin, G., & Waelkens, C. 2016, in Testing the theory of colliding winds: the periastron passage of 9 Sagittarii. I. X-ray and optical spectroscopy, Astronomy & Astrophysics, Volume 589, id.A121, 14 pp.

Rauw, G., & Naze,´ Y. 2016, in X-ray and optical spectroscopy of the massive young open cluster IC 1805, Astronomy & Astrophysics, Volume 594, id.A82, 19 pp. Rauw, G. 2016-2017, in High-energy astrophysics, ULg, Liege` Reale, F., Betta, R., Peres, G., Serio, S., & McTiernan, J. 1997, in Determination of the length of coro- nal loops from the decay of X-ray flares I. Solar flares observed with YOHKOH SXT., Astronomy and Astrophysics, v.325, p.782-790 Sana, H., Rauw, G., Sung, H., Gosset, E., & Vreux, J. -M. 2007, in An XMM-Newton view of the young open cluster NGC 6231 - III. Optically faint X-ray sources, Monthly Notices of the Royal Astronomical Society, Volume 377, Issue 3, pp. 945-956.

Schulz, N. S. 2012, in The Formation and Early Evolution of Stars: From Dust to Stars and Planets, eds. Springer Serio, S., Reale, F., Jakimiec, J., Sylwester, B., & Sylwester, J. 1991, in Dynamics of flaring loops. I - Thermodynamic decay scaling laws, Astronomy and Astrophysics, vol. 241, no. 1, Jan. 1991, p.197-202

Shakhovskaya, A. N., & Akhtemov, Z. S. 2013, in The relationship between the duration of solar X-ray flares and the mass of associated coronal ejections, Bulletin of the Crimean Astrophysical Observatory, Volume 109, Issue 1, pp. 86-89 Siess, L., Dufour, E., & Forestini, M. 2000, in An internet server for pre-main sequence tracks of low- and intermediate-mass stars, Astronomy and Astrophysics, v.358, p.593-599

Snowden, S., Valencic, L., Perry, B., Arida, M., & Kuntz, K. D. 2016, in The XMM-Newton ABC Guide: An Introduction to XMM-Newton Data Analysis

Stelzer, B., & Neuhauser,¨ R. 2001, in X-ray emission from young stars in Taurus-Auriga-Perseus: Luminosity functions and the rotation - activity - age - relation, Astronomy and Astrophysics, v.377, p.538-556

Sung, Hwankyung, Chun, Moo-Young, & Bessell, Michael S. 2000, in UBVRI and Hα Photometry of the Young Open Cluster NGC6530, The Astronomical Journal, Volume 120, Issue 1, pp. 333-348 Tothill, N. F. H., Gagne,´ M., Stecklum, B., & Kenworthy, M. A. 2008, in The Lagoon Nebula and its Vicinity, in Handbook of Star Forming Regions Vol. II, 533 http://www.physicsoftheuniverse.com/topics_blackholes_stars.html, Stars, Su- pernovas and Neutron Stars - Black Holes and Wormholes - The Physics of the Universe, accessed on February 23, 2017 http://www.messier-objects.com/messier-8-lagoon-nebula/, Messier 8: Lagoon Nebula — Messier Objects, accessed on May 3, 2017

61 BIBLIOGRAPHY BIBLIOGRAPHY http://chandra.harvard.edu/about/, Chandra :: About Chandra, accessed on April 8, 2017 http://www.cosmos.esa.int/web/xmm-newton, XMM-Newton > XMM-Newton SOC Home Page, accessed on October 1, 2016 https://www.cosmos.esa.int/web/xmm-newton/gallery-level3?p_p_id=48_INST ANCE_6yJhevwhnoHH&_48_INSTANCE_6yJhevwhnoHH_iframe_id=1017, XMM-Newton > XMM-Newton Image Gallery, accessed on February 10, 2017 https://www.cosmos.esa.int/web/xmm-newton/technical-details-mirrors, Technical Details - Mirrors - Cosmos, accessed on April 30, 2017 https://www.cosmos.esa.int/web/xmm-newton/technical-details-spacecraft, Technical Details - Spacecraft - Cosmos, accessed on May 3, 2017 ”XMM-Newton Users Handbook”, Issue 2.14, 2016 (ESA: XMM-Newton SOC) http://www.cosmos.esa.int/web/xmm-newton/technical-details-epic, XMM- Newton > Technical Details - EPIC, accessed on November 21, 2016

”Users Guide to the XMM-Newton Science Analysis System”, Issue 12.0, 2016 (ESA: XMM-Newton SOC) http://xmm-tools.cosmos.esa.int/external/sas/current/doc/emevents/node 4.html, Patterns, accessed on February 14, 2017 ”Specifications for individual SSC data products”, Issue 4.2, 2015 http://ds9.si.edu/site/Home.html, SAOImage DS9, accessed on February 15, 2017 http://cxc.harvard.edu/ciao/ahelp/dmimg2jpg.html, Ahelp: dmimg2jpg - CIAO 4.9, accessed on April 30, 2017 http://vizier.u-strasbg.fr/viz-bin/VizieR, VizieR Service, accessed on February 15, 2017 http://www.spitzer.caltech.edu/mission/32-The-Mission, Mission Overview - NASA Spitzer Space Telescope, accessed on April 30, 2017 http://www.spitzer.caltech.edu/mission/398-The-Infrared-Array-Camera-I RAC-, The Infrared Array Camera (IRAC) - NASA Spitzer Space Telescope, accessed on April 30, 2017 http://www.astro.princeton.edu/˜rhl/sm/, SM, accessed on April 6, 2017 https://hesperia.gsfc.nasa.gov/sftheory/questions.htm, Questions and Answers, accessed on May 15, 2017 http://www.the-athena-x-ray-observatory.eu/mission.html, Athena X-ray observa- tory, accessed on May 17, 2017

62