University of Amsterdam
MSc ’ Physics and Astronomy’
’Astronomy and Astrophysics’
Master Thesis
’The Swift GRB afterglow legacy survey’
’VLT/X-Shooter spectroscopy of the afterglow of GRB 160203A’
by
’Daan van Rest’
’10187464’
Februari’ ’2017’
’60’ ECTS
Supervisors:
’Prof. dr. L. Kaper’
’Dr. G. Pugliese’
Examiner:
’Prof. dr. L. Kaper’
’prof. dr. R.A.M.J. Wijers’
Contents
1 Introduction 3
1.1 The discovery of GRBs and history of GRB research . . . 3
1.2 The origin of GRBs . . . 5
1.3 Information gained from observing GRBs . . . 7
1.4 Models for the column density evolution of GRB 060418 . . . 11
1.5 UV pumping model for GRB 100901A . . . 16
2 Goal and current status of GRB research 18 2.1 The current sample of GRBs and its characteristics . . . 18
2.2 The VLT/X-Shooter GRB follow up program . . . 25
2.3 Properties of long GRB host galaxies . . . 25
2.4 Properties of short GRB host galaxies . . . 29
3 Observations of GRB 160203A and data reduction 31 3.1 BAT analysis of GRB 160203A . . . 31
3.2 GROND observation and confirmation of GRB afterglow candidate . . . 34
3.3 Skynet observations of GRB 160203A . . . 34
3.4 VLT/X-Shooter observations . . . 35
3.5 Table of detected lines . . . 39
3.6 Equivalent width and its measurement . . . 45
3.7 Error estimation . . . 45
3.8 Spectrum normalization . . . 45
4 Results 46 4.1 Spectral lines and their profiles . . . 46
4.2 The fine structure of Fe II lines . . . 51
4.3 Variability between each frame separately . . . 57
5 Discussion 63
6 Summary 64
Appendices 69
1
Introduction
Gamma Ray Bursts (GRBs) are the most energetic phenomena we know in the distant universe. The events are brief flashes of gamma-rays, from a few ms to more than 1000 s, which can be more luminous than all stars in the visible universe combined. These events are associated with the formation of black holes and magnetars. They have a low occurrence rate of 10−5 to 10−6 per galaxy per year (Podsiadlowski et al., 2004), but because of their high luminosity, they are detectable throughout the visible universe.
GRBs are thought to be the product of collapsing massive stars and compact object mergers. The gamma-rays are produced by a relativistic outflow of matter and radiation in a jet pointing towards us. This outflow can interact with the GRB environment, creating a GRB afterglow. This afterglow emits syn-chrotron radiation, radiation caused by the acceleration/deceleration of charged relativistic particles in a magnetic field, which is visible from X-rays to radio waves. After the GRB afterglow is formed, its spectral energy distribution (SED) shifts in time towards longer wavelengths. Study of these afterglows are crucial in determining the basic properties of GRBs, such as the distance scale and the chemical abundances in their environments.
These afterglows can be very bright when they are formed, almost visible by the naked eye, but their brightness decays in time according to a power law, causing the afterglow to disappear in a couple of days. When observing these events, it is important to observe them in the earliest possible stages. This, combined with the rare event rate of GRBs, made it relatively hard to observe GRB counterparts when they were discovered in the late 1960s. The first optical counterpart was found by van Paradijs et al. (1997), one of the major discoveries at the Anton Pannekoek Institute, an institute affiliated to the University of Amsterdam. Nowadays, with the help of satellites and their coordination with instruments at ground-based telescopes, it is possible to study the afterglows in detail. The Swift satellite for example produces light curves from GRB detections, and X-Shooter does a follow up to obtain an optical/NIR spectrum. These measurements give information about the energetics of the explosion and the interaction with the surrounding interstellar medium (ISM).
These events allow us to observe high-redshift galaxies and intervening systems, especially when observa-tions are done over a broad wavelength band, because the redshift of the GRB is unknown (0 < z ≤ 9). The VLT/X-Shooter is an excellent instrument to do this: it has a large spectral range, from 3000 up to 24800 ˚
A and a spectral resolution of ∼ 30 km/s, depending on the slit width (Vernet et al., 2011). This allows us to observe spectral line features in a relatively large redshift range. For example, it can observe Ly-α (1216 ˚
A) from z = 2.5 up to z = 20.5.
In this report I discuss the results of the observations of GRB 160203A, observations made with VLT/X-Shooter in Rapid Response Mode (RRM). This mode was developed to automate the coordination between a GRB detection and accurate localization by Swift and follow up by VLT/X-Shooter to observe the optical afterglow as soon as possible after the GRB trigger. GRB 160203A is the first event that two epochs were observed with one in the RRM with X-Shooter, just 18 minutes after the burst, and a second observation 5 hours later.
1.1
The discovery of GRBs and history of GRB research
GRBs were first detected by the Vela satellites during the cold war. The United States of America built these satellites to monitor compliance with the Partial Test Ban Treaty by the Soviet Union. This treaty
the Vela satellites formed a network across the (Earth) sky it was possible to use the time of arrival of the signal to roughly localize the origin of the signal. It was concluded that GRBs did not have a terrestrial or solar origin (Klebesadel et al., 1973).
Many models were developed in the following years that tried to explain the origin of GRBs, but ob-servational measurements were lacking. Colgate (1968) proposed for example that the prompt gamma-ray emission could come from shock ejection of the outer layers of supernovae in distant galaxies. Stecker and Frost (1973) suggested that a GRB could come from a superflare from a nearby star. It was difficult to localize a GRB, study GRB counterparts and estimate the distance towards a GRB. Without knowledge about the distance scale of GRBs, it is impossible to asses the energy scale of these events. Therefore models ranged from neutron stars colliding with comets to events happening at cosmological distances (Nemiroff, 1994).
The first important clues to determine the origin of GRBs came from BATSE observations. BATSE stands for Burst And Transient Source Experiment (BATSE). BATSE is an instrument on board NASA’s Compton Gamma Ray Observatory (CGRO), which was launched in 1991. BATSE observed GRBs sys-tematically and made the first GRB sample. From this sample (figure 1) it was concluded that the GRB distribution on the angular plane is isotropic. This means that GRBs are not concentrated in the galactic plane, but have an extragalactic origin, as they are equally likely to happen in every direction. Another result was that GRBs could be catalogued in two groups: short-hard GRBs and long-soft GRBs (Meegan et al., 1992). The first group of GRBs lasting shorter than 2 sec and the second group lasting longer than 2 sec. The hardness of the GRB is calculated by dividing a higher energy channel by a lower one. Short GRBs are associated with higher values for hardness, while long GRBs are relatively soft.
homo-The next step in GRB research was done with the Italian-Dutch satellite named BeppoSAX. This satellite had Wide Field Cameras (WFCs) that allowed observers to localize GRBs with an arcminute accuracy, ac-curately enough for an optical follow up. GRB 970229 was the first burst for which an optical and soft X-ray counterpart was found (van Paradijs et al., 1997; Costa et al., 1997). The Hubble Space Telescope (HST) studied the coordinates of this GRB and found a faint galaxy, which was assumed to be the origin of the GRB. GRB 970508 was the next burst to be localized. This burst showed the first observed radio afterglow (Frail et al., 1997). The optical spectrum of this burst was obtained with the Keck telescope on Hawaii and included metal absorption lines shifted to a redshift of z = 0.8349, which corresponds to a lookback time of 7.084 Gyr. This confirmed that GRBs have a extragalactic origin at cosmological distances (Metzger et al., 1997).
Nowadays the process of detecting a GRB and doing a follow up is automated. We have the Swift satellite which looks for gamma-ray sources over a wide field of view. Swift has 3 instruments, the Burst Alert Telescope (BAT), the X-ray Telescope (XRT) and the Ultraviolet/Optical Telescope (UVOT). When Swift detects a burst with the BAT it tries to localize it with the XRT and UVOT automatically. After localization, if the Rapid Response Mode (RRM) is triggered, the coordinates are sent to the VLT/X-Shooter which then slews to the gamma-ray source. This method allows us to observe GRB afterglows in its earliest stages, for GRB 160203A just 18 minutes after the BAT trigger.
1.2
The origin of GRBs
GRBs are one of the most luminous sources in the universe, reaching isotropic energies up to 1054erg/s. For comparison, all stars in the visible universe, including the sun, radiate up to 1055 erg/s. This makes GRBs easy to detect at large distances, ideally for the study of the high-redshift universe. This study requires knowledge about the basic properties of GRBs.
From detections done by BATSE and Swift, GRBs are classified into two groups, based on their duration. This duration is quantified by T90, which is defined as 90% of the duration of the GRB event. One of the mentioned groups is the short-hard GRBs, the other group consists of long-soft GRBs. Both classes of GRBs can be seen in figure 2, where the short-hard GRBs are represented by the upper left group of points, and the long-soft GRBs are found in the right lower group of points. Levan et al. (2014) propose that there is a third group of ultra long GRBs. All three groups are described consecutively.
Short-hard GRBs are defined to last for less than 2 seconds and have a higher spectral hardness than long GRBs. Some short GRBs are associated with late-type galaxies, such as elliptical galaxies. Late-type galaxies do not contain the gas needed for star formation, which rules out a direct link to young massive stars. Secondly, there is no association with supernovae, which also indicates that massive stars are not the progenitors for this kind of burst (Bloom et al., 2006; Zhang et al., 2009). An alternative model is a merger between two neutron stars or a neutron star and a black hole, resulting in a GRB and a black hole as final product. An overview of GRB formation scenarios can be seen in figure 3.
The second group consists of long GRBs, as they have a longer lifetime than two seconds. Most GRBs are long GRBs, but this could be a selection effect, as short GRBs are more difficult to detect. Currently only Swift and CGRO were able to detect short GRBs. Because long GRBs are more often detected and observed, their afterglows have been studied in greater detail. GRB 160203A, the GRB studied in this report, is a long GRB. Long GRBs have been associated to supernovae and star forming galaxies, meaning that they are probably linked to the death of massive stars (Woosley and Bloom, 2006; Pontzen et al., 2010).
Figure 2: The spectral-hardness (ratio of fluence in 50–100 keV divided by 20–50 keV) versus duration diagram for CGRO/BATSE GRBs (red points) and Swift GRBs (blue points). GRBs are grouped generally into short and hard GRBs and long and soft GRBs. This plot also contains 3 GRBs which had a longer duration than any other GRB measured before (Levan et al., 2014; Kouveliotou et al., 1993).
Figure 3: This is an overview of GRB progenitor scenarios. Short GRBs are probably created by mergers of neutron stars or a black hole and a neutron star, while long GRBs are created via a collapsing massive star. Both scenarios result in the formation of a black hole and a relativistic outflow of radiation and matter. This outflow contains shells that move at a variety of relativistic velocities, which causes these shells to interact. These colliding shells emit low energy gamma-rays. This outflow can also interact with the circumburst medium, resulting in an afterglow that can be seen from X-ray to radio (Gomboc, 2012).
core of about 10 M and is the size of the sun. After nuclear fusion stops in the core, it becomes unstable
and collapses to a black hole. This black hole is surrounded by an accretion disc, existing of material with a too large angular momentum to fall directly into the black hole. Along the rotation axis, matter undergoes almost free-fall. In this region it is also possible to push material through the star surface, if enough energy is injected into this region. This outflow would then cause the GRB. This model does not describe how the jet is launched (Levan et al., 2014).
Levan et al. (2014) state that there could be a third group of ultra long GRBs, with durations of several thousands of second (figure 2). These events show unusual X-ray and optical lightcurves with high amounts of variability, showing low correlation with the behaviour seen in the X-ray. Their host galaxies seem to be faint, compact dwarf galaxies with a high star forming rate. At this moment confirmed ultra long GRBs are GRB 091024A, GRB 101225A, and GRB 111209A. The origin of this new ”group” of GRBs is unknown.
1.3
Information gained from observing GRBs
Due to the high luminosity of GRBs, they can be detected up to very high-redshifts (z = 8 - 9) (Tanvir et al., 2009; Cucchiara et al., 2011). This allows us to probe the abundance patterns of high z galaxies, which gives insight to the nucleosynthetic enrichment history, providing information about the environment that produces GRBs. Moreover, because long GRBs are linked to massive stars, they could be used to probe distant star forming regions. Thirdly, they could be used to probe the intergalactic medium (IGM) up to the epoch of re-ionisation (figure 4).
Figure 4: This schematic visualizes that photons from the GRBs pass through the IGM till they reach Earth. All intervening systems leave their imprint on the spectrum, which makes GRBs probes of the IGM up to the youngest star populations. The age of the universe is labelled on the horizontal axis, from the birth of the first stars up to 13.7 Gyr, current time (Gomboc, 2012). Measuring GRBs and their afterglows thus contains information about GRBs themselves, their surroundings and about every intervening system in the line of sight.
Mg, S, Si, Ar, Ca and Ti, are produced in core collapse type II SNe. These kind of SNe come from massive stars with an initial mass of ∼ 10 M . Iron-peak elements (V, Cr, Mn, Fe, Co, Ni) are mainly made during
supernovae of Type Ia, which originates from stars with low initial masses and longer lifetimes (Hartoog, 2014). By identifying which elements were produced at a specific time in the universe, we can determine what kind of stars existed back. If no α-elements were found in a host galaxy, we know that there were no core collapse type II SNe in that system.
It takes time for these elements to be produced at detectable levels and at certain lookback times we expect to not see specific elements anymore. There should be a point in history that the universe did not contain the iron peak elements, or at least in smaller quantities. By detecting these elements in the host galaxy, using the GRB afterglow, we can constrain the evolution of its ISM.
Because GRBs can be observed at high-redshift, we expect at some point to see primordial gas (no met-als). This could help us to understand when the first stars formed, how they formed and how they looked like. The first stars (Population III stars) are thought to be progenitors of GRBs as well (Bromm and Loeb, 2002, 2006; Hartoog, 2014).
Although it is theoretically expected that in the early universe there should be a larger abundance of α-elements compared to iron peak elements, this is not statistically proven yet: the set of GRB hosts with a determined metal abundance is small, and usually limited due to low spectral resolution and S/N (Hartoog, 2014; Bromm and Loeb, 2002, 2006). Figure 5 shows measured metallicities of GRBs and quasar DLAs. By fitting both samples with a linear fit, Cucchiara et al. (2015) found that GRB metallicity measurements result in higher values at redshift of 1 and higher than quasar metallicities.
GRB afterglow observations can also be used to constrain when the epoch of reionisation started and ended. The epoch of reionisation is the second change phase in the universe since the Big Bang. In this phase baryonic matter condensed into UV-radiating objects, whose radiation ionized the neutral hydrogen in their environments. Gnedin and Ostriker (1997) models predicted that the epoch should start between redshifts 10 to 20, while it ends at z ∼ 6. Observationally, it still needs to be determined when and how this happened (Loeb and Barkana, 2001; Barkana and Loeb, 2001; Hartoog, 2014).
One method to measure the status of the epoch of reionisation is to look at the Lyman α forest of a GRB afterglow. At the beginning of this epoch, just after the epoch of recombination, we expect to find a relatively large abundance of neutral hydrogen. This should affect the strength of the Lyman series, as they are expected to be strong, and because of its redshift, spread over a large wavelength range. This is called the Gunn-Peterson effect.
As the amount of ionised hydrogen increases, and while the redshifts decreases, the Lyman lines should decrease in strength, and its spread. This effect can be observed using a sample of GRB afterglows over a range of redshifts. It is also possible to put constrains on the ionised fraction by fitting the red wing of the DLA (Miralda-Escud´e, 1998).
Figure 5: The graph shows how the metallicity evolves over time measured with GRBs (red) and quasar DLAs (gray). Triangles represent upper limits, while filled/open symbols indicate if these values come from high/low spectral resolving power instruments. Both samples were used to make a linear fit. The red line represents the GRB fit, while the black line comes from fitting the quasar metallicities. The shaded area represents the 1σ error. It seems that the environment of GRBs is more metal rich than the environment of quasars from a redshift of 1 and upwards. (Cucchiara et al., 2015)
1.4
Models for the column density evolution of GRB 060418
Due to the highly energetic nature of GRBs, their hosts show not only resonance metal lines (line transi-tions between the ground state and an excited state), but some also show transitransi-tions from excited states, which are called fine structure lines (see figures 7 and 8). Study of how these lines change over time gives information about how a GRB interacts with its host galaxy. Vreeswijk et al. (2007) modelled variability in absorption lines of GRB 060418, observed with VLT/UVES, to find out whether collisions, excitation by infra-red photons or fluorescence following excitation by ultraviolet photons are the dominant mechanism in GRB afterglows. They found that the UV pumping model was the best fit to the column densities and concluded that the effects of collisional and infrared excitation were negligible. Figure 6 shows the UV model fit. UV pumping models are therefore used to model variability of spectral lines in GRB afterglows.
The ISM is a low density environment, where the gas particles undergo a negligible amount of collisions. This gas is therefore in its ground state and if it gets excited by a photon, it de-excites back to its ground state, which we see as a resonance transition line. If a GRB ionizes and excites the ISM by UV pumping, which is far more energetic than the usual ISM circumstances, we may observe resonance and fine structure lines. Fine structure lines are lines in atomic spectra caused by the spin orbit coupling in an atom and can be seen as the transitions between different excited states. These excited states are less populated because of their oscillator strengths are lower than those of resonance lines. Therefore, if fine structure lines are seen in a spectrum, this is an indication that the matter is in an energetic environment, for example close to a GRB. Figures 7 and 8 show energy level diagrams of Fe II and Ni II, respectively.
The UV pumping model assumes that there is a source of UV radiation and a cloud with thickness l at a distance d from the source, filled with atoms with a column density of N . When a massive star collapses and a GRB is formed, its relativistic outflow can excite and ionize the circumburst medium, bring the atoms into an excited or ionized state. As the afterglow fades in time, excited levels start to depopulate. The depopulation is determined by the oscillator strengths of the transitions. The UV pumping model returns column densities for a set of parameters per timestep. If these column densities fit the observed values for a given set of parameters, then these parameters could describe the system we observe. Typical distances for absorber systems found by this model are 100 pc up to 5000 pc (Hartoog et al., 2013). The UV pumping model allows us to gain knowledge about the structure of high redshift galaxies at distances of ∼ 10 Gly. Methods to learn about the structure of optical unresolved systems at large distances are of great importance if we want to learn about the distant universe. A list of modelled GRBs can be found in table 1.
The model also requires extinction (AV) as input, because dust absorbs UV photons. Extinction lowers
therefore the number of photons we observe, which in turn influences the lightcurve and spectrum we obtain. This affects the measured column densities and thus the best fit model. It is therefore necessary to account for the extinction in GRB afterglow modelling.
In the case of GRB 060418 the best fit parameters were: log N (Fe II) = 14.75 ± 0.06, log N (Ni II) = 13.84 ± 0.02, d = 1.7 ± 0.2 kpc and a Doppler broadening factor of 25 ± 3 km/s. This implies that no significant amount of Fe II and Ni II is closer to the origin of the GRB than ∼ 1.7 kpc. This also applies for any H I regions, because neutral H ionizes more easily than Fe or Ni. The model also constrains the magnitude of the possible UV flash, because UV radiation correlates with the modelled column densities (Vreeswijk et al., 2007).
Do note that the models do not predict that the velocity of the outflow is faster than the speed of light. The model assumes that the GRB causes an afterglow. The optical lightcurve of this afterglow is then used to calculate the afterglow flux (formula 1). The afterglow flux is distributed over an expanding sphere,
Figure 6: The top panel shows the observed total column densities for Fe II lines and the fitted UV pumping model. Fe II fine structure lines are represented by open circles: 6D 7/2 ,6D 5/2 , 6D 3/2 , and 6D 1/2
respectively. The best fit for the ground state of Fe II is shown by the dotted line. The filled triangles
Figure 7: This is an energy level diagram for selected levels of Fe II. Not all transition lines are shown for clarity. Arrows indicate the most likely transitions between levels. The figure gives for each transition the
Figure 8: This is an energy level diagram for selected levels of Ni II. Not all transition lines are shown for clarity. Arrows indicate the most likely transitions between levels. The figure gives for each transition the wavelength and Einstein coefficient, Aul in s−1. Electric dipole allowed transitions are indicated with a solid
line. Forbidden transitions (magnetic dipole or electric quadrupole) are shown with a dotted line (Vreeswijk et al., 2007).
Table 1: This table is a list of GRBs for which the distance to an absorber has been derived from the variability in the afterglow (Hartoog et al., 2013). The columns are respectively the Burst ID, number of epochs, total observing time between the epochs, telescope and spectrograph, ion which for which variability
1.5
UV pumping model for GRB 100901A
Hartoog et al. (2013) used the UV pumping model to fit the variability caused by GRB 100901A, which can be seen in figure 9. The other models were not used because Vreeswijk et al. (2007) concluded that UV pumping is the dominant excitation mechanism. The top plot shows the R band lightcurve. It is suggested that the column densities per timestep correlate with the R band magnitudes over time (formula 1). In figure 9 Fe II* follows the R band magnitude, which is UV in the rest frame. The Ni II column density reacts later than Fe II* to the GRB and decays at a slower rate.
Fνrest(t) =F0· 10 (mR(t)−AR,gal)/−2.5 1 + zh [λrest(1 + zh) 6415˚A ] βν[DL d ] 2 (1)
Formula 1 (Hartoog, 2014) describes the flux in the host galaxy frame at the GRB-facing side of the absorbing cloud (Hartoog, 2014). F0 = 3.02 · 10−20 erg/s/cm2/Hz is the flux of Vega at the wavelength
of 6410 ˚A (Fukugita et al., 1995), AR,gal = 0.264 is the Galactic extinction (Schlegel et al., 1998), λrest
is the wavelength array of transitions for which the flux is calculated, βν = 0.82 is the spectral slope and
DL = 9.968 · 109 pc is the luminosity distance. The only variable in this formula that changes over time
is mR(t) (see figure 9). The calculated flux is used in the model to calculate column densities at each timestep.
The fit results of Hartoog et al. (2013) are described in table 2. The model without extinction fits a small cloud at a distance of 250 ± 75 pc, but when extinction is added the model favours large clouds (> 1 kpc) at a small distance (∼ 100 pc). It was impossible to discriminate between the the close-large and distant-small cloud case.
Table 2: This table contains the parameters used by Hartoog et al. (2013) to model GRB 100901A (figure 9). The input values were fixed or free. Free values change to the best fit after the model is run. The model is ran multiple times with different input values, until a good fit is found. The columns respectively describe the extinction, distance to the absorber, thickness of the absorber, distance to the absorber after the fit, thickness of the absorber after the fit, column density of Fe II, column density of Ni II and the chi-square of the fit. Column densities are on log scale in cm−2.
From figure 9 we can see that the model predicts that the fine structure levels depopulate in the order of days, but that change is visible in the order of ∼ 0.010 days. If such a change can be seen in the spectra of GRB 160203A, we could simulate the host galaxy with the UV pumping model. The UV flux calculated by formula 1 excites the Fe II lines, which depopulate over time. This translates to a weakening of absorption
Figure 9: The top panel shows the R band magnitudes of GRB 100901A. The middle and bottom panel the column density as a function of time for respectively the Fe II fine-structure and metastable levels and for a
2
Goal and current status of GRB research
This thesis is about one burst, GRB 160203A, and the information we can extract from its afterglow spec-trum. This observation is part of a bigger project. While Swift already detected 1237 bursts in total, the sample for which an optical spectrum is obtained of the afterglow is small. We know that long GRBs are associated with supernovae and thus the collapse of massive stars and that short GRBs could be related to compact mergers (Woosley and Bloom, 2006; Pontzen et al., 2010; Bloom et al., 2006). However, through GRB afterglows we can probe the environment in which GRBs explode. The metallicity, ionization and velocity structure give important information about the formation and evolution of galaxies.
Currently the dataset of GRBs consists of 1237 detections, of which 113 were observed with the VLT/X-Shooter spectrograph. That means that we have information on the redshifts (distances) and host galaxies for a significant fraction of the GRBs.
This dataset could provide an important view on what kind of light curves belong to what kind of GRB. GRBs at this moment are only classified by their T90 and hardness. This could be improved, because the number of peaks and their shape also describe physics happening in a GRB. These properties should therefore be included as well in the overall picture of GRBs. GRB light curves sometimes contain one peak, sometimes multiple, and nobody really understands what drives this behaviour.
Optical/NIR spectroscopy, although always some time after the GRB trigger due to localization and positioning time, also gives vital information about GRBs. While we cannot observe the GRB event at the same moment Swift detects a burst, we can probe its afterglow. This afterglow can be used to study the GRB environment as the X-ray and UV excites the surrounding material. At this moment, the GRB X-Shooter collaboration observes GRBs at a higher rate than that they are processed. This thesis includes one of those studies.
A table containing all X-Shooted GRBs can be found in Appendix A, in table 7. This Appendix also includes a list of papers in progress and published papers on the X-Shooter sample. This list is in table 8.
2.1
The current sample of GRBs and its characteristics
This subsection contains three plots describing the current status (6-2-2017) of the X-Shooter sample. Cur-rently, X-Shooter observed 113 GRB afterglows of 1237 Swift GRBs. Figure 10 shows the redshift distribution of both samples. Both distributions follow the same trend, with a peak at z = 1 and a tail towards higher redshifts. The increase in the number of GRBs till a redshift of 1 is caused because more GRBs can be detected in a larger volume of space.
The high-z tail in the redshift distribution is at least partly caused by bias. First, instruments are limited by their sensitivity. Secondly, redshifts are determined by the displacement of spectral lines compared to their rest wavelengths. If the displacement is too large, usual used spectral lines disappear from the observ-able window of the instrument, and thus complicate the calculation of a redshift. Another effect is the dust extinction of the host galaxies, absorbing photons from a GRB afterglow to reach the observer. If we do not observe an optical afterglow, the GRB is called a dark burst (Coward et al., 2013).
Figures 13 and 14 show the T90 and T90-hardness distribution of both the X-Shooter and Swift sample respectively. Both samples contain more long GRBs than short GRBs.
other means.
More GRB detections could provide more reliable statistics about general GRB properties, such as what the ratio of short and long GRBs is and how their afterglows differ in terms of metallicity, chemical abun-dances, redshift and hardness. Because short GRBs are proposed to originate from mergers of neutron stars and black holes, both the final stages of stars, we would expect to see a higher abundance of metals in their afterglow spectra compared to long GRBs. This can be validated if more short GRB spectra are procured, as the current X-Shooter sample contains only 7 short GRBs. All short GRBs have a weak or negligable afterglow, except for GRB 130603B (de Ugarte Postigo et al., 2014). GRB 160410A also has a good S/N and a paper is in preparation by Selsing et al.
Figure 11 shows the redshift distribution of Quasi-Stellar Objects (QSOs). As the GRBs in figure 10, the number first increases because of a larger volume of observable space up to a redshift of 1. At larger redshifts than 1, the QSO distribution shows a tail, similar to the redshift distribution of GRBs. Do note that the y-axis of figure 11 is in log scale. When comparing figures 11 and 10 we can conclude that GRBs and supermassive black holes are distributed similarly across redshift space.
Madau and Dickinson (2014) studied the history of star formation. They made figure 12 which shows how the star formation (SF) rate changes with redshift. This figure included SF rate measurements from multiple samples (table 1 in Madau and Dickinson (2014)). From this graph we can see that the SF rate peaks at a redshift of z = 2. When we compare figure 12 with figure 10, we see that for z < 2 the number of GRBs stays constant, while the SFR decreases at lower redshifts. For z > 2 both distributions decrease, which is expected if long GRBs do correlate with the SF rate (Hartoog, 2014; Woosley and Bloom, 2006; Pontzen et al., 2010). Note that the X-Shooter sample consists almost entirely of long GRB afterglows, as the sample only contains 7 short GRBs. It is therefore surprising that no correlation is found for z < 2.
Figure 10: The graph includes two distributions. The red distribution consists of all Swift GRBs with a determined redshift (N = 315). The blue distribution consists of the X-Shooted GRBs with a determined redshift (N = 93). Both distributions are similarly populated.
Figure 11: This graph shows the redshift distribution of QSOs. The total number of QSOs in the sample is 105783. The bin size is 0.1. This sample represents the fifth edition of the Sloan Digital Sky Survey (SDSS) Quasar Catalog (Schneider et al., 2010). The distribution is similar to figure 10.
Figure 12: Madau and Dickinson (2014) looked at how the SF rate depends on the redshift and produced these plots. The top right plot shows the SF rate calculated from UV measurements, the bottom right plot shows the SF rates calculated from IR measurements. The left plot includes both samples. Each color represents a different sample of galaxies, which are explained in table 1 in Madau and Dickinson (2014). Note that the x-axis (redshift) is in log scale.
Figure 13: This is a comparison between the T90 distributions of all X-Shooted GRBs and Swift GRBs. These distributions contain all GRBs with a known T90 and hardness ratio. It is clear that there are two groups of GRBs in both samples, short and long GRBs. Short GRBs are underrepresented in the X-Shooter sample.
Figure 14: This graph shows the X-Shooted GRBs and Swift GRBs in the hardness-T90 plane. The left group of points represents the short GRBs, the right group of points represents the long GRBs. The black dot is GRB 160203A and is one of the softest GRBs in the sample. Only two other burst are softer. These distributions contain all GRBs with a known T90 and hardness ratio.
2.2
The VLT/X-Shooter GRB follow up program
GRB afterglow observations characterize the properties of the host galaxy, such as: chemical abundances, star forming regions and other observables at large distances. This is only possible if there are satellites that are able to detect and locate GRBs rapidly enough for a follow up. In order to obtain the spectra from GRB afterglows there is the VLT/X-Shooter GRB follow up program. Prof. dr. Johan P.U. Fynbo is the principal investigator of the GRB research collaboration.
In this program Swift/BAT is used to detect GRBs, after which Swift/XRT and Swift/UVOT try to localize the burst. Swift is active since November 20, 2004, and was originally planned to operate for 2 years, but its mission time got extended till present. The GRB coordinates are then used by VLT/X-Shooter to obtain a spectrum. X-Shooter has a wavelength range of 3000 up to 25000 ˚A. This allows us to study GRB afterglows over a wide range of redshifts. For example, X-Shooter should be able to detect Ly-α (1216 ˚A) from z = 2.5 up to z = 20.5.
As mentioned before, although the sample already consists of 113 GRB afterglows, it is still a poor statis-tic. Further continuation of this program is needed to obtain statistically significant results. In order to study the chemical evolution in the universe, more than 25 metallicity measurements are needed. It should be noted that this program has the potential to discover unexpected phenomena, such as the extinction of type Ia SN (Fynbo et al., 2014).
2.3
Properties of long GRB host galaxies
Lyman et al. (2017) studied a sample of 39 long GRBs locations at a redshift of z < 3. Those locations were observed with the Hubble Space Telescope (HST) WFC3/F160W SNAPSHOT, an UVIS/IR imager, which resulted in 35 observed host galaxies. No hosts were detected for 4 bursts. For 31 of the 35 hosts it was possible to localize the GRB explosion site in the galaxy.
They compared their sample with a star forming galaxy sample (Santini et al., 2009) and found that long GRB host galaxies are significantly fainter than typical star forming galaxies in this redshift range (figure 15)). Because long GRB progenitors are massive stars, it was expected that their hosts would be compara-ble to star forming galaxies. However, this indicates that long GRBs are not unbiased star forming region (SFR) tracers for redshifts of z < 3. Long GRBs do explode in the brightest regions of their hosts (figure 16), which are the regions where most star forming happens. Half of the bursts explode in the brightest pixel. We also know that long GRB hosts are mainly spiral and irregular galaxies (Lyman et al., 2017; Hartoog, 2014). From figure 17 can be seen that most long GRBs are in spiral galaxies. This figure also shows that galaxies are more concentrated at lower redshifts.
Figure 15: This plot includes the long GRB sample Lyman et al. (2017) have studied and a sample of SFRs from the GOODS-MUSIC survey (Santini et al., 2009). The cyan points are long GRB hosts; the purple points are GOODS-MUSIC SFR galaxies. The black dotted line represents m = 24 mag. The dashed line is the average of the GOODS-MUSIC sample. From this plot we can see that long GRB hosts are less bright than the average SFR galaxies.
Figure 16: This graph shows UVIS/IR HST images of GRB hosts. + is the centre of the image, X the brightest pixel and the star is the GRB location. The color represents the brightness (Flight). Do note that Flight is normalized. From these images can be seen that long GRBs explode in the brightest regions of their host galaxies (Lyman et al., 2017).
Figure 17: This graph represents the morphology of the host galaxies studied by Lyman et al. (2017). The galaxies are parametrised by their concentration and asymmetry (Conselice et al., 2005). The colors represent the redshifts of the hosts. Most long GRB host galaxies are spiral galaxies if this parametrization is used.
2.4
Properties of short GRB host galaxies
We know that short GRBs form in early-type elliptical and late-type spiral galaxies (Hartoog, 2014; Berger, 2009, 2014). Long GRBs form only in late-type galaxies where there is enough gas to create massive stars. Short GRBs can form in regions filled with gas, and in regions without gas, because they are proposed to come from compact binary mergers. Galaxies lose gas through time, due to the conversion of gas into stars and black holes, or that gas is flung away when galaxies merge. Early-type galaxies, which are possible hosts for short GRBs, are therefore expected to have less gas and a low star forming regions. Short GRB hosts are thus assumed to have low star formation rates. This can be seen in figure 19 (Berger, 2014). Because short GRB hosts are on average older, it is expected that their average metallicity is higher than long GRB hosts. This is shown in figure 18 (Berger, 2014).
Figure 18: The graph compares the metallicity with the magnitude of galaxies (Berger, 2009). circles are long GRB hosts, squares are short GRB hosts, pentagrams are star forming galaxies studied by Kobulnicky and Kewley (2004). The grey bars present the 14-86 percentile range for galaxies at redshift z < 0.1 from the Sloan Digital Sky Survey (Tremonti et al., 2004). Short GRBs have higher metallicities on average than Long GRBs.
Figure 19: In this figure (Berger, 2014) is the star forming rate plotted against the brightness of long GRB hosts (circles), short GRB hosts (squares and triangles) and star forming galaxies (pentagrams) from the sample of Kobulnicky and Kewley (2004). The grey line is a linear fit for the long GRB hosts. The grey area is represents the 1 σ deviation. The top-left plot is the cumulative distribution of all three samples. Berger (2014) finds that short GRBs have a significantly lower star forming rate. Short GRBs do trace the sample of Kobulnicky and Kewley (2004).
3
Observations of GRB 160203A and data reduction
In order to get an as high S/N as possible, it is needed to observe a GRB afterglow in its most early stages. Figure 22 shows the optical lightcurve of GRB 160203A and the need to get early observations as the bright-ness decays exponentially by approximately ∼ t−2.
In order to be able to detect GRBs with a spectrograph, it first needs to be localized. This is done by the Swift satellite, which observes the sky over a broad field of view using the ’Burst Alert Telescope’ (BAT). After detection it slews the X-ray Telescope (XRT) and the Ultraviolet/Optical Telescope (UVOT) towards the estimated BAT position for more accurate localization, in the case of GRB 160203A, from 3 arcmin to 2.3 arcseconds accuracy.
Only after this report, follow up telescopes slew towards the coordinates to do further investigation of the source. Skynet, a network of smaller and automated telescopes, produced the optical lightcurve of figure 22. The VLT/X-Shooter was used to do spectral analysis, just 18 minutes after trigger, on the GRB afterglow, which is the product of the GRB event detected by the Swift/BAT.
The observation of a GRB event thus requires multiple observations being done. Reports about these observations are saved in the Gamma-ray Circulars Network (GCN) system. The GCNs about GRB 160203A that led to the VLT/X-Shooter observations in the rapid respond mode are described in this section.
3.1
BAT analysis of GRB 160203A
At 02:13:10 UT, The Swift Alert Telescope (BAT) detected and located GRB 160203A at RA(J2000) = 10h 47m 48s and Dec(J2000) = -24d 45’ 43” with an uncertainty of 3 arcmin. The peak count rate was ∼ 500 counts/s (P. D’Avanzo et al. GCN 18979). The image of the detection can be seen in figure 20.
137.3 seconds after BAT detection, the XRT slewed towards the source at 02:15:27.8 UT and found a fading X-ray source at a RA(J2000) = 10h 47m 48.27 and Dec(J2000) = -24d 47’ 19.2”. The uncertainty of this detection is 2.3 arcseconds, 96 arcseconds from the BAT position, but within the BAT error. The burst was not detected by the UVOT.
Figure 21 shows the BAT lightcurve. The GRB shows a clear pulse in the energy bands from 15 up to 100 keV. The highest energy band of 100-350 keV does not receive a signal, meaning that GRB 160203A is a soft burst. The profile of the burst seems to be single peaked: while the profile does seem to show sharp fluctuations in the right part of the peak, this could also be due to noise. The peak lasts longer than 2 s, which means it is a long GRB. The T90 of the burst is 20.2 ± 3.2 sec.
Figure 21: The BAT lightcurve of GRB 160203A. The upper 4 figures show the BAT lightcurve in different energy channels. The T90 of the burst is 20.2 ± 3.2 sec. The bottom figure is the stacked lightcurve, covering all energy bands (GCN 18998, Barthelmy et al. 2016).
3.2
GROND observation and confirmation of GRB afterglow candidate
The Gamma-Ray Burst Optical/Near-Infrared Detector (GROND) on the ESO 2.2 m telescope at La Silla in Chile started its observations at at 02:18:58 UT, 6 min after the BAT trigger, to see if the source is a GRB afterglow candidate. It found a source at RA(J2000) = 10:47:48.35 and Dec(J2000) = -24:47:19.8 (GCN 18980, Kruehler et al. 2016). These coordinates have both an uncertainty of 0”.5. The observed magnitudes per band are in table 3.
Band Magnitude Error
g’ 18.9 0.1 r’ 18.0 0.1 i’ 17.7 0.1 z’ 17.6 0.1 J 17.0 0.1 H 16.7 0.1
Table 3: Magnitudes measured with GROND
3.3
Skynet observations of GRB 160203A
After the confirmation of the GRB afterglow, A. Trotter et al. 2016 (GCN 18987) did a follow up using Skynet telescopes. Skynet is a network of automated telescopes. This network numbers 19 optical telescopes between 14 and 40 inches in diameter. The telescopes are located in Chile, Australia, the United States and Italy. Skynet observations were used to create a optical lightcurve, which can be seen in figure 22. The curve seems to show a bump near a Log(days) of ∼ -1.9. This bump could be the supernova usually associated with a long GRB.
Figure 22: This is the optical lightcurve made with Skynet, a network of automated telescopes. The lightcurve of GRB 160203A is measured with 3 different telescopes. The decay goes as ∼ t−2. The telescopes observed in the I, R, V and B bands. The curve seems to show a bulge near a Log(days) of -1.9, which could be a supernova associated to the burst (GCN 18987, A. Trotter et al. 2016).
3.4
VLT/X-Shooter observations
Optical and near-infrared spectra of GRB 160203A were obtained in the rapid response mode with the VLT/X-Shooter (GCN 18982, Pugliese et al. 2016). The layout of X-Shooter is described in figure 24. The unique feature of X-Shooter is that that it can acquire spectra in the UVB, VIS and NIR band simultane-ously. This covered a wavelength range from 3000 ˚A up to 24800 ˚A (Vernet et al., 2011).
The observation started at 02:31:35 UT on 03-02-2016, just 18 minutes after the burst. This is the first time that an afterglow spectrum is acquired with X-Shooter just minutes after the burst, giving vital infor-mation about the environment in the earliest stages of the GRB environment.
Two sets of observations were made, one in the rapid response mode, and one set 5 hours after the burst. This gives the opportunity to study the change in strength in spectral lines and thus measure the effect of the GRB on its host galaxy. The table of observations is in table 4. One of the acquisition images can be seen in figure 23. Detected lines can be found in table 5.
The S/N was found in table 4 was estimated by the astropy package found in python. the function sigma clipped stats was used to calculate the mean, median and standard deviation of a 2D spectrum. This function masks all pixels which deviates with more than 3 σ. The background can be estimated by the median and the noise by the standard deviation. The signal was calculated by averaging all rejected points up to 50 sigma to exclude cosmic rays. Signal was then divided by noise. This gives a rough estimate how good an image is. The sides of the frames were neglected in the S/N calculation. Sides were defined to be all pixels removed less than 100 pixels from the edges of a frame.
Figure 23: This is an acquisition image of GRB 160203A. The GRB is located in the middle bright spot. The size of the field is 1.5’x1.5’.
Figure 24: This figure describes the X-Shooter instrument setup (Vernet et al., 2011). X-Shooter consists of a central structure that has three spectrographs optimized for the UVB, VIS and NIR wavelength ranges. After the focus two dichroics reflect or transmit the UVB and VIS light to their arms, while the longer wavelengths continue on their straight path to the NIR spectrograph. This setup allows the instrument to have a large observable window between 3000 ˚A and 24800 ˚A.
Arm yr-m-d hr:min:s Exposure Time (s) Average Airmass Average Seeing S/N UVB 2016-02-03 02:31:39.908 180.0 1.753 0.845 4.52 VIS 2016-02-03 02:31:45.038 180.0 1.755 0.89 10.4 NIR 2016-02-03 02:31:47.8353 180.0 1.7545 0.89 33.3 UVB 2016-02-03 02:36:11.772 300.0 1.7015 0.98 4.91 VIS 2016-02-03 02:36:16.932 300.0 1.703 0.98 10.79 NIR 2016-02-03 02:36:20.2817 299.9 1.703 0.98 33.81 UVB 2016-02-03 02:42:43.526 600.0 1.626 1.11 6.11 VIS 2016-02-03 02:42:48.677 600.0 1.6275 1.03 13.89 NIR 2016-02-03 02:42:52.0232 600.0 1.6265 1.11 34.31 UVB 2016-02-03 02:54:15.578 1200.0 1.509 1.075 6.4 VIS 2016-02-03 02:54:20.738 1200.0 1.5095 1.075 17.04 NIR 2016-02-03 02:54:23.9118 600.0 1.54 1.03 34.43 NIR 2016-02-03 03:04:32.6691 600.0 1.4735 1.215 34.77 UVB 2016-02-03 03:15:48.362 1920.0 1.3575 0.875 6.65 VIS 2016-02-03 03:15:53.523 1920.0 1.358 0.875 18.53 NIR 2016-02-03 03:15:56.8566 480.0 1.4125 1.01 35.09 NIR 2016-02-03 03:24:04.3450 480.0 1.37 0.885 34.64 NIR 2016-02-03 03:32:12.5015 480.0 1.3305 0.825 34.52 NIR 2016-02-03 03:40:20.6566 480.0 1.2955 0.825 34.61
Arm yr-m-d hr:min:s Exposure Time (s) Average Airmass Average Seeing S/N
UVB 2016-02-03 07:42:23.129 600.0 1.039 1.065 4.22 VIS 2016-02-03 07:42:28.279 600.0 1.0385 1.04 11.67 NIR 2016-02-03 07:42:31.4133 600.0 1.039 1.04 37.0 UVB 2016-02-03 07:54:00.130 600.0 1.0535 1.02 4.16 VIS 2016-02-03 07:54:05.301 600.0 1.053 1.02 12.03 NIR 2016-02-03 07:54:08.6223 600.0 1.053 1.02 36.89 UVB 2016-02-03 08:05:37.483 600.0 1.071 1.135 4.14 VIS 2016-02-03 08:05:42.643 600.0 1.071 1.09 12.05 NIR 2016-02-03 08:05:45.8322 600.0 1.071 1.09 36.7 UVB 2016-02-03 08:17:15.674 600.0 1.0915 1.17 4.16 VIS 2016-02-03 08:17:20.855 600.0 1.0915 1.17 12.97 NIR 2016-02-03 08:17:23.7060 600.0 1.0915 1.17 36.64
Table 4: These are the details of the observations done with the VLT/X-Shooter. The first column shows with which arm an observation was done: UVB, VIS and NIR. The following three columns give the date, start time and exposure time of an observation. The last two columns the average of the beginning and final airmass and seeing. The S/N of the NIR spectra seems to be odd and is probably caused by telluric lines in the S/N calculation.
3.5
Table of detected lines
At the redshift of 3.517 we detect a large number of metal absorption lines. The ions of which resonance lines are observed are S I, S II, Si II, Si IV, O I, O II, Ni II, Al II, Al III, Zn I, Fe I, Fe II, Fe III, Mn I, Mn II, Mg II, Cr II and weak signatures of CO and Ti II. The spectrum also shows intervening systems with z = 1.267, 2.203 and 2.831. This thesis focuses on the lines found in the host system. Identified lines can be found in table 5. Do note that the EWs found in the table are not in the rest frame. To calculate the EW in rest those columns need to be multiplied by a constant factor based on the redshift: EWrest = 1+zEWh = EW4.517.
The rebinned spectrum is plotted in figures 25, 26 and 27.
The IRAF package noao/splot was used in combination with a self written python program to calculate the redshift. IRAF measured the center of the lines using Voigt profiles found in the function splot. Python code read table 5, averaged the redshifts and produced RV plots of each line. This calculation gives an average redshift of 3.517. This redshift was then used to calculate the wavelengths where we would expect to find the center of the spectral lines. This is mandatory to be able to compare RV profiles of lines with each other.
linename λ0(˚A) λz(˚A) redshift EW epoch 1 error EW epoch 2 error
SII 1250.58 5648.88 3.517 0.1254 0.1004 0.2028 0.1792 SII 1253.52 5663.31 3.5179 0.1513 0.1072 0.2095 0.1864 SII 1259.52 5689.62 3.5173 0.2109 0.0824 0.2804 0.1513 SiII 1260.42 5693.83 3.5174 0.4941 0.1287 0.4396 0.1944 SiII* 1264.74 5712.73 3.5169 0.2543 0.1258 0.2443 0.195 OI 1302.17 5881.62 3.5168 0.4465 1.0838 0.3694 2.2361 SiII 1304.37 5892.07 3.5172 0.302 0.1037 0.3553 0.2136 OI* 1304.86 5894.5 3.5173 0.1596 0.0474 0.0859 0.0792 NiII 1317.22 5949.79 3.5169 0.0411 - - -CII 1334.53 6027.87 3.5168 0.4469 0.2897 0.3935 0.0998 CII* 1335.71 6033.18 3.5168 0.2995 0.3312 0.3298 0.1126 NiII 1370.13 6188.97 3.5171 0.0035 - - -SiIV 1393.76 6294.57 3.5163 0.1374 0.0489 0.1427 0.0818 SiIV 1402.77 6336.57 3.5172 0.5684 0.0872 0.626 0.1611 GaII 1414.4 6387.92 3.5163 0.0137 - - -CO 1419.0 6409.49 3.5169 0.0141 - - -NiII 1454.84 6571.89 3.5173 0.0526 - - -ZnI 1457.57 6582.66 3.5162 0.0104 - - -CO 1477.5 6673.73 3.5169 0.0019 - - -SiII 1526.71 6895.77 3.5168 0.4906 - 0.4564 -SiII* 1533.43 6925.47 3.5163 0.071 - - -CoII 1539.47 6965.9 3.5249 0.0317 - - -CIV 1548.2 6992.25 3.5164 0.1677 0.0442 0.1468 0.0778 CIV 1550.77 7003.84 3.5164 0.1457 0.0496 0.1607 0.0815 FeII* 1570.25 7088.8 3.5144 0.0115 - - -FeII 1608.4509 7265.39 3.517 0.3296 0.0575 0.3500 0.143 FeII 1611.2004 7276.92 3.5165 0.0728 0.0878 0.0821 0.118 FeII* 1618.47 7309.61 3.5164 0.0122 - 0.0422 -FeII* 1631.13 7369.45 3.518 0.014 - 0.0122 -FeII* 1634.35 7381.32 3.5164 0.0075 - -
-linename λ0(˚A) λz(˚A) redshift EW epoch 1 error EW epoch 2 error NiII 1703.41 7693.97 3.5168 0.0175 - - -NiII 1709.6 7722.89 3.5174 0.0502 - - -NiII 1741.55 7866.17 3.5168 0.0146 - - -NiII 1751.92 7913.7 3.5172 0.0413 - - -SI 1807.31 8162.19 3.5162 0.0392 - - -SiII 1808.01 8166.87 3.517 0.2654 0.0136 0.2823 0.0472 SiII* 1817.45 8211.45 3.5181 0.0235 - - -NiII 1842.89 8323.57 3.5166 0.0156 - - -AlIII 1854.72 8376.87 3.5165 0.1219 0.0290 0.0487 0.1093 AlIII 1862.79 8412.94 3.5163 0.0918 0.0839 0.0467 0.0781 FeII 1875.16 8465.4 3.5145 0.0064 - - -FeI 1883.78 8508.02 3.5165 0.0109 - - -SiII 1892.03 8558.98 3.5237 0.1005 - - -TiII 1906.24 8610.11 3.5168 0.008 - - -CoII 1941.29 8768.23 3.5167 0.0237 - - -CrII 2056.26 9288.73 3.5173 0.1566 0.0499 0.0772 0.0812 CrII 2062.23 9316.24 3.5176 0.2452 0.0658 0.1935 0.0940 ZnII 2062.66 9319.07 3.518 0.1377 - - -CrII 2066.16 9333.78 3.5175 0.3588 0.0148 0.1167 0.0246 FeIII 2107.99 9522.21 3.5172 0.1526 0.0348 0.1434 0.1071 NiII* 2166.23 9782.54 3.5159 0.0509 - - -FeI 2167.45 9791.22 3.5174 0.033 - - -NiII* 2175.22 9825.97 3.5172 0.0506 - - -MnI 2185.59 9872.22 3.517 0.0987 - - -NiII 2221.72 10013.0 3.5069 0.2836 - - -NiII 2223.0 10070.7 3.5302 0.0174 - - -FeII 2260.7793 10210.5 3.5164 0.212 0.1750 0.4714 0.786 FeII* 2328.11 10515.65 3.5168 0.0626 0.0300 - -FeII* 2328.11 10516.5 3.5172 0.0894 - - -FeII* 2333.5147 10539.1 3.5164 0.098 0.0178 - -FeII 2344.2129 10589.1 3.5171 0.5732 0.0508 0.4091 0.643 FeII* 2345.0 10596.6 3.5188 0.207 - - -FeII* 2349.0215 10611.93 3.5176 0.677 0.0320 - -FeII* 2359.83 10659.5 3.5171 0.01 - - -FeII* 2365.5508 10688.8 3.5185 0.0731 0.0261 - -FeII 2374.4604 10725.5 3.517 0.5156 0.1472 0.2075 0.750 FeII* 2381.4877 10755.08 3.5161 0.0344 0.0198 - -FeII 2382.7641 10762.84 3.517 0.6396 0.0474 0.1508 0.234 FeII* 2382.77 10764.0 3.5174 0.6687 - - -FeII* 2383.7873 10768.83 3.5175 0.0204 0.0124 - -FeII* 2396.15 10823.1 3.5169 0.09 - - -FeII* 2396.3551 10825.2 3.5174 0.3903 0.0363 - -FeII* 2399.9718 10842.16 3.5176 0.1824 0.0978 - -FeII* 2405.16 10862.7 3.5164 0.3379 - - -FeII* 2405.6173 10869.15 3.5182 0.2542 0.0281 - -FeII 2586.6495 11683.94 3.517 0.5864 0.0383 0.6919 0.530 FeII* 2599.1457 11737.1 3.5158 0.124 0.0378 - -FeII 2600.1722 11744.7 3.5169 0.7425 0.0417 0.1505 0.964
linename λ0(˚A) λz(˚A) redshift EW epoch 1 error EW epoch 2 error FeII* 2607.8658 11780.9 3.5174 0.0513 0.0439 - -FeII* 2612.6536 11802.4 3.5174 0.0826 0.0223 - -FeII* 2618.3984 11828.71 3.5175 0.0225 0.0129 - -FeII* 2621.19 11838.5 3.5165 0.0804 - - -FeII* 2629.08 11876.1 3.5172 0.0417 - - -FeII* 2632.1077 11897.63 3.5202 0.023 0.0161 - -MgII 2796.35 12631.0 3.517 0.1792 - - -MgII 2803.53 12664.0 3.5172 0.451 - -
-Table 5: -Table of identified elements. Features that were studied in detail have calculated errors and EW measurements in the second epoch if the feature was still visible. Features that do not exist in the second epoch are marked with a -. Fine structure lines have an * next to the name of the line.
3.6
Equivalent width and its measurement
Because GRBs are transient in nature, we expect that their features vary over time, getting fainter as time progresses. Measuring the strength and shape of a spectral line, quantified by the equivalent width (EW), is a measurement of the kinematics of the host galaxy. The equivalent width is defined in formula 2.
In formula 2 I is the measured intensity at wavelength i, while C is the continuum at wavelength i. The ratio of C(i)I(i) is a normalized intensity. If this value is subtracted from 1 and integrate that over the wave-length (λ-range) range of a absorption line, we get a normalized area. If EW = 0 there is only continuum within the wavelength range.
EW = 1 itot n X i=1 1 − I(i) C(i) (2)
EW measurements were done with the IRAF package splot and its deblend function. This interactive function needs a λ-range and an indication of where absorption features are to operate. It calculates the con-tinuum using a linear fit in the defined λ-range. Deblend also needs input about where absorption features are in the spectrum, as these parts are not taken into the calculation of the continuum. The boundaries of an absorption line are defined by a Voigt profile.
3.7
Error estimation
The error on the calculated EW is approximated by formula 3, where δλ is the spectral bin width, σi the
error spectrum and Fc the continuum flux. i is the range of wavelengths containing the feature, starting at
where the continuum intersects with the wing of the feature, till the other end. The error spectrum is an array of the estimated error per wavelength bin and this spectrum is made during the data reduction process with the X-Shooter pipeline. Calculated error spectra are then automatically saved in the fits files produced by the pipeline. σWλ,obs= δλ v u u t n X i=1 (σi/Fc)2 (3)
3.8
Spectrum normalization
For the graphs below the spectra were normalized. This was not needed for the above mentioned calculations as they already contained a component to correct for the normalization. The function continuum of IRAF was used to fit a spline function of order = 13 to the raw 2D spectrum. By dividing the spectrum by the fitted continuum, we get a normalized spectrum.
4
Results
The goal of this project is to study the optical and near infrared spectrum of the afterglow of GRB 160203A to measure the redshift of the host galaxy, to search for intervening systems and to find evidence for the response of the host galaxy to the GRB event. This research was done by studying the radial velocity (RV) structures of spectral lines in the GRB afterglow. Firstly, this sheds light on what kind of elements there were at the redshift of the burst (z=3.517). Secondly, the RV structure relates to the structure of the ISM of the host galaxy, as each intervening system leaves its trace in the afterglow signal. This should give us an indication of the environment that produces GRBs. This section elaborates on what techniques were used in the analysis. From the redshift and equivalent width (EW) calculations, up to how the variability is studied.
4.1
Spectral lines and their profiles
In this section the spectral lines of the host galaxy are described. From RV profiles we can gain information about the structure and contents of the host galaxy, as each velocity component and their position tells about how material is distributed throughout the galaxy. Abundances of heavy elements also constrain the number and what types of stars lived in the galaxy. For example, if no metals are detected, it is a primordial galaxy where massive stars and their supernovae did not enrich the interstellar matter yet. If we detect many and strong metal lines, the host is probably an older galaxy where in the past supernovae enriched its medium. This provides knowledge about the environment that has produced GRBs. This subsection contains graphs of the RV profiles of C, Si, Al and Cr in figures 28, 29, 30 and 31.
As the GRB pumps UV radiation into its surrounding, the environment gets excited. The transition lines with the weakest oscillator strengths are formed closest to the GRB, and decay the fastest. Low energy transitions can happen at a larger distance from the GRB, as well as close to the GRB. Consequently, these lines should have the broadest RV profiles. This can be seen for example in the C II and C IV profiles in figure 28. C II stretches out up to 100 km/s, it is saturated and could have 3 velocity components. On the contrary, C IV has one component and is only 50 km/s broad. C IV is higher ionized and thus would require a more energetic environment to form, and thus is the C IV component probably closer to the GRB. We see the same behaviour in Si (figure 29) and Al (figure 30). Cr II (figure 31) also shows 2 components, comparable with Si II and Al II. The red component of Cr II is stronger than its blue component. Cr II 2062 has only a red component.
We detect fine structure lines of Ni II, Fe II, Si II and C II. The fine structure line of C II can be seen in figure 28, which has the same profile as the resonance lines of C II. C II* and Si II* was also detected by Christensen et al. (2011); Th¨one et al. (2013). These transitions are likely made by UV pumping caused by the GRB. Si II* is also detected in star forming regions where young stars radiate UV (Christensen et al., 2011), and is therefore not unique to GRBs. Fe II* and Ni II* are not found in other environments than GRBs (Vreeswijk et al., 2007; Christensen et al., 2011). Multiple lines of Fe II* were detected and are described in subsection 4.2.
Figure 28: These are resonance lines of C II and C IV. The red line is the observation that started just 18 minutes after the burst, the blue line represents the observation taken 5 hours later. Both C IV and C II do not vary. C II is more saturated than C IV and shows 3 components, which could be made by noise. C IV shows one component, which could be close to the GRB.
Figure 30: These are resonance lines of Al III and Al II. The red line is the observation that started just 18 minutes after the burst, the blue line represents the observation taken 5 hours later. The Al II line is stronger than the Al III lines. All three profiles have 2 velocity components. Do note that the Al III lines show variation in time.
Figure 31: These are resonance lines of Cr II. The red line is the observation that started just 18 minutes after the burst, the blue line represents the observation taken 5 hours later. Cr II 2062 is the strongest feature of the three and does not show significant variation in time. The other Cr II lines do vary significantly in time. All three lines have 1 component, except for Cr II 2062, which has only a red component.
4.2
The fine structure of Fe II lines
The observations of the afterglow of GRB 160203A contain Fe II resonance lines in both epochs; Fe II fine structure lines can be found in the first set of observations as well. The 2nd epoch has too much noise to calculate EWs and thus determine the variability (figures 34, 35 and 36).
Information about the equivalent widths and how they vary over time provides the opportunity to derive the distance between the burst location and the regions where the lines are created (D’Elia et al., 2007; Vreeswijk et al., 2007). If the GRB is the sole source of energy, it is expected that the higher energetic tran-sition lines are formed only near the GRB. These lines should be weaker than the resonance lines, because of their weaker oscillator strengths, and should have less velocity components, as they are formed only in the direct environment of the GRB. This can also be seen in the figures below, as fine structure lines have one component, while resonance lines are broader and usually have more components.
Figures 32 and 33 present the RV profiles of Fe II resonance lines. From these images we can see that the RV profile has 2 components, meaning that iron is distributed over two distinct regions in the host galaxy. Moreover, the right component is stronger than the left one. One reason for this structure is that we detect two clouds of which the one moving faster away from us is more metal rich.
Figures 34, 35 and 36 present fine structure lines of Fe II. They are all relatively weak compared to the resonance lines. Fe II* lines have only been detected in GRB host galaxies (Vreeswijk et al., 2007; Chris-tensen et al., 2011). Although Fe II* is detected, it is weak in the 1st epoch, and gone because of the noise in the 2nd epoch. This makes it impossible to study Fe II* variability due to GRB 160203A, which could be done in order to obtain additional properties of the host galaxy (Vreeswijk et al., 2007).
Figure 32: These are RV profiles of Fe II resonance lines. The red line is the observation that started just 18 minutes after the burst, the blue line represents the observation taken 5 hours later. The observed equivalent widths can be found in the legend. Fe II 1608, 2261 and 2344 have two components, while 1611 only has one component. The emission peak in 1611 is a telluric. 2261 is observed at the start of the NIR arm of X-Shooter, where the S/N is low.
Figure 33: These are RV profiles of Fe II resonance lines. The red line is the observation that started just 18 minutes after the burst, the blue line represents the observation taken 5 hours later. Fe II 2374, 2383, 2587 and 2600 all show 2 velocity components. These lines are observed in the NIR arm of X-Shooter, whose spectra have a lower S/N ratio.
Figure 35: These are RV profiles of Fe II fine structure lines with a lower level of D5/2. The red line is the
observation that started just 18 minutes after the burst, the blue line represents the observation taken 5 hours later. All lines are weak and have one component, except for Fe II 2400.
Figure 36: These are RV profiles of Fe II fine structure lines with lower level D5/2. The red line is the
observation that started just 18 minutes after the burst, the blue line represents the observation taken 5 hours later. Fe II 2406 has 2 strong features, the other Fe II lines are weak.
4.3
Variability between each frame separately
In this project two definitions of variability are used. If EWs can be measured in multiple epochs, it is possible to see if they vary in time. It is expected that the EW of a spectral profile is correlated with the light curve of a GRB afterglow (see figure 9). Do note that EW variability does not scale linearly with the column density. The conversion factor between these quantities is the curve of growth, which describes how the EW changes as the column density varies. Although column density and EW do not convert linearly, a significant change in EW does mean a change in column density. A change in EW is therefore a valid method to measure the variability caused by a GRB.
Another method is to calculate the excess variance per wavelength over time. The excess variance is the variance over time per wavelength in a data set, minus the expected variance from the continuum. This unit is calculated by formula 4 (Vaughan et al., 2003).
σNXS =2 S
2− σ2
x2 (4)
In formula 4 is σ2
NXSthe excess variance, S2the variance of a wavelength bin, σ2 the expected variance measured in the continuum and x2the average value squared for normalization. NXS quantifies the difference between the expected variance and the observed variance. Fvar =qσ2
NXS is comparable of the root mean square of the variance and is contains the same information, but is also sometimes used to define variability (Vaughan et al., 2003). Do note that σ2NXScan be negative.
The excess variance method is preferred to the method of comparing EWs between epochs, as it quanti-fies the variability per wavelength bin, which is necessary to localize variability. The plots below show the variability of C IV, Fe II, Fe II* and Fe III.
In this subsection plots of frames were studied individually. Previously, both epoch images were stacked into two images. Now only the second epoch is stacked, as its S/N is too low for studying the individual frames. This way variability between the frames of the 1st epoch individually and the stacked frame of the second epoch can be studied. All graphs below excluded the 1st observation because of its short exposure time and thus low S/N. The last observation in the graphs is the stacked spectrum of the 2nd epoch, while observations 1-4 are spectra from the 1st epoch (see table 6). Overall, it does not seem that the RV profiles show variability.
Figure 37 presents the variability in C IV 1548. The top plot shows observed line profiles. From this plot we can see that C IV has one constant component. The middle plots show the variance between two images. This tells us when change happened. The variance taking all images into account is shown in the bottom plot, which is a method to find at which RV happens the most variability.
C IV 1548 is constant, but its environment does show variance. This indicates that more variability is happening in the continuum than in the line profile. The variance in the continuum is noise. This is especially visible in the bottom plot of figure 37 at 100-200 km/s. The right wing of C IV 1548 does show variability (σ2
NXS∼ 0.01).
We expect to see variability at the edges of the profile, because the center of the profile is saturated, while the wings are not. When a saturated spectral line decays, the wings decay first. Variability in the center of a line profile can only be seen in unsaturated lines.
Figure 37: This is C IV 1548 in 5 consecutive frames. The exposure time is mentioned per observation. The first plot shows the line RV line profiles. The second plot the variance between just two frames to
Fe II 1608 (figure 38), Fe II* 2383 (figure 39) and Fe III 2108 (figure 40) were studied because they are strong lines, which could be used in the models of Vreeswijk et al. (2007) to calculate properties of the GRB host galaxy. From subsection 4.2 we know that it is impossible to do this, because the UV pumping model needs to be fitted to multiple Fe lines to get reliable results. We do not have a comparable amount of usable lines as in Vreeswijk et al. (2007); Hartoog et al. (2013). We can however look at the variability in these specific lines.
Fe II 1608 (figure 38) has 2 components. Both centers are constant, comparable with C IV 1548, but show variability in their wings (σ2
NXS ∼ 0.02). The peak at ∼ −25 km/s moves to the left, which could indicate that the blue component becomes weaker, while the red component becomes stronger. This is only detected between frame 1-2. The peak does not change in observation 2-5.
Figure 39 presents Fe II* 2383, which is a fine structure line. It has one broad component, which probably exists of two saturated components comparable with Fe II 1608. Again we see that the wings have variability (σNXS2 ∼ 0.05). At 200 km/s we see that the continuum varies with σ2
NXS∼ 0.20, which means that the variability we see in the wings could be just noise.
Fe III 2108 can be seen in figure 40. This is a strong but unsaturated line, where we would expect to see variability across the entire profile. We do see variability in the centre (σ2
NXS∼ −0.01), but it is insignificant compared to the variability in the wings and the continuum (σ2
NXS∼ 0.02).
Overall it seems that the wings of the Fe profiles do show variability. It could be that the 1st epoch and the 2nd epoch are too close observed after each other to see change. Still, considering the optical lightcurve in figure 22, which tells us that the magnitude drops with 2 between frames 2 and 5, we expect to see decay. The individual normalized spectra of the 1st epoch were made by Jonatan Selsing, a PhD student from Copenhagen. He wrote a Python program to fit the continuum automatically for large collections of fits images. I used his spectra of the individual frames to make the plots below.
Observation Notes
1 2nd RRM image, because the first image had a bad S/N
2 3rd RRM image
3 4th RRM image
4 5th RRM image
5 Stacked images of the 2nd epoch
Figure 38: This is Fe II 1608 in 5 consecutive frames. The exposure time is mentioned per observation. The first plot shows the line RV line profiles. The second plot the variance between just two frames to see
