Skip banner and navigation tools.

 |  site map

About the automated XRT analysis of Gravitational Wave follow up

Contents


Overview

The analysis of XRT data collected in response to a Gravitational Wave (GW) event is largely automated, building on the library of automated analysis software created at the UKSSDC, and used for automated GRB products, on-demand analysis and 1SXPS catalogue. The process is outlined in Evans et al., 2016 which is currently the citable reference for these products. There have been a few small changes to the procedure since that paper was written, in that the source detection code has been revised, and the 2MPZ catalogue (Bilicki et al., 2014) is also used in the ranking process. A brief summary of the analysis procedure is given below.

Back to contents | Back to GW results


Using these data

The data available from this repository are available for use in research and publications, provided that this is acknowledged. We request two forms of acknowledgement. Where the data are first introduced, please cite Evans et al. 2016 (MNRAS, 460, L40) which describe the analysis procedured. In the acknowledgements section, please use the following text:

This work made use of data supplied by the UK Swift Science Data Centre at the University of Leicester.

Back to contents | Back to GW results


Overview of the analysis process

The basic concept of the XRT analysis of GW follow up is simple:

  1. Search for sources.
  2. Analyse sources.
  3. Rank sources by likelihood of being the counterpart.

This is made more complicated by two factors: the data arrive piecemeal (and we don't want to wait until it's all in to start the analysis) and many fields overlap. In order to ensure we get the best results as quickly as possible the same piece of sky can thus be analysed multiple times, and the results must be consolidated. Also, where fields overlap it is necessary to ensure that, ultimately, all data covering the same pice of sky are analysed togetehr to maximise sensitivity at that point. To enable this, the data are split into ‘analysis blocks,’ essentially groups of observation with a maximum size, defined so as to ensure that every field and every overlap between fields is analysed with the minimum number of separate analyses possible. More detail is given for the interested reader in the Analysis blocks section below.

Back to contents | Back to GW results


Source detection

As soon as an observation is received at the UKSSDC, the source detection code runs. The analysis blocks containing the observation are identified, and independent analysis jobs are executed for each such block. The source detection system is that developed for the 2SXPS catalogue (Evans et al., in prep), which is itself a modified version of the 1SXPS catalogue code (Evans et al., 2014). In summary: a 0.3—10 keV image, exposure map and event list are created for the analysis block. A sliding-cell source detection is performed, initially using a locally-estimated background and then, once the putative sources are masked out, using a background map. An iterative process follows whereby source searching is carried out, a PSF fit is carried out for any detected sources, the background map is recalculated accordingly, and the process begins again until no sources are found. The PSF fits are then repeated using the final background map, such that each source is fitted with knowledge of the other sources (to help reduce the effects of confusion). Sources are then assigned a flag indicating how likely it is they they are real sources rather than background fluctuations. These flags are defined in Table 1.

Flag Name
(Value)
False
pos rate
Cum. False
pos rate
Good (0)0.3%0.3%
Reasonable (1)7%1%
Poor (2)35%10%

Table 1. The flag schema. The names are referred to in the documentation, but are stored as numbers in the database. The cumulative false positive rate refers to the false positive rate in flags up to this value. i.e. the false positive rate when considering all Good and Reasonable sources is 1%, although the false positive rate among just the Reasonable sources is 7%.

Once source detection is complete, the list of detected sources is merged with the existing list of sources from other detect runs, to produce a single, consolidated list of sources in the GW analysis.

Back to contents | Back to GW results


Source analysis

When a source is first detected, re-detected, or an area of sky containing a source is observed but the source itself is not detected, the source is (re)analysed. Source analysis is fairly simple. A light curve of the source is created (using the tools from Evans et al. 2007), and both the mean and peak count-rates are determined. No spectrum is built since the large-area search for GW counterparts has only short exposure times and does not yield enough data for a reliable spectrum to be constructed. Count-rates are converted to flux using a canonical power-law spectrum, with a photon index of Γ=1.7 and an absorption column NH=3×1020 cm-2.

Various checks are carried out to attempt to identify whether the source is likely to be spurious, due to artefacts such as diffuse X-ray emission, contamination by optical photons from an optically-bright object, or stray light. If any of these checks come back positive, a warning flag is set so that XRT team members know to manually verify the reality of the source.

Finally, a search is done for historic X-ray emission from this source, from the 1SXPS catalogue and the HEASARC X-ray master catalogue. If a match is found, the catalogued flux is compared with the flux in the new detection. If not, an upper limit is produced from 1SXPS, the Rosat All Sky Survey, and the XMM-Newton upper limit server, where those facilities cover the source location. These data, along with the light curve, are used to rank the source accordingly to how likely it is to be related to the GW event.

Back to contents | Back to GW results


Source ranking

Each source is automatically classified. This process considers the source brightness compared to historical detections/flux limits, the light curve behaviour and whether the source is within 200 kpc (in projection) of a known galaxy. There are four possible classifications, listed here in decreasing order of ‘interest’. Note that, for public consumption, these classes are represented by a numerical ‘rank’ of 1—4.

Afterglow Candidate (Rank=1)

An afterglow candidate is source which is either:

  • Uncatalogued and at least 5-σ above the 3-σ upper limit from RASS or 1SXPS1.
  • A known X-ray source which is 5-σ above the catalogued flux1.

It must also lie within 200 kpc of a known galaxy (assuming it is at the distance of that galaxy).

Afterglow candidates meet the criteria to trigger an interrupt, that is, an urgent AT should be uploaded to gather more data on this source ASAP, typically for one snapshot, and then the survey of the GW region should be resumed (this requires recreating the observing plan, ODS have instructions for doing this).

Interesting source (Rank=2)

An interesting source is source which is either:

  • Uncatalogued and: at least 3-σ above the 3-σ upper limit from RASS or 1SXPS, or fading
  • A known X-ray source which is 3-σ above the catalogued flux.

Unlike afterglow candidates, an interesting source need not be near a known galaxy.

Interesting sources are not of enough merit to warrant interrupting the ongoing observations, however they should be prioritised; that is, they must be observed in the second round of observations. This is handled automatically by the GW planning software. If the source was only detected, or marked as interesting, in the second set of observations, it should be re-observed as soon as possible.

Uncatalogued X-ray source (Rank=3)

An uncatalogued X-ray source is an object which is not catalogued in X-rays, but also meets none of the criteria above to differentiate it from a field source unrelated to the GW trigger.

Uncatalogued sources should be re-observed in the PPST for 1 ks each, once all the AT observations have been completed and all data analysed, if no confirmed afterglow has been found.

Known X-ray source (Rank=4)

A known X-ray source is something which has been detected in X-rays before, and has a flux consistent with or below that from the previous observations. No further action is required for these sources.

1Where the historical count-rate/upper limit was not derived from XRT data, they have been converted to equivalent XRT PC mode 0.3—10 keV count-rates using PIMMS, assuming a typical AGN spectrum: NH=3×1020 cm-2, Γ=1.7.

Back to contents | Back to GW results


Source publishing

Newly-detected sources are not automatically made available on these public web pages. The short exposures means that the automated processeds to identify spurious detections are less reliable than is normal. Therefore sources have to be manually verified by members of the XRT team before they are pushed to these public web pages.

Back to contents | Back to GW results


Analysis blocks

This section can be safely ignored, unless you are interested in probing the detection history of a specific source or are especially interested in the details of how the analysis is conducted.

The XRT fields are spread over a large region of the sky, and form many disconnected fields or clumps of fields. Where a field contains a known galaxy with a distance consistent with the GW distance, the pointings are optimised to ensure full galaxy coverage with the UVOT, which has a smaller field of view, and thus a number of XRT fields can closely overlap. There is a limit to how many of these fields it is sensible to combine into a single image, for software reasons and because of the coordinate inaccuracy inherent to large tangent-plane projection images. At the same time, where multiple XRT fields of view overlap, we want to take advantage of this overlap to ensure the deepest exposure.

To address all this, we use the concept of analysis blocks (hereafter just ‘blocks’).

A block is a group of XRT fields of view, no larger than a standard 7-point tile (i.e. no more than 3 fields across). Blocks can overlap with each other, and blocks are defined such that, with the minimum possible number of blocks, every XRT field, and every overlap between fields is part of a block.

For example, consider a row of four XRT fields of view, 20′ apart (i.e. slightly overlapping), all with the same declination. Call these fields, A, B, C and D. These fields would form 2 blocks: ABC, and BCD. This means that some parts of the sky will be analysed as part of two blocks: those covered by only field B, only field C, or the BC overlap. Some parts of the sky will be analysed in only one block: those covered only by field A or only field D. And some parts of the sky are more confusing: the overlap between fields A and B is analysed in block ABC; that part of the sky is also analysed by block BCD, but only the data in field B are included in that analysis.

Each time data are received, the computer works out which block(s) those data corresponds to, and requests an analysis of that/those block(s). If any of these blocks are already being analysed, the request is queued, and a new analysis of the block, taking into account the new data, will be carried out once the ongoing analysis completes. The ongoing analysis will not be aware of the new data.

The use of blocks can lead to various subtle effects. First, (using the above fields/blocks), a source lying in the AB overlap which is only just detected in block ABC will be undetected in block BCD, even though it lies in the part of the sky covered by both blocks. The second effect concerns source variability, and the fact that the data are not received simultaneously. Again using the above example, imagine that the data are received consecutively, in alphabetical order. When the data for field A are received, block ABC is analysed. When the data for field B are received, blocks ABC and BCD are analysed. Imagine that there is a fading source in field A, which is only just detected in the first analysis of block ABC. When the second analysis runs, the background will be higher in the AB overlap, and the source has faded: this may mean that the source is not detected in the second analysis of block ABC, even though that contains more exposure. Because of this, once a source has been identified it cannot be removed from the source list, even if later analyses fail to find it; its light curve will also be updated every time a block in which it lies (even if it was not detected that block) is analysed. Therefore there is no danger of sources being ‘lost’ because of the above issues, but these subtleties must be borne in mind when considering the history of a source.

Back to contents | Back to GW results


Human intervention

As noted above, while the analysis is automated, the short exposures necessitate human verification in some cases. Humans also check each source to confirm the automated rank is correct. These checks are done before the source is published to the public website.

Verifying the detection

There are a number if issues the XRT team will check for before accepting a source is real. In most cases, sources affected by these issues are automatically flagged as likely to be spurious.

Optical loading
Optical loading arises when a bright optical source lies in the field of view, and the accumulation of optical photons are registered by the XRT CCD. Depending on the optical brightness, this can result in an apparent source being detected: usually easy to identify as it is much more concentrated than the standard PSF. For very bright optical sources, a ‘ring of fire’ artefact can be produced, an example is in the 1SXPS documentation. The source detection software checks the bright star catalogue used by UVOT to identify any bright sources capable of producing optical loading, and a warning is given for any X-ray source close to such a star. However, this catalogue does not contain all of the red sources which can induce optical loading, and the optical sky is variable, thus some sources will not be flagged. Equally, not all flagged sources are optical loading induced: the flag is to draw human attention to the possibility, not to confirm the source as spurious.
Optical loading in the XRT is discussed in detail on this dedicated page. XRT experts check the data for evidence of optical loading.
Stray Light
X-rays from a bright X-ray source 35′—75′ from the XRT boresight, can undergo a single reflection from the X-ray optics which directs them onto the camera. This is manifested as a set of concentric rings. Due the Poisson effects, these rings are not homogeneous, and these scattered photons sometimes give rise to spurious source detections. The automatic software checks for the presence of a catalogued X-ray source 35′—75′ from the XRT boresight and will warn if a source may be affected by stray light. As with optical loading, this is not perfect since X-ray catalogues are incomplete, and the X-ray sky varies. Also, genuine X-ray sources can exist within the stray light rings, therefore human verification is needed of each source to confirm whether it is affected by stray light.
The new source detection code developed for 2SXPS and employed from O3 onwards automatically seeks to model stray light and is thus much more robust against such spurious detections. However, in the short exposures employed in GW follow up, the stray light rings are often so sparesely sampled that this fitting dsoes not succeed.
Extended sources
The source-detection system looks for point sources, so extended sources can easily give rise to multiple spurious detections.If several sources are detected very close together, they are flagged as likely spurious, but XRT experts will again check and confirm if any of the sources are real point-sources, or if they should all be rejected.

Back to contents | Back to GW results


Identifying the GW counterpart.

As discussed earlier, the software attempts to automatically classify how likely each source is to be the counterpart to the GW trigger. However, this system cannot be perfect and humans attempt to verify this. In particular, given the large area of sky covered it is not impossible that XRT will serendipitously observe a recently-discovered transient that is not yet in any catalogues but has been reported, e.g. as an ATEL. Thus humans must check for any such matches to any transients flagged up by our source ranking.

Back to contents | Back to GW results


Web pages

The results of the XRT automated analysis are presented online, via this website. The front page lists all of the triggers followed to date, most recent first, and provides links to view the XRT results, or view/search the list of observations carried out.

The XRT results are presented over three pages:

Trigger summary pages.
The main page for each trigger, summarises the results.
Source pages.
Gives details of a specific XRT source.
Source image pages.
Allows images of each analysis run of each block to be analysed.

Back to contents | Back to GW results


Trigger summary pages

The trigger summary pages exist, one per trigger, to give a top-level overview of the results so far. At the top left of the page some details are given of the GW trigger (taken from the LVK GCN notices), along with links to the GraceDB page for this trigger. Below this are details of the XRT observations to date. This lists how many fields were in the original observing plan, and of these, how many have been observed so far. The targetID range covered by the planned fields is also listed. At the end of the table the number of XRT sources found is given, both the total, and the number per classification. A link is also given to view the individual analysis blocks for this trigger.

Below this is an expandable section labelled, “Externally-detected sources”. If Swift has carried out ToO observations of potential counterparts found by other people, then these are listed here along with either an XRT upper limit or a link to a source matching the external source. XRT team members may also register the positions of other sources reported in GCNs, even where no ToO was requested: if those sources happened to lie within the area covered by XRT, then the upper limit or matching source will be shown here.

To the right of this table an Aitoff-projection all-sky image is shown, with the planned XRT fields marked. Red circles are fields for which data have not yet been received and analysed, green fields are those for which they have. A (much) higher resolution image can be reached by clicking on the small image.

Next, a table gives details of all of the sources which have been detected by XRT. Above the table there are controls to:

Regarding these last two points: XRT source detection is fairly reliable but not perfect, and in particular issues such as optical loading, stray light or extended sources can lead to spurious detections. The XRT team will manually check each source and indicate when they have passed. Note: passing the vetting process does not mean the source is real, only that there is nothing obviously wrong with it. The source detection flags, Good, Reasonable and Poor are set based on the source statistics, using the system developed for the SXPS catalogues. Roughly speaking, these correspond to 3- 2- and 1- σ detections respectively. Since Poor sources are very low significance, they are hidden by default, but this control lets you show them if you wish.

The table can be sorted on any column by clicking on it. Those columns are:

#
The source identifier
Position
Ra/Dec of the source in the XRT coordinate frame (i.e. this is not an enhanced position)
Err
90% confidence error radius in the position.
Flag
The detection flag from the source detection system, which identifies how likely it is that the source is a genuine astrophysical object, rather than a spurious detection. Table 1 gives details of the flags.
Class (‘All X-ray sources’ table only)
The automatic classification of the source. See Source classes, above.
Cat source?
Whether the source matches a previously catalogued X-ray source, from the 1SXPS catalogue, or those contained in the ‘X-ray master’ database table, maintained by HEASARC. For space reasons, this column simply contains ‘Y’ or ‘N’, for details of the source matches, see the Source pages.
Fdiff (σ)
This reflects the brightness of the source, compared to previous observation, in σ. i.e. \( \left(R - R_c\right)/\sigma_R \), where \(R \pm \sigma_R \) is the measured peak XRT count-rate with its uncertainty, and \(R_c\) is either the historical count-rate (for known sources) or 3-σ upper limit (from 1SXPS or RASS, whichever is deeper). Where the historical count-rate/upper limit was not derived from XRT data, they have been converted to equivalent XRT PC mode 0.3—10 keV count-rates using PIMMS, assuming a typical AGN spectrum: NH=3×1020, Γ=1.7.
Near Galaxy
Whether the source lies within 200 kpc of a known galaxy in the GWGC or 2MPZ catalogues, assuming it is at the same distance as that galaxies. For space reasons, this column simply contains ‘Y’ or ‘N’, for details of the galaxy matches, see the Source pages.
Near 2MASS
Whether the source lies near any 2MASS objects. This is important because very red objects may be flare stars, which are a likely source of transient X-ray phenomena unrelated to the GW trigger (see human checks). As with the galaxies and X-ray catalogue matches, this column simply contains ‘Y’ or ‘N’, for details of the galaxy matches, see the Source pages.
Near SIMBAD
Whether a SIMBAD object matches the source position. If the answer is ‘Yes’ the word also links to the SIMBAD entry for the object in question. Note: only the nearest match returned by SIMBAD is linked to in this way.
Search
This field provides links to cone searches centred on the source, and with a radius 3 times the 90% confidence error. There are 6 links:
V
Vizier
S
SIMBAD
N
NED
X
X-ray master catalogue (via LEDAS)
1S
1SXPS
3X
3XMM-DR5 (via LEDAS)
Notes
Any extra notes associated with the source.

You can sort the table by any column by clicking on the column header. The column on which the table is currently sorted will have a red or blue header, red meaning that the sort is in ascending order, blue meaning descending.

Back to contents | Back to GW results


Source pages

A source page exists for each detected X-ray source, giving details of the source, possible catalogue matches, and the ability to mark (or unmark) a source as spurious.

At the top right of the page is an image of the source, approximately 6.3′ to a side (a scale is given on the image: the cross-hair lines are each 1′ long), showing the source in the block and processing version that yielded the strongest detection. A link beneath this image allows you to view all images covering the location of this source.

At the top left is a table, split into two sections. The first, “Source details,” gives the basic details of the source (position, error, detection flag etc.). The peak count-rate of the source is also given, taken from the light curve. If no curve was produced, this field will be undefined. The peak flux is also given, converted from the count-rate assuming a standard AGN spectrum (NH=3×1020, Γ=1.7).

The second section of the table concerns information from other catalogues. Note that any catalogued sources reported here are identified purely based on spatial coincidence, and this does not guarantee that they are related to the X-ray source. When constructing the 1SXPS catalogue we investigated the probability of spurious matches, by randomly shifting the positions of all 1SXPS sources, and performing a catalogue cross-match: the probabilities of spurious matches are shown in Table. 2.

The first field reports whether the source matches a known X-ray object, and if so, what object that is. Where multiple matches exist (a common phenomenon) preference is given to the 1SXPS catalogue, because that is a Swift-XRT catalogue, making comparisons of source fluxes etc. easier. A list of all matching X-ray catalogued objects can be obtained using the links at the end of the table. If the source was catalogued, the catalogued count-rate and flux are also reported, as well as the ratio of observed to catalogued flux. This is given in σ, i.e. \( \left(R - R_c\right)/ \sqrt{\sigma_R^2 + \sigma_C^2} \), where \(R \pm \sigma_R \) is the measured peak XRT count-rate with its uncertainty and \(R_c \pm \sigma_C\) is the historical count-rate and error. If the XRT source is fainter than the catalogued source, the ratio is simply given as “≤0” Of course, the catalogued flux represents only a snapshot of the source's flux, so the possibility must be considered that the source is variable. The link at the end of the table to search for all X-ray matches should be used to consider the range of fluxes at which the source has been measured.

If no matching X-ray source is known, the RASS and 1SXPS 3-σ upper limits (if available) are given and the ratio \( \left(R - U_C\right)/ \sigma_R \) is given, where \(R \pm \sigma_R \) is again the measured peak XRT count-rate and uncertainty, and \(U_C\) is the historical 3-σ upper limit.

Next, whether the source matches known galaxies or 2MASS objects is also reported, and a link is provided to a table further down the page, giving details of those matches. Similarly, if a SIMBAD object is spatially coincident with the source, details of that match (and a link to SIMBAD) are provided.

Finally, a series of links are given to cone searches centred on the source, with a radius 3 times ‘Err’.

Below the table a light curve of the source will be given if one was successfully created. This light curve is produced by the tools that make the GRB light curves, and the binning is set to the standard GRB binning (i.e. minimum of 15 count/bin), with two minor modifications:

At the foot of the page, tables may exist giving details of any galaxies within 200 kpc (in projection) of the XRT source, and any 2MASS objects spatially coincident with the source. For each table controls exist to select whether coordinates are reported in sexagesimal or decimal format, and clicking on any column header sorts the table on that column (click again to invert the search order). The column on which the table is currently sorted will have a red or blue header, red meaning that the sort is in ascending order, blue meaning descending.

For the galaxy table, the position of the galaxy and the angular distance from the galaxy centre to the XRT source are given, along with the projected distance in kpc, assuming the source is at the same distance as the galaxy. The major and minor radii of the galaxy, and the position angle (in degrees from North, via East) is given, and the galaxy distance, absolute B magnitude, and which catalogue it came from.

For 2MASS objects, the object ID is given, and this links to the catalogue entry for that object (in Vizier). The 2MASS position, the angular distance of this from the X-ray source, and the JHK magnitudes of the 2MASS object are also given.

Back to contents | Back to GW results


Source image pages

Note that in order to understand these pages, you must first read the explanation of analysis blocks.

The source pages show a single image of the XRT source, but also provide a link to view all images. That link opens the source image page.

By default, the source image page shows one image for each block in which the source is detected, the image shown being the most recent processing of the block in which the source was detected. Yellow cross-hairs indicate the source position. A label above the image indicates which block and processing version the image refers to. Clicking on this will reveal which obsIDs were used in that block/processing version, and how much exposure each obsID had at analysis time.

At the top of the page are a series of controls. The first allows you to decide whether to view the image, or the exposure map. The second lets you choose whether or not to show non-detections, that is all blocks/processing versions in which the source location was covered, but the source was not detected. In these images again the cross-hairs indicate the location of the source, even though it was not detected. The label above the image appears in red for non-detections. Finally, you can choose which images for each block are shown: this can be only the latest processing that yielded a detection (default; this option is not available if non-detections are included), the latest processing only, or all processing versions.

The main purpose of this page is diagnostic, to give XRT experts access to all images regarding a source, which may be helpful in determining if marginal detections are real or not.

Back to contents | Back to GW results


Galaxy checker

The galaxy checker allows you to identify catalogued galaxies ‘consistent with’ a given position. Each catalogued galaxy has a position, major and minor axis radius, and a position angle. To this, a 200 kpc halo has also been added. A galaxy is deemed to be ‘consistent with’ with your input position if your position lies within the ellipse formed by the galaxy+halo.

By default this will only return galaxies with a distance consistent with the GW distance estimate along the specified sightline, at the 3-σ level; a checkbox lets you remove this contraint.

Back to contents | Back to GW results


Observing schedule

Links on both the source index page and the overall index page allow you to view the list of fields observed to date or to search to see if a specific position is covered within the XRT field of view. The table controls let you determine the format of the coordinates to return. When viewing the full plan you can also see the time at which the observation occured. By default this is in Swift Mission Elapsed Time; this can be shown instead as a UT date but this is very slow!

Back to contents | Back to GW results