Survey of Globular Cluster 47 Tucanae

Science question

Globular cluster cores have revealed themselves as fertile grounds for stellar X-ray activity at both bright and low luminosities, offering up a variety of types of X-ray point sources for study. The lower luminosity sources include several kinds of systems with different X-ray emission mechanisms, and the bright sources have been identified as low-mass X-ray binaries containing neutron stars. In this thread, we survey the X-ray photometric and spectral characteristics of the point-source constituents of Galactic globular cluster 47 Tucanae, as seen through the eye of Chandra and recorded in the CSC. This results in a comprehensive table of source properties and simple spectral fits which can be used for source classification and analysis.

Creating a source list

We begin by using the catalog query interface, CSCview, to compile a preliminary list of Chandra observations which targeted 47 Tucanae. Later, we use the catalog to filter this list and retrieve pre-calculated source properties and data products for all sources in the selected ObsID(s).

CSCview is launched from a link in the navigation bar of the CSC web pages (CSC homepage: http://cxc.harvard.edu/csc/). CSCview opens on the Query tab, where we may specify our search criteria and desired search results.

Figure 1.Query tab of CSCview

To begin our query, we first clear any example entries which may be present in the interactive windows of the Query tab by selecting the "File->New->Empty Form" menu option - or by highlighting these items and selecting the '-' button next to the appropriate window. (You may choose to have the query form appear empty upon startup, or populated with an example query with the "Startup Query" option in "Edit->Preferences"; the latter is the default startup option.) Then, we drag-and-drop several "Source Observation->Observation-Specific Information" properties from the provided list into the Search Criteria window to define the following:

• (o.targname LIKE 47 TUC%) AND
• ((o.detector LIKE ACIS-%5%) OR
• (o.detector LIKE ACIS-%7%))

This search statement will locate all sources in the catalog in Chandra ACIS observations which targeted 47 Tucanae (or, more specifically, observations with source target names beginning with "47 TUC"). We choose observations with sources lying on at least one of the back-illuminated ACIS-S chips (ACIS chip numbers 5 and 7), since they maximize sensitivity to low-energy sources.

In the Result Set window, we list the properties we would like returned for each observation. Even though all properties contained in the catalog are tied to individual sources, we can create a preliminary list of full-field observations by selecting only observation-specific properties for our result set:

• o.obsid
• o.targname
• o.ra_targ
• o.dec_targ

The ObsID is a number assigned to each Chandra observation which indicates a specific target and observational configuration provided in a Chandra proposal. The target name and target equatorial coordinates match those of the primary object targeted by each observation (in this example, 47 Tucanae). Choosing the "Select: 'all' 'distinct'" rows option above the Result set window allows us to access a table of results containing only one row for each distinct ObsID found in the search, instead of multiple rows of identical observation information for each source in each observation. Finally, in the Sort Order window we specify that the table of results to be returned should be sorted on ObsID, in descending order.

The completed query form is shown in Figure 2.

Figure 2. Completed query form in the Query tab of CSCview

The results of the query appear in the Results tab after selecting "File-->Search" in the Query tab, as shown in Figure 3.

Figure 3. Table of query results in the Results tab of CSCview

We see that eight ACIS-S and/or ACIS-S/ACIS-I imaging observations of 47 Tucanae were made with Chandra during the first eight years of its mission. To narrow the list of observations down to one or two for an in-depth study of the sources in 47 Tucanae, we first reject those which are not confined to the back-illuminated ACIS-S chips, and then those with relatively few source detections overall. We can do the former by re-visiting the Query tab and modifying the original query accordingly: 1) we add the 'o.detector' property to the Result Set window, set 'Select' to 'all' 'rows', and finally submit the modified query. The table of results - which now contains one row for each source in each observation which satisfies the search criteria - shows us that ObsIDs 3387, 3386, 3385, and 3384 have few sources on ACIS chip 7 relative to the rest of the ObsIDs in the list, and none of the observations in the list have sources on ACIS chip 5. ObsIDs 338* are now candidates for rejection from the list of observations.

Figure 3. Table of query results with detector information for each source

Next, we clear the query form in the Query tab with the menu item "File-->New", and define a new query:

• Search Criteria: 'o.obsid = 3387'
• Result Set: 'o.posid'
• Select: 'count' 'rows'

The 'Select: 'count' 'rows'' setting will return not a table of source properties after the query is submitted, but one number: the total number of individual sources detected in ObsID 3387, each identified by the 'posid' catalog property. Repeating this query for the rest of the ObsIDs in the list tells us that ObsIDs 3387, 3386, 3385, and 3384 each contain less than 20 source detections in total, while ObsIDs 2738, 2737, 2736, and 2735 each contain more than 150 sources. As a result, ObsIDs 338* can now be rejected on account of having a suspiciously small number of total source detections relative to the other observations, ObsIDs 273*.

To make the final cut to our observation list, we reject observations with relatively high background counts. This will give us higher signal-to-noise for our sources, as well as maximize the number of point source detections.

We can make a source significance cutoff in a number of ways with CSCview; we choose to download and compare the background counts images for the ObsIDs in our list, to find one with a smooth, low background. (See also the CSC significance source properties 'o.detect_significance', 'o.flux_significance', and master 'significance'.) We can examine and analyze the downloaded background images with SAOImage DS9, e.g. with the DS9 menu items "View-->Horizontal/Vertical Graph" and "Analysis-->Pixel Table".

Level-3 data products, such as images, event lists, and spectra, may be downloaded for each source contained in the CSC from the Products tab of CSCview (see the Data Products page for the full list of data files provided by the CSC). To download background images in various science energy bands for each ObsID in our final list, we first re-visit the Query tab and add the 'dataset_id' catalog column to the Result Set window, and re-submit the query (the 'dataset_id' column is used to access data products associated with sources in the catalog). Then, we select the appropriate rows in the resulting query results table in the Results Tab, the desired data products in the Data Products window, then "File-->Search" (see Figure 4). This opens the Products tab, where the list of selected files appear ready for download. Refer to the CSCview thread Retrieving Data Products for details on download options.

Figure 4. Selected CSC data products in Results tab of CSCview

After unpacking the background images for our ObsIDs, we can load many of them simultaneously into a DS9 display for easy comparison:

unix% ds9 acisf0273*_000N001_b_bkgimg3.fits.gz &

Figure 5 shows the resulting display, a comparison of the CSC broad-band background counts images for ObsIDs 2738, 2737, 2736,and 2735:

Figure 5. DS9 display of CSC full-field, broad-band background images

Inspection of the backgrounds leaves ObsID 2386 as the focus for our study. We now have a preliminary source list: all sources detected in the 47 Tucanae Chandra observation ObsID 2386. After refining the preliminary source list, we will conduct two different catalog queries for these sources: one specific to imaging and photometric analysis, and the other better suited to spectral analysis.

Refining the source list

In order to minimize the number of contaminating foreground and background sources in our analysis, we decide to restrict our survey to point sources contained within the half-mass radius of 47 Tucanae, equal to 2.79 arcminutes. Since globuar clusters like 47 Tucanae are believed to be old enough for dynamic relaxation and mass segregation to have already occurred, this volume of space in the cluster should contain the most massive sources. As a result, it should maximize the detection of the types of sources we expect to find there, namely millisecond radio pulsars (MSPs), X-ray active binaries (ABs), accreting cataclysmic variables (CVs), and quiescent low-mass X-ray binaries (qLMXBs).

To search for these core sources, we clear our most recent query in the Query tab ("File-->New") and start a new one; we enter "o.obsid = 2736" in the Search Criteria window, the equatorial coordinates of the center of 47 Tucanae in the Cone Search 'ra' and 'dec' spaces, and 'radius= 2.79 arcmin'. The Name Resolver feature of the cone search can be used to automatically enter these coordinates by searching on the name "47 Tucanae". These conditions serve as the search criteria for the new query; they are related by "AND" search logic.

Figure 6. CSCview cone search of sources within the half-mass radius of 47 Tucanae

Retrieving CSC source data

To complete the query and retrieve CSC source data for analysis, we specify in the Result Set window all of the master source and source observation characteristics we wish to know. A list of example queries are available as templates via the "File-->Open-->Standard Query" menu item; they include queries which may be used for photometric and timing analysis. We decide to enter a modified version of the Master Photometry Summary query, which includes source position, aperture source energy fluxes and background-subtracted source counts in various energy bands, the hard-to-soft hardness ratio, and the upper and lower confidence limits associated with each of these quantities (not shown in the list below for brevity). Finally, we include all available source flags to isolate sources which are potentially unreliable for scientific analysis.

• ra
• dec
• match_type
• o.src_cnts_aper_b
• o.src_cnts_aper_s
• o.src_cnts_aper_m
• o.src_cnts_aper_h
• flux_aper_b
• flux_aper_s
• flux_aper_m
• flux_aper_h
• hard_hm
• hard_hs
• hard_ms
• extent_flag
• conf_flag
• pileup_flag
• var_flag
• streak_src_flag
• sat_src_flag
• man_inc_flag
• man_reg_flag
• man_match_flag
• var_inter_hard_flag
• posid

We choose master source properties along with source observation properties in order to have any best-estimate properties which may be available for the sources in ObsID 2736. (For brief definitions of each source property contained in the catalog, refer to the Master Chandra Source Table and Table of Individual Source Observations; for high-level explanations, see the Catalog Descriptions pages.) If the 'match_type' column has a value of 'a' for any of the sources, then we will know that the source is ambiguously matched to at least two master sources in the catalog; this means it is a confused source and its properties are not used in the calculation of any best-estimate master source properties. If 'match_type=u', however, the source is uniquely matched to a single master source, and if it is not saturated, the source has a set of best-estimate properties. For more on the distinction between master source and source observation properties, as well as the nature of master sources/source observations associatons, refer to the Catalog Organization page and the CSCview thread "Using the Source Property Associations".

Before submitting the completed query, we set the "Edit-->Preferences-->Output Coordindate Format" preference to "decimal", so that the RA and Dec values in the search results table each occupy one, not three, columns; this will make the save file of results easier to manipulate with analysis tools later. After submitting the query with the "File-->Search", we save the table of source data to a text file by selecting "File-->Save" while the Results tab is open. In the section "Analyzing CSC source data with CIAO, Sherpa, and ChIPS" we use the plotting program ChIPS to plot our sources in X-ray color-magnitude space.

First, we select and download all desired data products for our sources in the Results and Products tabs, respectively, so that we may exit CSCview and begin our analysis with CIAO, ChIPS, and Sherpa.

We can download source-specific and full-field data products as described in the section "Downloading CSC data products". This time, we choose the full-field event file, field-of-view file, the b-, s-, m-, and h-band sensitivity maps, as well as the source region files, which contain both the source and background region ellipses. Finally, we select the exposure-corrected fluxed event images in the b, m, s, and h energy bands, each containing multipe images at varying resolution.

Figure 7. List of selected CSC data products for sources in 47 Tucanae, in the Results and Products tabs

Since there is a limit to the number of files which can be downloaded at once with CSCview, we choose to utilize the "File-->Generate Download Script" feature, which is more time-consuming but will access all desired files in one download session. The generated script, which has a default name of 'cscbatch', contains a list of GNU Wget commands, one for each data product to be downloaded. The selected data products can be downloaded by executing the batch script at the command line, as follows:

unix% chmod 755 cscbatch   #make executable
unix% ./cscbatch
Beginning download of CSC files.
--2009-03-13 16:41:23--  http://cda.harvard.edu/l3services/archiveFile.do?level=3&filetype=evt3&filename=acisf02736_000N001_evt3.fits&dataset=flight
Resolving cda.harvard.edu... 131.142.185.171
Connecting to cda.harvard.edu|131.142.185.171|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: unspecified [image/fits]
Saving to: `acisf02736_000N001_evt3.fits.gz'

    [  <=>                                ] 4,057,080   2.92M/s   
 2009-03-13 16:41:35 (1.59 MB/s) - `acisf02736_000N001_evt3.fits.gz' saved [17006673]

--2009-03-13 16:41:35--  http://cda.harvard.edu/l3services/archiveFile.do?level=3&filetype=srcreg&filename=acisf02736_000N001_r0010_reg3.fits&dataset=flight
Resolving cda.harvard.edu... 131.142.185.171
Connecting to cda.harvard.edu|131.142.185.171|:80... connected.
HTTP request sent, awaiting response... 


[rest of output omitted]          

Visualizing the source list with CIAO and DS9

Armed with data products for our sources, we can now visualize all detected source region ellipses overlaid on an events image of the cluster. The Level-3 reg3.fits files we downloaded for each source contain both source and background region descriptions, so we simply compile these regions into one DS9-compatible ASCII file to be loaded into DS9 along with the evt3.fits file for ObsID 2736:

ciao% foreach ff (acisf02736_000N001_r*_reg3.fits)
foreach? dmcopy ${ff}"[SRCREG]" ${ff}.src.reg 
foreach? dmcopy ${ff}"[BKGREG]" ${ff}.bkg.reg 

ciao% foreach ff (*.reg)
dmmakereg "region(${ff})" ${ff}.ascii kernel=ascii 

ciao% more *src.reg.ascii | grep "physical" >  just_source_ellipses.reg
ciao% more *bkg.reg.ascii | grep "physical" >  just_bkg.reg

ciao% more just_bkg.reg | grep -v "Polygon" > just_bkg_ellipses.reg

First, the source and background region descriptions contained in each reg3.fits file are copied to individual region files, *src.reg and *bkg.reg, using the CIAO DM tool dmcopy. These *.reg files are then converted from FITS to ASCII format with dmmakereg, and subsequently compiled into DS9-compatible ASCII region lists. (Note that the following comment must be added to the top of files just_source_ellipses.reg and just_bkg_ellipses.reg:

# Region file format: DS9 version 3.0 global color=blue font="helvetica 10 normal" select=1 edit=1 move=1 delete=1 include=1 fixed=0

). These files may be loaded into DS9 with the '-region' option, along with the field-of-view for the observation, as follows:

unix% ds9 acisf02736_000N001_evt3.fits -region just_source_ellipses.reg  -region just_bkg_ellipses.reg -region acisf02736_000N001_fov3.fits &

The resulting figure is shown in Figure 8, with a zoomed-in version for better visualization of the cluster source and background region ellipses.

Figure 8. CSC source regions and full field-of-view overlaid on a Chandra events image of 47 Tucanae

The events image with all 93 source detections plus background regions in Figure 8 is so crowded with CSC region ellipses that the cluster itself is not visible! Therefore, we load only the source regions without the background components for visual clarity, and then just a small subset of the source regions:

unix% ds9 acisf02736_000N001_evt3.fits -region just_sources.reg &
unix% ds9 acisf02736_000N001_evt3.fits -region just_sources_short.reg &

The resulting figures are shown in Figure 9.

Figure 9. Events image of cluster 47 Tucanae with overlaid source region ellipses

Refer to the CIAO threads Using CIAO Region Files and Using SAOImage DS9 to learn more about how region files are used in CIAO and DS9.

Each spatial region defining a source and its corresponding background in the CSC is determined by scaling and merging the individual wavdetect source detection regions from all spatial scales and source detection energy bands. The result is a single elliptical source region which excludes any overlapping source regions, and a single, co-located, scaled, elliptical annular background region (not to be confused with the error ellipses which describe source position uncertainty). Most CSC "source" properties, such as spectral fit fluxes, light curves, and the aperture photometry "*_aper" values (but not "*_aper90") are extracted from this source region.

Next, we create a smoothed, 3-color image of the cluster with the img3.fits files we downloaded. The CIAO thread True Color Images in DS9 lists step-by-step instructions on how to create such images in DS9, starting with the manipulation of an event file. However, since the CSC comes pre-packaged with exposure-corrected fluxed events images in 6 science energy bands for each observation, all we have to do is select three of the img3.fits images we dowloaded earlier and load them into DS9 with the '-rgb' option:

unix% ds9 -rgb -red acisf02736_s_img3.fits -green acisf02736_m_img3.fits -blue acisf02736_h_img3.fits &

The resulting 3-color image is shown in Figure 10.

Figure 10. 3-color image of 47 Tucanae, produced by combining CSC img3.fits files

Here we have combined the soft-, medium-, and hard-band CSC images into a 3-color image, where we have smoothed the data with the DS9 "Analysis-->Smooth" feature, using a Gaussian function with a kernel radius of 2. Note that if we had wanted each image to encompass an energy range different than that used in the CSC, we could have used dmcopy as shown in the CIAO thread to modify the evt3.fits file accordingly.

Now we can use CIAO to examine the CSC source properties we retrieved with CSCview, to conduct photometric and spectral analyses of the sources in 47 Tucanae.

Analyzing CSC source data with CIAO, Sherpa, and ChIPS

CSCView output format

Tab-Separated-ValueCIAOkernel=text/tsvUsing a CSC Save File in CIAO

---

Plotting CSC source properties with ChIPS

To convert the CSC broad-band energy fluxes for our sources to X-ray luminosities, we assume a distance to 47 Tucanae of 4.85 kpc (refer to the Source Fluxes page and the memo "Computing Flux and Flux Significance" to learn how CSC fluxes are calculated). This distance corresponds to a flux-to-luminosity conversion factor of 2.8e45 cm²; we multiply this value by the 'flux_aper_b' column in the file 2736_src_data.tsv, using the tool dmtcalc:

dmtcalc "2736_src_data.tsv[opt kernel=text/tsv]" 2736_src_data.ciao expression="luminosity_aper_b=((float)flux_aper_b)*(2.8e45)"

Note: the default output format for CIAO is FITS, regardless of the input format. So the file 2736_src_data.ciao is now in FITS binary table format. Now we use ChIPS to plot the broad-band X-ray luminosity versus the CSC hard-to-soft hardness ratio, computed from the source counts in the hard and soft CSC science energy bands; this should help distinguish some of the thermal emitters from the non-thermal sources in the cluster:

ciao% chips

chips> add_curve("2736_src_data.ciao[cols hard_hs, luminosity_aper_b]"
chips> set_curve(["line.style","none"]) # remove line connecting data points

chips> set_plot_xlabel ("source hs hardness ratio")
chips> set_plot_ylabel ("L_x(0.5-7.0), ergs s^{-1}")

The resulting plot is shown in Figure 11.

Figure 11. ChIPS plot of CSC broad-band X-ray luminosity versus hard-to-soft hardness ratio

To be consistent with the literature, we can plot the broad-band luminosity against a hard/soft hardness ratio defined differently, as 2.5*log(soft counts/hard counts). We use dmtcalc again, this time to add a new column to 2736_src_data.ciao with the new hardness ratio values:

ciao% dmtcalc 2736_src_data.ciao 2736_src_newhr.ciao expression="NEW_HR=2.5*log(o.src_cnts_aper_s/o.src_cnts_aper_h)"

ciao% chips

chips> add_curve("2736_src_data.ciao[cols new_hr,luminosity_aper_b]") 

chips> set_curve(["line.style","none"])

chips> set_plot_xlabel ("2.5log([0.5-1.2keV]/[2.0-7.0keV])")
chips> set_plot_ylabel ("L_x(0.5-7.0), ergs s^{-1}")

The X-ray CMD is show in Figure 11.

Figure 11. X-ray color-magnitude diagram displaying CSC flux data for 47 Tucanae sources

The sources brighter than L_x ~ 10³¹ ergs s^-1 in the diagram appear to coincide with previously identified qMLXBs, CVs, and ABs in 47 Tucanae; the region below L_x ~ 10³¹ ergs s^-1 possibly includes the lower temperature CVs, ABs, and MSPs found in these studies, as well as background sources which are likely AGN. We can examine the CSC power law and black body model spectral fits for these sources with Sherpa, to see if the suggested physical emission mechanism reinforces the photometric classification.

---

Fitting CSC source spectra with Sherpa

We can download the PHA spectra and the ARF and RMF instrument response files for our sources, as illustrated in the section "Downloading CSC Data Products", as well as in the thread Retrieving Data Products. As an example, we download the spectrum files of the third brightest source in our list, which in our X-ray CMD appears to match a known qLMXB candidate in 47 Tucanae. (Note that the Level-3 spectrum files are compatible with Sherpa versions 4.1 and higher.)

To access the CSC spectral properties of this source, we return to the Query tab and modify our last query as follows: 1) we change the sort order of the result set by replacing 'match_type' with 'flux_aper_b' in the Sort Order window (in 'descending' order); 2) we add the master source black body and power law spectral fit properties to the result set list, which can be found in the "Model Spectral Fits" section of the master source properties list; and 3) we select "File-->Search" to access the new set of results; the modified query form is shown in Figure 12. In the Results tab, we select the third source from the top of the list of results, as well as the "Spectrum" files from the list of data products, and then "File-->Search". We then download the selected files from the Products tab and unpack the resulting tar file.

Figure 12. CSCview query which accesses source spectral properties

Each Chandra Level=3 PHA file (pha3.fits) available in the Chandra Source Catalog contains both a source and background spectrum in separate FITS HDUs (CXC Data Model blocks). When a pha3.fits file is loaded into Sherpa with load_data or load_pha, the background spectrum is automatically recognized and read in as well, with the same filename as the source spectrum. The resulting source and background Sherpa data sets may be handled in the usual way in Sherpa.

First, we use Sherpa to separately fit an absorbed 1-D power law and black body model to the selected 47 Tucanae source, using the best-fit power law and black body model parameters recorded in the CSC for this source. The CSC black body model spectral fits are performed over the energy range 0.5-7.0 keV, where the free parameters fitted are the total integrated flux, total neutral Hydrogen absorbing column density, and black body temperature. The power law model spectral fits are also performed over the energy range 0.5-7.0 keV; the free parameters fitted are the total integrated flux, total neutral Hydrogen absorbing column density, and power law photon index.

unix% sherpa
-----------------------------------------------------
Welcome to Sherpa: CXC's Modeling and Fitting Package
-----------------------------------------------------
CIAO 4.2 Sherpa version 2 Tuesday, July 6, 2010

sherpa> load_pha("acisf02736_000N001_r0038_pha3.fits.gz")
read ARF file acisf02736_000N001_r0038_arf3.fits
read RMF file acisf02736_000N001_r0038_rmf3.fits
read background file acisf02736_000N001_r0038_pha3.fits

sherpa> calc_data_sum()
          935.0

sherpa> subtract()
sherpa> group_counts(1)  #require minimum of 1 count per bin
sherpa> ignore()
sherpa> notice(0.5, 7.0)

sherpa> set_model(xsphabs.abs1*powlaw1d.p1)
sherpa> abs1.nh = .3389
sherpa> abs1.nh.max = .3622
sherpa> abs1.nh.min = .3163

sherpa> p1.gamma = 3.92 
sherpa> p1.gamma.max = 4.04
sherpa> p1.gamma.min = 3.81

sherpa> show_model()
Model: 1
apply_rmf(apply_arf((58519.8509903 * (xsphabs.abs1 * powlaw1d.p1))))
   Param        Type          Value          Min          Max      Units
   -----        ----          -----          ---          ---      -----
   abs1.nh      thawed       0.3389       0.3163       0.3622 10^22 atoms / cm^2
   p1.gamma     thawed         3.92         3.81         4.04           
   p1.ref       frozen            1 -3.40282e+38  3.40282e+38           
   p1.ampl      thawed            1            0  3.40282e+38    

sherpa> set_method("neldermead")
sherpa> set_stat("chi2xspecvar")

sherpa> fit()
 Solar Abundance Vector set to angr:  Anders E. & Grevesse N. Geochimica et Cosmochimica Acta 53, 197 (1989)
 Cross Section Table set to bcmc:  Balucinska-Church and McCammon, 1998
Dataset               = 1
Method                = neldermead
Statistic             = chi2xspecvar
Initial fit statistic = 3.56813e+11
Final fit statistic   = 291.378 at function evaluation 608
Data points           = 164
Degrees of freedom    = 161
Probability [Q-value] = 1.42318e-09
Reduced statistic     = 1.8098
Change in statistic   = 3.56813e+11
   abs1.nh        0.3622      
   p1.gamma       3.81        
   p1.ampl        4.51114e-05 

sherpa> plot_fit_delchi()

We find that the source is not well-represented by an absorbed power law, as indicated by the reduced χ² statistic of the fit and the plot of the fit Δχ (the residuals divided by the uncertainties):

Figure 13. Plot of a poor power law model fit to the CSC source

Next, we try the absorbed black body model fit, again using the best-fit CSC parameters for this model and source:

sherpa> load_pha(2, "acisf02736_000N001_r0038_pha3.fits.gz")
sherpa> subtract(2)
sherpa> group_counts(2, 1)  #require minimum of 1 count per bin
sherpa> ignore()
sherpa> notice(0.5, 7.0)

sherpa> set_model(2, abs1*xsbbody.b1)
sherpa> abs1.nh = 1e-07
sherpa> abs1.nh.max = 0.001677
sherpa> b1.kT = .2963
sherpa> b1.kT.max = .3009
sherpa> b1.kT.min = .2918

sherpa> show_model(2)
Model: 2
apply_rmf(apply_arf((58519.8509903 * (xsphabs.abs1 * xsbbody.b1))))
   Param        Type          Value          Min          Max      Units
   -----        ----          -----          ---          ---      -----
   abs1.nh      thawed        1e-07            0     0.001677 10^22 atoms / cm^2
   b1.kt        thawed       0.2963       0.2918       0.3009        keV
   b1.norm      thawed            1            0        1e+24 L39 / (D10)**2

sherpa> fit(2)
Dataset               = 2
Method                = neldermead
Statistic             = chi2xspecvar
Initial fit statistic = 2.66444e+15
Final fit statistic   = 315.675 at function evaluation 567
Data points           = 164
Degrees of freedom    = 161
Probability [Q-value] = 4.03846e-12
Reduced statistic     = 1.96071
Change in statistic   = 2.66444e+15
   abs1.nh        0.001677    
   b1.kt          0.3009      
   b1.norm        4.73533e-07 

sherpa> plot_fit_delchi(2)

Figure 14. Plot of a poor black body model fit to a source in 47 Tucanae

The source is also poorly fit by an absorbed black body model. After some more experimentation, we try including a hydrogen atmosphere model component in the fit:

sherpa> clean()
sherpa> load_pha(3, "acisf02736_000N001_r0038_pha3.fits.gz")
sherpa> subtract(3)
sherpa> group_counts(3, 1)  #require minimum of 1 count per bin
sherpa> ignore()
sherpa> notice(0.5, 7.0)

sherpa> set_model(3, xsphabs.abs1*(powlaw1d.p1+xsnsagrav.ha1))
sherpa> guess(3, ha1)
sherpa> guess(3, p1)

sherpa> show_model(3)
Model: 3
apply_rmf(apply_arf((58519.8509903 * (xsphabs.abs1 * (powlaw1d.p1 + xsnsagrav.ha1)))))
   Param        Type          Value          Min          Max      Units
   -----        ----          -----          ---          ---      -----
   abs1.nh      thawed            1            0       100000 10^22 atoms / cm^2
   p1.gamma     thawed            1          -10           10           
   p1.ref       frozen            1 -3.40282e+38  3.40282e+38           
   p1.ampl      thawed  2.20766e-05  2.20766e-07   0.00220766           
   ha1.logt_eff thawed            6          5.5          6.5          K
   ha1.nsmass   thawed          1.4          0.3          2.5       Msun
   ha1.nsrad    thawed           10            6           20         km
   ha1.norm     thawed  0.000331772  3.31772e-07     0.331772           


sherpa> set_method("neldermead")
sherpa> set_stat("chi2xspecvar")

sherpa> fit(3)
Dataset               = 3
Method                = neldermead
Statistic             = chi2xspecvar
Initial fit statistic = 1.2724e+09
Final fit statistic   = 186.465 at function evaluation 1896
Data points           = 184
Degrees of freedom    = 177
Probability [Q-value] = 0.298186
Reduced statistic     = 1.05347
Change in statistic   = 1.2724e+09
   abs1.nh        0.784476    
   p1.gamma       4.71802     
   p1.ampl        0.000160586 
   ha1.logt_eff   5.50093     
   ha1.nsmass     2.46324     
   ha1.nsrad      10.8931     
   ha1.norm       3.38319e-07 


sherpa> plot_fit_delchi(3)
sherpa> plot("fit", 3, "ratio", 3)

Figure 15. Better fit to the source spectrum with a hydrogen atmosphere plus power law model

The fit residuals and data-to-model ratio are shown in Figure 15. Clearly, fitting a hydrogen atmosphere model plus a power law produces a nicer fit - at least at the soft end of the spectrum - which is how one might expect to model a candidate qLMXB.