GALEX PII Paper

From Tauwiki

Jump to: navigation, search

Contents

[edit] Introduction

Studies of the diffuse ultraviolet sky have been an important part of interstellar dust studies over the last four decades (Bowyer, 1991; Henry, 1991; Murthy, 2009) but were limited by the difficulty of observing faint diffuse sources near the limit of the instrumental sensitivity. It has been generally agreed that the low and mid-latitude diffuse radiation is due to the scattering of starlight by interstellar dust but with a baseline at high galactic latitudes which was variously attributed to either high latitude dust (Bowyer 1991) or to an extragalactic source (Henry 200x).

Just as the Infrared Astronomy Satellite (IRAS) revolutionized the study of the diffuse infrared emission (Low et al.), data from the Galaxy Evolution Explorer (GALEX) has the potential to change our view of the diffuse ultraviolet emission. Over the last 6 years, GALEX has amassed many deep observations of the ultraviolet sky with a spatial resolution (for the faint diffuse emission) comparable to the infrared data. We have embarked on a survey of all GALEX deep observations (> 2000 s) to map the diffuse UV radiation field over the sky beginning with observations of a region of nebulosity first observed by Sandage (1976) in the vicinity of M82. This region had a comparatively high optical depth and we found a flat UV emission (Sujatha et al. 2009) despite the IRAS 100 micron emission increasing by a factor of 2.

In this work, we examine a set of observations of a region in Draco where the optical depth is much lower but where there is a ridge of dust emission extending through the field. As with the Sandage observations of Sujatha et al. (2009), this field is at high Galactic latitude (34 - 38 degrees) but is about 60 degrees away. The data are from the GALEX Deep Imaging Survey (DIS) and overlap with the Spitzer first look extragalactic survey.

[edit] Observations and Data Analysis

The GALEX spacecraft was launched in 2003 as part of NASA's Small Explorer (SMEX) program with a primary science objective of observing star formation in galaxies at low redshifts (Martin et al. 2005). Light from the sky is collected through a single 50 cm telescope and then separated into two bands (far ultraviolet: 1350 - 1750 A; and near ultraviolet: 1750 - 2850 A) using a dichroic mirror. Independent low noise delay-line detectors record every photon with an overall effective spatial resolution of 5 -- 7" in the sky over a 1.25 degree field. The data products from the mission include, amongst other files, FITS images of the FUV and NUV fields and a list of point sources in each field. A complete description of the data processing, the calibration and the data products may be found in Morrisey et al. (2007).

This work follows on our observations of diffuse emission in a region near M82 (Sujatha et al. 2009) and focuses on a set of 11 observations covering an area of almost 3 square degrees in the constellation of Draco with cumulative exposure times of 5,000 to 50,000 seconds (Table obs_log). These observations were taken by the GALEX team as part of a program to map the Space Infrared Telescope Facility (SIRTF: now the Spitzer Space Telescope) first light locations - hence the target name of "SIRTFFL". This region (Fig. IR_img) contains the high velocity cloud (HVC) Complex C (Miville-Deschenes et al.) at a distance of more than 800 pc which was the first high velocity cloud in which dust was detected. Much closer to us is LVC 88+36-2, seen as a ridge in the IR emission, which is only 60 pc away (Lillienthal et al.) from the Sun and, most significantly, was the first cloud discovered to cast an shadow in the X-ray background (Burrows & Meadenhall). Because of the then upcoming Spitzer observations, Lockman & Condon mapped the region in the 21 cm line of HI finding several components (Table xx). This wealth of detail has proven invaluable in our understanding of the UV observations.

Each observation was comprised of a number of visits spread over a period of weeks, or even months, all of which were coadded by the standard GALEX pipeline (Morrissey et al.) to produce a single image in each of the two bands. Point sources in each image were extracted using a standard point source extractor (Sextractor - [ref]) and a merged point source catalog was created. We note here that the exposure time in the FUV detector was often significantly less than that in the NUV because of power supply problems on the spacecraft. Our processing uses the FITS image files and the merged point source catalog from the GALEX pipeline and has been described by Sujatha et al. (2009). These image files have been fully calibrated and flat fielded but not background subtracted. Although the GALEX program does provide files containing the background in each observation, these were made by fitting a multi-dimensional surface to the image and therefore show structure related to the pinning points of the surface. While adequate for their intended purpose of subtracting the background from point sources in the field, they introduce large scale artifacts which make them unsuitable for the study of the diffuse radiation field.

We therefore created our own background files for each observation by blanking out each of the point sources in the merged GALEX point source catalog and binning the observation into 2' pixels (80 x 80 GALEX pixels). We also restricted our analysis to the central 1.15 degrees of the 1.25 degree field of view because the edges were dominated by instrumental artifacts, both from the detector and scattering from off-axis stars. The total number of pixels rejected was about 20% of the total in each field. The remaining signal was comprised of the foreground emission (instrumental dark count, airglow and zodiacal light) and the astrophysical signal (atomic and molecular emission, dust scattered starlight, and any extragalactic contribution).

[edit] Results

[edit] Foreground Emission

A large field of view imager such as GALEX has distinct advantages in observations of the diffuse background in that stars can be easily rejected and, anyway, contribute only a small percentage of the total signal in the field (less than xx% in our observations). However, without spectra, it is not possible to directly separate the different components of the diffuse radiation field and we have to infer their contributions. Instrumental dark count is negligible, contributing less than 5 photon units in either band (Morrissey et al) but airglow, primarily due to the OI lines at 1356 A and 2471 A, is expected to contribute about 200 photon units to either band (Brune et al.; Boffi et al. 2007). The airglow is the only one of the components of the diffuse radiation field which could vary over a single visit and we have used the TEC (Total Event Counter) to track this as a function of time from local midnight (Fig. TEC_time). A baseline has been subtracted from each visit so that the total count rate is 0 at local midnight. The airglow is surprisingly well fit with a quadratic as a function of time from local midnight.

The baseline levels, which represent the integrated emission at local midnight including residual airglow, are strongly correlated with the 10.7 cm solar flux (ref) which is used as a proxy for solar-terrestrial interactions (Chatterjee ref). We have subtracted an offset from each set of visits, corresponding to the y-intercept and representing the astrophysical signal in an observation, and plotted the remainder in Fig. solar_flux. Note that the NUV data have had the zodiacal light subtracted from each visit. Putting Fig. TEC_time and Fig. solar_flux together, we find the following equations for the airglow in the FUV and NUV, respectively, with an uncertainty of about 50 photon units: 

FUV_AG = 3.4*SF + 25*t2 + 12*t ------ (4)
NUV_AG = 3.7*SF + 16*t2 + 5.9*t ------ (5)

Our interest in the airglow is only to subtract it from the astrophysical background and we limit ourselves to saying that this emission is consistent with solar photons resonantly scattering from geocoronal oxygen atoms (personal communication: L. J. Paxton)

The remaining foreground contributor, zodiacal light, is important only in the NUV band because of the rapidly fading solar spectrum at wavelengths of less than 2000 A. There is no UV map of the zodiacal light but we have used the optical tabulation of Leinert et al. (1998) with grey scattering to predict the zodiacal light in each visit. We have tabulated the average contribution of the three foreground components --- the dark count, airglow, and the zodiacal light --- in Table xx. The foreground emission ranges from 20% to 50% of the total emission in any field and we can estimate its level to about 50 photon units. It is important to note that this is only an uncertainty in the offset; the foreground emission is uniform across the field and will not affect the spatial variability of the diffuse radiation field.

[edit] Sources of Error

One of the greatest problems affecting our understanding of the diffuse radiation field has been in characterizing the errors in the observations. The large number of observations in this data set and the spatial overlap between completely different observations has allowed us to explore these errors, at least as they relate to GALEX data. The first of these is the difficulty of setting the level of the foreground emission. We have compared the actual offset between different observations with some degree of spatial overlap with that calculated using our empirical models for the airglow and zodiacal light (Fig. overlap_plot) and have an agreement to within about 30 photon units. It should be reiterated that this is an uncertainty in the absolute level of the astrophysical emission and not in the spatial variability.

More interesting is the scatter in the data. For a photon counting instrument such as GALEX, the instrumental scatter would be only due to photon noise or to errors in the flat fielding of the instrument. We have empirically derived the instrumental scatter by dividing each observation into two sets of visits, which may well be separated by several months, and finding the scatter between them. We have plotted the actual scatter against the calculated photon scatter in Fig. scat_plot finding excellent agreement between the two. As an independent test, we also took the overlap regions between different observations and calculated the scatter between them. These give slightly higherare consistent with It was also possible to test the scatter using the overlap regions, although there were fewer points which may have been affected by edge effects. Although, the scatter for the overlap regions is somewhat higher than the calculated values, this is likely due to the many fewer points in the overlap regions and their location near the edge of the detector. It should be noted that these comparisons are in sky coordinates because there are arbitrary roll angle differences between the different visits.

[edit] Discussion

Our basic result is the FUV and NUV images of the SIRF First Look field and is shown in Fig. diffuse_image at a spatial resolution of 2 arcminutes. These may be compared with the 21 cm map of Lockman and Condon (2005) and the IRAS 100 micron map (Fig. radio_ir). This particular field has been of some interest and was chosen for the SIRTF (now Spitzer) first look observations - hence the target name of SIRTFFL. The Draco Nebula is just out of the field but our observations include both the high velocity cloud (HVC) Complex C (Miville-Deschenes et al. ApJ 631, L57 2005) at a distance of more than 800 pc as well as the nearby cloud LVC 88+36-2. This latter cloud is known to cast a strong X-ray shadow (Burrows & Meadenhall) and is only 60 pc away (Lillienthal et al.) from the Sun. We have identified the different components of the ISM in this field in Table xx.

There are several possible contributors to the astrophysical UV emission. The greatest part is dust scattered starlight which will contribute to both the FUV and the NUV bands. This is reflected in the good correlation between the FUV and NUV bands (r = xx) and further in the good correlation between the IRAS 100 micron fluxes and the UV radiation (Fig. FUV_IR and NUV_IR). This should be contrasted with the essentially flat UV-IR curves obtained by Sujatha et al. (2009) around the Sandage nebulosity. The IR emission is due to thermal radiation from an optically thin layer of dust, as the cross-section of the grains is low in the IR. On the other hand, the cross-section of the grains is much higher in the UV and the optical depth transitions from being optically thin in these Draco observations to being optically thick in the Sandage region. A different way of looking at this is seen in Fig. uv-ir where we have plotted the ratio between the UV bands and the IR. There is a clear trend from the low optical depth (low 100 micron intensity) Draco region to the high optical depth in the UV (but still low in the IR) Sandage region.

It is interesting that the UV/IR ratio in our GALEX data follows a continuous curve very similar to that found by Murthy et al. (2001) in Orion using data from the Midcourse Space Experiment (MSX) even though the environment is very different. Both the UV and the IR fluxes in Orion were greater by a factor of about 200, reflecting the intense radiation field there. However, these ratios are quite different in other determinations (references) with values ranging from near 0 to almost 500 with little dependence on the IR. It is likely that these relations are only apparent when observed at a high enough spatial resolution; the MSX data were at a resolution of 20" and these data are at a resolution of 2' while the other observations are at resolutions of 0.5 degrees or worse. Because both the IR and the UV vary on smaller scales, the measured UV/IR ratio may not be a reliable estimator of the true ratio. In fact, Sasseen et al. found a total UV/IR ratio when they calculated the slope using all their data, higher than any of the individual data sets.

Readily apparent in Fig. UV_IR_ratio is the emission from a ridge of gas (LVC 88_36-2) running through our field where the FUV/IR ratio is greater than in other areas, although without a corresponding enhancement in the NUV/IR ratio. Indeed, this reflects a general increase in the FUV/NUV ratio with the FUV surface brightness (Fig. ratio_fuv) seen both here and in the Sandage region. The most likely explanation for this is is that there is an additional component in the FUV band which is not seen in the NUV.unlikely to be due to dust scattering alone in these two nearby bands and we expect variations of no more than xx% in the ratio (see below). Rather, it is likely to be due to an additional component in the FUV which does not appear in the NUV. Sujatha et al. (2009) attributed this to emission from the Werner band of molecular hydrogen, a reasonable assumption in the Sandage region where Martin et al. had already observed widespread H2 fluorescent emission.

If we make the assumption that the FUV/NUV ratio for dust scattering alone does not vary across the field, we can estimate the amount of excess emission by setting the ratio to the minimum observed in the field (0.7) and subtracting the scaled NUV from the FUV data (Fig. excess_emission). The ridge is prominent in this image and the level of the emission is correlated with the low velocity HI emission (Fig. ratio_hi) where the HI emission is from the survey of Lockman & Condon (200x). We have calculated the amount of H2 fluorescent emission expected using eq. xx of Martin et al. finding good agreement with the observed values in the ridge (Fig. H2_ridge) but with too low a prediction near the center of the cloud, where more emission is observed. We obtain a significantly better fit if we allow R (the H2 formation rate) to increase toward the cloud center, also plotted in Fig. H2_ridge. We have found that R varies from xx near the edge of the cloud to xx near the center. Other determinations are ...

The situation is much less clear outside of the ridge. We have not been able to convincingly fit the data with molecular hydrogen emission and there are other possibilities which we cannot distinguish with our imaging only data. For instance, Park et al. have observed atomic emission lines of both Si II (1533 A) and C IV (1550 A) around the nearby Draco molecular cloud which would contribute an effective level of about 50 photon units in the FUV band. Even a low resolution spectral image of this region would help us to separate the different components.

Until the advent of GALEX data, the quality of the data have not been sufficient to warrant sophisticated models which incorporate many hidden assumptions. However, we are now finding that the simple models we have used in the past (Murthy & Henry 1995; Sujatha et al. 2008; Sujatha et al. 2009) are no longer sufficient to adequately fit the data, particularly at the high spatial resolution of the present observations. This model has been described fully by Sujatha et al. and uses Kurucz models for the stars in the Hipparcos catalog for the interstellar radiation field. This radiation is then scattered from dust in the line of sight, taking into account self-extinction. Because the optical depth is low in these observations, we have used a single scattering model which provides a reasonable fit to the data (Fig. model_fit).

The primary limits (Fig. ag_contour) on the optical constants of the grains come from the NUV channel which is purely due to dust scattered radiation. We find 3 sigma limits of 0.24 < a < 0.32 and 0 < g < 0.45 for the optical constants of the dust grains in the NUV band but were not able to determine the optical constants independently for the FUV channel because of the excess emission. Draine has predicted a=0.48; g = 0.54 in the NUV band and a = 0.39; g = 0.66 in the FUV band for a mixture of spherical graphite and silicate grains. Although our limits are not formally consistent with these values, there is enough uncertainty in our model that minor changes in the model will make it consistent.

More important is to note that the scatter in the data is much larger than the photon noise. We have empirically derived an 1 sigma error bar of about 40 photon units in the model fit to the data which is much greater than the photon noise, reflecting the incompleteness of the model.

[edit] References

(1) Sandage, A. 1976, AJ, 81, 954

(2) Sujatha et al. (2009)

(3) Martin, D. C. et al., 2005, ApJL, 619, L1

(4) Morrissey, P. et al. 2007, ApJS, 173, 682

(5) Boffi, F. R., et al. 2007

[edit] Figures

Fig.2. GALEX Images of Observed Locations.
Fig.2. GALEX Images of Observed Locations.
Fig.iras_obs. IRAS 100 micron map of the region with GALEX FOV's.
Fig.iras_obs. IRAS 100 micron map of the region with GALEX FOV's.
Fig.TEC_timea. Total FUV count rate as a function of local time.
Fig.TEC_timea. Total FUV count rate as a function of local time.
Fig.TEC_timeb. Total NUV count rate as a function of local time.
Fig.TEC_timeb. Total NUV count rate as a function of local time.
Fig.solar_fluxa. FUV AG_C level versus solar flux at the Earth.
Fig.solar_fluxa. FUV AG_C level versus solar flux at the Earth.
Fig.solar_fluxb. NUV AG_C level versus solar flux at the Earth.
Fig.solar_fluxb. NUV AG_C level versus solar flux at the Earth.
Fig.scatter_plot. Intrinsic scatter versus predicted level.
Fig.scatter_plot. Intrinsic scatter versus predicted level.

[edit] Tables

Table.obs_log - Details of the Observations
Tile_nameRA_CentDec_CentglcgbcNuv_exptimeFuv_exptimeStart_TimeEnd_TimeNuv_VisitsFuv_Visits
SIRTFFL_00259.11 59.91 88.8435.0552917.15 52016.95 03/07/03 25/08/08 41 39
SIRTFFL_01260.41 59.34 88.0834.44 26006.1 30922.1 04/07/03 26/07/04 20 24
SIRTFFL_02260.09 58.5 87.0834.66 39037.05 26859.35 18/08/03 02/09/07 30 25
SIRTFFL_03258.33 58.86 87.6135.55 39830.4 29570.9 19/08/03 01/09/07 30 28
SIRTFFL_04256.98 59.72 88.7636.13 3874.45 3874.45 01/05/04 01/05/04 3 3
SIRTFFL_05 260.68 60.7 89.7134.2 5305 5305 01/05/04 03/05/04 4 4
SIRTFFL_06257.58 60.45 89.635.74 27658.75 3668 25/07/05 07/04/08 22 9
SIRTFFL_07260.54 60.81 89.8534.26 34376.55 21276.1 27/07/05 02/09/07 25 20
SIRTFFL_08262.61 59.15 87.7833.33 40639.6 22540 19/06/05 01/09/07 28 20
SIRTFFL_09 257.2 59.72 88.7436.02 15737.4 3197.5 29/07/05 07/04/08 12 9
SIRTFFL_10 256.99 58.8 87.6336.24 27383.75 10757.7 25/07/05 28/08/07 21 15
SANDAGE 142.04 70.386 142.338.2 35210 14821 11/01/05 04/01/07 10 22
Table.1 - Airglow & Zodiacal Contribution in each Observation
Tile_NameMinimum Solar Flux (SF_MIN)FUV AG (pu) = 3.4*SF_MIN + AG_VNUV AG (pu) = 3.7*SF_MIN + AG_V λ − λO βZL (pu) NUV_AG + ZL (pu)
SIRTFFL_006530129784.17-12881.76367 664
SIRTFFL_018034936090.67-193.581.59407 767
SIRTFFL_027033231577-206.5480.76373 688
SIRTFFL_036931830871.46-19980.67382 690
SIRTFFL_0475313320182.62-183.281.01342 662
SIRTFFL_0590362376187.47-190.282.84342 718
SIRTFFL_067030630492-195.6581.78358 662
SIRTFFL_077033031465-201.282.9381 695
SIRTFFL_086932730686.3-163.181.87365 671
SIRTFFL_097733333392.5-144.0481.08367 700
SIRTFFL_107030330572.65-147.5180.24365 670
SANDAGE8139138485.61-176.251.4440 824
Retrieved from "http://tauvex.iiap.res.in/tauwiki/index.php/GALEX_PII_Paper"
CSS 2.1 Xhtml 1.0 Last Modified: February 22, 2007 GooglePagerank