It’s Not All Bad News

The good news is that the GALEX spectra are free of cosmic rays, since the detectors are’t sensitive to cosmic rays, and artifacts from bad pixels, since the detectors don’t have pixels and, more importantly, each point in the spectrum is averaged over many locations on the detector.

The bad news is that the extensive masking of overlapping objects can leave a number of deceptive artifacts in the spectra. The key message of this page is that almost all of these artifacts, while often difficult to identify as artifacts in the 1D spectra, reveal their true nature in the 2D spectral strips. Therefore, I strongly recommend that you look at the 1D spectra and 2D spectral strips together by, for example, using the GALEXSpec code.

Problems Not Due to Masking

As you look through GALEX spectra, you will notice that most of the spectra have non-physical (to say the least) continuum slopes at the edges (within 100 - 150 Å) of the bands. This is generally due to slight misalignment of the response function, which is very steeply falling at the band edges, and the spectrum. The spectrum and background apertures are fixed (this may change with updated versions of the spectroscopy pipeline). For bright extended objects, the spectrum extraction aperture is too small, and so the flux recovered in the spectra will be less than the true flux (as measured by the photometry). Finally, the sky subtracted at each pixel is just the average of the sky values in the background apertures on either side of the spectrum, not a fit. If there are overlapping spectra in the background apertures that have not been properly masked, these overlapping spectra will be erroneously included in the estimated background.

Problems Due to Masking

Especially in the NUV where the source density is higher, the GALEX spectra suffer from contamination from nearby overlapping spectra. Most GALEX spectra (the exception being some Guest Investigator observations) are obtained with at least 10 grism angles in order to disentangle the overlapping spectra. The spectroscopy pipeline masks regions where there is overlap and throws away all photons in a masked region. The masking is not perfect, however, leading to a number of artifacts in the affected spectra, some of which are difficult to distinguish from real features in the 1D spectra. To reiterate the key point, while some of the artifacts do look like genuine features in the 1D spectra, they are easy to identify as artifacts in the 2D spectra.

The figures below illustrate some of the artifacts that are found in GALEX spectra. The red, green, and blue numbers along the top of the 2D spectral strips mark the 1st, 2nd, and 3rd spectral order wavelengths, respectively.

spectrum

NUV strip

Figure 1. There is a strong absorption feature at ~2500 Å (marked with the blue arrow) in the 1D spectrum. However, a look at the 2D spectral strip shows that the feature is in fact an artifact caused by a masking error.

Spectrum

Strips

Profiles

Figure 2. The 1D spectrum (top panel) has clearly had too much background subtracted. In the spectral strips (middle panel), one can see that a bright source (marked with a barely visible orange arrow) has not been properly masked. The spatial profiles are shown in the bottom panel, where the blue line is the median profile from the bluest third of the strip, the green line is the median profile from the middle third, and the red line is the median profile from the reddest third.

Spectrum

NUV strip

Figure 3. The blue arrow marks a broad feature in the 1D spectrum (top panel) which could conceivably be a quasar emission line. However, inspection of the NUV spectral strip (bottom panel) reveals that the flux responsible for the feature in the 1D spectrum lies slightly above the object spectrum and is an artifact.

One Visit Spectra

Since the photons in overlap regions are simply discarded, spectra from one (or a few visits) will often have “drop outs” in the masked regions where there are no data. With more visits, such “drop outs” become much rarer as wavelength ranges masked at one particular grism angle are filled with data from other grism angles. Stated more plainly, data with only one or a few visits (e.g., guest investigator spectra) will often look much worse than the many-visit data collected for the GALEX surveys.

Where to Get Help

If you need help deciding whether or not a particular feature is an artifact, please feel free to contact me. But, remember to look at the 2D spectral strip first!