NOAO Extremely Wide-Field Infrared Imager
Hα nebular emission is one of the most direct tracers of star formation. As such, a great deal of our current understanding of star formation in local galaxies is based on a long and rich history of Hα observational studies. However, there is a dearth of analogous work in the intermediate redshift universe - a critical period in galaxy evolution where the overall star formation activity reaches its maximum - because the line is shifted out of the optical window past z~0.4. Thus, we have begun a campaign to extend deep Hα galaxy surveys to earlier cosmic times by taking advantage of the new capabilities offered by the NOAO Extremely Wide-Field Infrared Imager (NEWFIRM) at the KPNO 4m. Our strategy is to obtain (1%) narrowband imaging of key extragalactic fields in low-OH airglow windows at 1.19 and 2.09 μm (Hα @ z~0.8 and 2.2 respectively). Coupled with IMACS at the Baade 6.5m, which is well-matched for obtaining optical spectroscopy for emission-line galaxy candidates identified in the NEWFIRM narrowband imaging, the data from this survey will enable a wide range of investigations including:
NEWFIRM makes detections by taking images with a narrow-band (at 1187 nm) and J-band filter. The J-band flux (which represents the continuum), scaled to account for the differences in broadband/narrowband filter widths, is subtracted from the narrow-band flux (which represents the emission from Hα at z=0.81 and the continuum), which results in an estimate of Hα emission at z=0.81. Hα is a tracer for star formation rate and since lookback time is a function of redshift we can use data from NEWFIRM to determine the star formation rate for the z=0.81 epoch of the universe. Unfortunately NEWFIRM cannot detect every Hα source so the data from NEWFIRM alone will not result in an accurate star formation history of the universe. To compensate for this, galaxies are simulated with known parameters and then injected into mosaics of science data from NEWFIRM. Then the sources are extracted using the normal method and a ratio of simulated sources to detected sources is found. This is called the incompleteness of the data and can be used to find an accurate count of detected galaxies at different luminosities.
Parameters used to simulate galaxies:
1. Luminosity Bins – [33.4, 33.6, 33.8, 34.0, 34.2, 34.4 ,34.6, 34.8 ,35.0, 35.2, 35.4, 35.6, 35.8, 36.0] in log(Hα/W) the simulations are done once for each bin (the calibration process is described here)
2. X and Y position on the chip – between 50 and 5,350 pixels for both X and Y on the 5,400 by 5,400 chip assigned by a uniform distribution
3. Equivalent Width – log(EW/Å)=1.17, 1.52, and 2.00 assigned from Lee et al. (2007) with standard deviations also from Lee et al. (2007)
3. N[II] Ratio – The Hα line is brightened to include the N[II] lines according to a distribution from Villar et al. (2008) that assigns N[II] by equivalent width
4. Position Angle – 0-360° assigned by a uniform distribution
5. Axial Ratio – 0.15 to 1 (we estimate 0.15 to be the minimum because of the intrinsic "fatness" of the disk) assigned by a uniform distribution
6. Half Light Radius R0.5 – as defined by the equation -ln(0.5)*( ρ + randn*[ρ /3]) where the disk scale length ρ is 1.18 for z=0.81, the redshift our filter is calibrated to detect, and randn is a random number from a normal distribution with a standard deviation of 1 and a mean of 0
7. Extinction – Extinction is done using the Hopkins et al. 2001 prescription
8. The number of sources simulated – 1,000 sources per image (any more and overlap dominates the statistics)
9. The Distance Thresholds – Coordinate matching is set to 3 pixels or 1.2 arcseconds for these simulations. Also, any source found within 1 pixel of a source detected in the original mosaic is excluded
10. The Signal to Noise Threshold – Set to 3 for these simulations
Parameters used to SExtract galaxies:
1. Detection Threshold – 2.0σ
2. Number of Contiguous Pixels – 5 pixels
After all of these parameters have been determined galaxies are added to a base image by the IRAF task mkobjects. The base image is a mosaic of science data collected by NEWFIRM. The noise of NEWFIRM is dependent on X and Y position on the chip so it is impossible to accurately simulate the noise of NEWFIRM with IRAF and a single noise value. This is unfortunate since this means there is a chance that a simulated galaxy will overlap with an actual galaxy, what we call uncertainty due to overlap (which may also occur if two or more simulated sources overlap as well). To minimize overlap the number of sources per simulation is limited. After the simulated galaxies have been created Source Extractor is run exactly as if the image were a normal science image. Any source which lies inside a mask made up of bad sections of the detector and stars is excluded. The simulated galaxy counts are then compared with the counts from SExtractor. Sources found by SExtractor which are not within a certain distance threshold of the simulated position are excluded. SExtractor is also run on the mosaic image and any galaxies in the image with simulations that are within one pixel of a galaxy found in the mosaic image are considered to be real galaxies and are excluded. Simulated sources which do not have SExtracted sources within the distance threshold are considered non-detections. If multiple SExtracted sources lie within the distance threshold, then the source with the closest magnitude to the simulated source is considered the match. The signal to noise ratio for every matched source is then found by converting counts to flux and then using the following equations:
Parameters used to calculate the final completeness:
1. Mask Files – Any simulated source within a mask file is eliminated.
2. Real Distance Threshold – Any simulated source within 1 pixel of a source detected by sextractor from the original image is eliminated.
3. Sextractor Distance Threshold – Any simulated source that has no sExtractor detection is considered a non-detection.
4. Signal to Noise Threshold – Any sExtracted source with a S/N less than 3 is considered a non-detection.
The completeness is then detections/(detections+non-detections). We simulate 250 galaxies/simulation/quadrant which ends up around 150 galaxies/simulation/quadrant due to the real distance threshold and mask files. We run the simulation 30 times so in the end there is ~4500 galaxies/field/quadrant.
|log(Hα/W)||% for log(EW/Å)=0-2||% for log(EW/Å)=1-2|
|33.4||0.6130 +/- 0.0008290||0.6680 +/- 0.001497|
|33.6||0.7720 +/- 0.001516||0.6950 +/- 0.0007420|
|33.8||0.9850 +/- 0.002119||0.7280 +/- 0.001108|
|34.0||1.413 +/- 0.002762||0.9400 +/- 0.001625|
|34.2||2.865 +/- 0.002810||1.535 +/- 0.002885|
|34.4||5.900 +/- 0.01060||3.920 +/- 0.005930|
|34.6||12.92 +/- 0.02025||10.46 +/- 0.01001|
|34.8||34.80 +/- 0.05200||30.67 +/- 0.01981|
|35.0||72.90 +/- 0.02920||71.70 +/- 0.04520|
|35.2||93.80 +/- 0.007580||92.90 +/- 0.01360|
|35.4||96.00 +/- 0.005750||96.00 +/- 0.008700|
|35.6||96.80 +/- 0.002745||96.30 +/- 0.006690|
|35.8||97.40 +/- 0.001501||97.30 +/- 0.003820|
|36.0||97.80 +/- 0.002142||97.90 +/- 0.003360|
Note: Percentages were found by averaging completness found by 25 simulations (we used 25 different random number seeds). The uncertainty is the standard deviation taken from the 25 simulations and has no other propagated uncertainty.
Danny and I have written a small section for an upcoming paper which can be seen here.
In order to test these results a number of plots were created:
1. Taken from the NEWFIRM wiki: http://newfirm.ociw.edu/wiki/Main_Page
Last Updated: April 13, 2010