The Millimeter Array Deep Field -- Yun

Min Yun's `very crude' simulation of the MMA Deep Field at 850 microns
V1. Min Yun's very crude simulation of the MMA Deep Field at 850 microns All this includes so far are the 5 HDF SCUBA sources plus 1656 model sources brighter than 30 microJy, which is 5 sigma upper limit after spending about 1 hr per primary beam -- about 200 hrs total -- covering 4' by 4' field. The sizes of the circles are logarithmic representation of their 850 micron brightness (radius = log S(mJy) + 3 -- we can come up with more sensible things later). The total number is about a factor 2 larger than Al's analysis because I am using the latest source count papers the Ivison and Co: N(>S) = 7900 S^{-1.1} per square degree. Oh, the circle in the middle is roughly the FOV of SCUBA, so you can compare directly with Hughes et al. The brighter sources in the circle are simulations of the actual sources seen by them.

Millimeter Array Deep Field Simulations -- Gallimore

Millimeter Array D-array 350 GHz continuum simulation plotted with a linear stretch. The color channel allocation corresponds to redshift following red: z > 3; green: 1.5 < z < 3; and blue: z < 1.5. The field size is 4.2 arcminutes. The assumed cosmology and evolution match the model Peak-5 of Blain et al. (1999) (astro-ph/9806062). For simplicity, the sources were taken to be unresolved. No noise is added in this simulation, but I imposed a flux cutoff of 0.01 mJy. The same simulation plotted with a logarithmic stretch to emphasize the number of detectable sources.
Same simulation as above, except that the sources are now distributed in randomly oriented 20~kpc diameter disks. The stretch is linear. Disk simulation displayed with a logarithmic stretch.
To do list:
  • More extensive testing of the algorithms wouldn't hurt.
  • Check the math on the cosmology.
  • Need to consider realistic source sizes and possible evolution with redshift?
  • Possibly allow the morphology of the sources to vary?
  • Add noise to images.
  • Vary the cosmology & evolution parameters.
Here's the source code:
cosmo1.f Performs all of the source counts as a function of redshift. All of the parameters are hard-wired right now, but they will be converted to run-time inputs eventually.
cosmoplot.f Generates FITS images based on the output of cosmo1. All of the sources are placed on the image as single pixels. I use AIPS tasks to convolve with a gaussian beam (CONVL) and to generate RGB images (TVRGB). Again, all of the parameters are hard-wired, but will be softened later. Compilation requires the fitsio library for FORTRAN and a suitable random number generator for the function ran1 (I used the Numerical Recipes routine).
cosmo2.f Like cosmo1, but also prints out the angular size distance for use in cosmoplot2.f.
cosmoplot2.f Like cosmoplot.f, but places randomly oriented disks of a fixed diameter around the image.
Millimeter Array 350 GHz continuum simulation plotted in pseudo-color with a linear stretch. The background rms is 0.01 mJy, the image resolution (beam) is 1.5 arcseconds, and the field size is 4.2 arcminutes. The assumed cosmology and evolution match the model Peak-5 of Blain et al. (1999) (astro-ph / 9806062). For illustration, the galaxies were simulated by 20 kpc diameter uniform disks. The same simulation, except that galaxies in redshift bins have been assigned different color channels: red: z > 3; green: 1.5 < z < 3; and blue: z < 1.5. The image stretch is linear, and the simulated noise appears as grey.
To do list:
  • More extensive testing of the algorithms wouldn't hurt.
  • Check the math on the cosmology.
  • Need to consider realistic source sizes and possible evolution with redshift?
  • Possibly allow the morphology of the sources to vary?
  • Vary the cosmology & evolution parameters.
The latest source code:
cosmo3.f Performs all of the source counts as a function of redshift. Inputs are through the file cosmo3.par. The output file, cosmo3.dat, can be used as input for cosmoplot3.
cosmoplot3.f Generates FITS images based on the output of cosmo3. Inputs are through the file cosmoplot3.par. This version of the code allows convolution with a circular Gaussian beam and the addition of Gaussian noise. Compilation requires the fitsio library for FORTRAN and the Numerical Recipes subroutines ran1, gasdev, rlft3, and fourn .
nsvs.f This code simply checks the source counts, N(>S) as a function of S, based on output from cosmo3. Requires the Numerical Recipes subroutine indexx.

Ways out:
NRAO Charlottesville Homepage
Jack's Homepage

Changed by: Jack Gallimore, 29-Jan-1999

Spectral Lines in the Millimeter Array Deep Field Simulations -- Wootten, Radford, Yun

The above simulations are medium deep mosaicked observations taking a few weeks at several hundred pointings. We have decided that best sensitivity is at 350 GHz. How would an array as specified in the Project Book observe this? First, the total coverage from edge to edge of the upper and lower sidebands is 24 GHz, so the pseudo continuum band should be centered in the window, which isn't much wider than this. The observations will be taken in spectral line mode but with coarse resolution covering all 8 GHz centered on 339 GHz in the LSB. In the USB, the 4-12 GHz IF centers us at 355 GHz, again with 8 GHz coverage. This results in the total band matched well to the atmospheric window. Only 16 GHz can be processed at a single time, so maximum frequency coverage involves using only one polarization for each sideband. The minimum frequency resolution per spectral channel is 31.25 MHz, or 27.2 km/s; 90% of the analog bandwidth will be usable for a total spectral coverage of 12,500 km/s, though non-contiguous. Up to J=8-7 CO at z=1.75, different transitions of CO will fall within the band for different z. For 25% of z-space, a transition of CO will fall within this band for some transition. Granted, J=8-7 isn't the most expected search line, but lines of H2O or CI will also be shifted into the band for moderate redshifts. Therefore, we will not only detect over 1000 galaxies, we'll get redshifts for some hefty percentage of them for free. The optical HDF is dominated by galaxies with z<1.7 though the submillimeter HDF would be expected to have a somewhat different redshift distribution. Gallimore's simulations above show this dramatically. But we may well also obtain redshifts for dozens if not hundreds of galaxies simultaneously with the continuum observations. We must determine what the sensitivity will be to line emission for these few hours per pointing integrations, and what fraction of galaxies we might measure line emission for. This can be quite rough for the first cut. Perhaps we should show this somehow on the simulation--randomly pepper the percentage of galaxies for which we expect CO detections with rainbow colors, to indicate spectral detection?

V24. From Silk and Spaans ApJ 488, L79 (1997) Fig 2.: Redshift dependence of the CO e mission spectrum for a starburst galaxy containing 3 x 10^5 Orion regions. The t hree red lines indicate the range in line intensities resulting from metallicities eq ual to 4 and 1/4 times solar. Note the rough constancy of line luminosity with r edshift. For such galaxies, CO emission should be observable to the earliest epochs of galaxy formation. Note (see end of this page) a new preprint by Combes et al. which totally disagrees with this result.

V3. This is shown in the figure to the left, showing z as abcissa with coverage color coded to lines, listed on the ordinate. Each CO line is named on the y axis as in column 1 of the table which I emailed around. The first CI is 492 GHz, the second 809 GHz. The abcissa (z) is given on my sketch as two line segments, the lowest z from column 4 going to column 5, then another segment from column 2 value to that in column 3.

This figure is being modified to show the actual expected CO intensities for the galaxies in the field shown above as a 'pie slice' diagram.

Min points out: Of course, the far infrared atomic cooling transitions dominate the spectra of many galaxies, if not the most luminous ones. However, in many cases they are still comparable to the CO lines in brightness. Any of these lines within the z range given in the table will fall within the spectrometer passband during the deep field integration above. If intense enough for detection, the deep field integration will result in measurement of the redshift of the corresponding galaxy. Note that all z between 3 < z < 8 are covered by one of these lines.

LineRest Wavelength (microns)Frequency (THz)z range
[N II]2051.4632.96 - 4.32
[N II]1222.4615.15 - 7.95
[C II]1571.9114.16 - 5.95
[O I]634.7611.9 - 16.4

We conclude that the spectrometer coverage which produces the continuum image above, made with a Large Millimeter Array, also provides the opportunity to detect emission from bright lines over a large range of z. Hence, in contrast to the Hubble Deep Field, the LMA Deep Field will image distant galaxies in much of the third dimension, distance, in addition to their location on the plane of the sky.

V4. Min S. Yun plots five sigma detection limits for CO lines in Milky Way under two different cosmologies. These are for a Large Millimeter Array of 48 12m antennas. Furthermore, the 5 sigma detection is calculated over the entire 300 km/s line but for a single beam. The LMA should be able to detect CO from the Milky Way out to z=4 even in the q=0.5 no evolution cosmology. The lower magenta colored line is for a five sigma detection in a single 25 km/s wide line, perhaps relevant to detecting resolved clouds in individual galaxies.
See also: F. Combes, R. Maoli, and A. Omont, CO lines in high redshift galaxies: perspective for future mm instruments, accepted in A&A

which you can get off the astro-ph web site (xxx.lanl.gov/archive/astro-ph) as astro-ph/9902286

Further material relating to CO lines in these fields is available from Hi-Z Molecule Page . GBT simulations may be found at the GBT Deep Field Page . Last modified 17 Nov 1999