Next: GOLDMiNe Up: Minutes of the 1st Previous: HIDEEP/HIJASS

Minutes of the talk on the:
HI Mass Function

Steve Schneider (U. Mass.)

I note that "Z" plays an important role in what the ALFA surveys will look for:

HI rich objects few metals low Z

nearby low mass low redshift low z

Arecibo low zenith angle Low Z

Arecibo Blind Surveys

Several surveys have been undertaken in the recent past at Arecibo:

AHISS (Arecibo HI Sky Survey): (Sorar 1994).
- Scanned 19h at fixed declinations (rms of 0.9 mJy)
- 1 object with M_H < 10⁸ solar masses.
Slice Survey: Spitzak & Schneider.
- Scanned 60 sq deg area in Pisces-Perseus region at rms 1.7 mJy
- 3 objects with M_H < 10⁸ solar masses.
ADBS (Arecibo Dual Beam Survey): Rosenberg & Schneider 2002.
- Scanned 450h in drift scans.
- A few sources with unobvious optical counterparts.

HI Selected Galaxies:

For the ADBS, we used the J band luminosities to try to understand the underlying population detected by the survey. At any given HI mass, there are a wide range of stellar luminosities (J band) $L_J \propto {\rm M}_{HI}^2$ almost.

There is lots of gas relative to the stellar mass in the detected objects.

Deriving the HIMF:

The HIMF derived from the ADBS is quite different from that derived from the AHISS by Zwaan et al. Among other things, we used different methods. So the questions are:

Are the difference real?
Could they be due to environment? We find a hint that the HIMF is different in Virgo.

How tricky is it to actually derive HIMF? Coverage wins for volume, but if you do a shallow survey you are sampling only the nearby volume at low masses. The environment locally may be atypical. In particular, you have to worry about the Eddington bias.

Some caveats

5-sigma sounds good, but:
- Sources scatter in and out (flux bias).
- Baselining & interference excision introduces bias.
- Completeness tests fail at this level.
So need artificial test sources to find completeness.

What if you integrate only long enough to see half as far?
- You can cover twice as much volume in same time. so you get more volume by going shallow over area.
But what is the tradeoff?
- Would like to learn about HI flux versus redshift.
- For a given flux limit, you are sampling very small volume for low mass objects.
- Need to ask: how far out do you have to go in order to cover enough volume?

$V_{tot}$ method:
- Total volume within survey in which source could have been detected.
- Need to account for beam shape, frequency dependence, sensitivity variations, line widths
- Must add up the volumes properly.

LSS in the ADBS:
- Can see difference between galaxies within $27^\circ$ of Virgo vs outside Virgo.
- Very different galaxy distribution because of LSS; this needs to be taken into account

Eddington bias:
- Introduced through worrying about Virgo.
- Sources get shifted to lower cz and then their distances are misinterpreted.
- Not a significant result where the velocity dispersion ~ 300 km/s, but can be a problem in cluster regions where the dispersion is more like 1000 km/s.

Completeness:

5 sigma limits turn out not to be enough. You have to be cautious doing a simple cutoff, thinking you are complete to that limit, because sources get scattered in and out. You probably need to cut at a limit of more like 10 sigma to be sure.
You need to spend time worrying about testing completeness.
- Inject fake sources and test if they are found. .
- Methods for quality control of source extraction methods.
- It is surprising how strong a source can be and sometimes be missed.

Questions/Discussion:

Riccardo	Don't you have a bias at low M_HI so that you don't detect many?
Steve	This is an HI selected sample.
Steve	The same effect is seen in optical surveys because of the small volume.