Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.naic.edu/alfa/ealfa/meeting2/handouts/ealfa_whitepaper.ps Äàòà èçìåíåíèÿ: Sat May 13 00:11:08 2006 Äàòà èíäåêñèðîâàíèÿ: Sun Dec 23 04:30:58 2007 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: 2df survey

E{ALFA White Paper DRAFT Version 2.1 07Jul03
1. Introduction
ALFA will enable a variety of survey projects which were not possible at NAIC before ALFA's advent. A number
of those survey projects will be quite ambitious, requiring thousands of hours of telescope time and several
calendar years to complete. These large ALFA surveys will need to be carefully evaluated and regulated, so that
the enormous commitment by the NAIC has the greatest scientic impact.
Large E{ALFA surveys will likely present a sharp departure from the standard proposal process exercised at
NAIC over the past decades. They will produce massive amounts of data in specic areas of research and will
provide an important astronomical reference source for years to come, constituting a major legacy of the Arecibo
Observatory. In a manner similar to eorts of comparable scope at other major observing facilities, the E{ALFA
surveys should be expected to conform to a few basic criteria:
 They will be large, coherent eorts with fundamental scientic goals that cannot be satised by a combi-
nation of regular proposal eorts.
 The data products of the surveys must be of general interest to a broad scientic community. Therefore, a
special eort must be made in the delivery of high quality data products and in the production of robust
software tools that facilitate their access.
 All data products should be made accessible by the general community within the shortest possible time
after data acquisition at the telescope, thus enabling maximally fertile and competitive scientic follow{up.
This document presents a brief summary of the science issues that would be tackled by E{ALFA surveys in the area
of extragalactic Astronomy, the hardware and software requirements for their completion, the synergy with other
ALFA and non{ALFA surveys, the scope of possible follow{up observing programs and general organizational
details. The main goal of this document is to provide NAIC with useful input in its decisions leading to the
delivery of an ALFA system, ancillary hardware and software, and its planning for optimally interfacing with the
E{ALFA user community.
Note that some of these same principles are similar to those aecting other ALFA users, while others may be
complementary, re ecting the special requirements of exploring the extragalactic universe.
2. Science justication, survey strategies, and requirements
The main categories of current scientic inquiry for extragalactic L{band science are:
 The low-mass end of the HI mass function: The relatively small number of low mass (< 10 8 M )
sources detected in blind surveys has left large statistical uncertainties in the number density of the lowest
HI mass sources. There are up to factors of 10 disagreement in the estimated number of sources at 10 7 M .
Because of their intrinsically weak signals, presently these sources have been detected only to small distances.
 High-mass HI sources: These are rare objects and may have surprising properties. One of the highest
mass HI sources, Malin 1, has an extremely low-surface-brightness disk at optical wavelengths. The fre-
quency of the rare sources that fall in the exponential tail of the HI mass function is uncertain, and given
the peculiar nature of Malin 1, it is unlikely that optical surveys will provide the best means for identifying
the highest-mass sources.
 Environmental variations of the HI properties of galaxies: Current observations hint at dierences
in the HI characteristics of galaxies from location to location. The cluster HI mass function appears to be
shallower than the eld mass function, which may be caused by a variety of environmental in uences that
might be explored in detail in nearby groups and clusters.

 The low HI column density environment of galaxies: How extended are galaxy disks? Are any
subpopulations of the Milky Way's High Velocity Clouds (HVCs) extragalactic? Is there a system of HVCs
around other galaxies? How far out and what is the lling factor?
 Evolutionary changes in the HI mass function: There is evidence that star formation has been
changing rapidly over the past few billion years, perhaps changing the number density of HI systems as
strongly as (1 + z) 3:2 . Thus, the overall density of HI in the universe and the shape of the HI mass function
may be quite dierent even at a redshift of z = 0:15.
 The link to Lyman- absorbers: The connection between the HI systems that generate QSO absorption
lines and the HI systems seen in emission at 21 cm remains unclear. Observations at other wavelength
regimes that permit the detection of low-redshift absorption against faint QSOs oer the possibility of
determining where these absorption lines mostly arise.
 Local large scale structure in the Zone of Avoidance (ZOA): Detecting galaxies at 21 cm allows
us to map out the galaxy distribution in regions of the sky where interstellar extinction makes it diôcult
to detect galaxies with optical techniques and where confusion with the galactic foreground hinders the
identication of infrared tracers. Nearby structures ll large fractions of the sky, and the ZOA often
interrupts them, making it unclear how they are interconnected.
 OH Megamasers: Observations of OH Megamasers hold the potential to trace the merger history of the
Universe through cosmic time. ALFA surveys would have an important impact in the detection of such
systems at low to intermediate redshifts.
Planned Major Surveys: Addressing the above topics involves balancing areal coverage on the sky against
the required integration time. Long integrations allow the detection of a given mass HI source to a greater
distance, but at a signicant cost: the total volume coverage obtained in a xed observing time drops inversely
as the limiting distance achieved. In other words, integrating 16 times longer achieves 4 times lower noise, which
permits detecting the same source twice as far away. However, the volume covered in, say 100 hours, will be only
half as big as for a shallow survey that reaches half as far.
If the goal were simply to detect the greatest number of galaxies, the best strategy would be simply to divide the
total available integration time uniformly over the area of the Arecibo sky. On the other hand, this strategy will
sample only relatively nearby volumes for low-mass sources, potentially telling us nothing about low-mass sources
in dierent environments. A strategy of observing deeply at a smaller number of widely-separated positions will
maximize the variety of environments that are sampled. This could be measured by, for example, counting the
number of 10 Mpc volumes intercepted by the survey.
Without going into great detail about integration times and survey strategies, we can identify ve types of projects
that address dierent aspects of the major science goals. They are summarized as follows; their basic requirements
are given in Table 1.
A. All-Arecibo-Sky shallow survey: This survey is designed to maximize the number of galaxies detected.
It will help to better establish the HI mass function in the local universe and give the most complete view of the
Arecibo sky at 21 cm. ALFA will operate approximately 3 times faster than HIPASS, and will have the advantage
of a signicantly smaller beam, so fewer follow-up observations will be needed to disentangle confused sources. The
all-sky coverage will provide the best opportunity for mapping out HVCs, identifying the HI environment around
low-z Lyman- absorption systems, and will provide low-sensitivity coverage of nearby groups and clusters. This
is also the best strategy for nding the rare high-mass sources, and because these are detectable to > 100Mpc
in only  10s integrations, these will be detectable in a wide variety of independent volumes. In addition, a
wide area shallow survey would potentially double the number of known OH Megamasers. Coverage of 10,000
sq. deg. in single-beam spacing (for a Gaussian beam solid angle of 15 sq. arcmin | 3.9'x3.4' at HPFW), with
10 s integrations, will require about 1000 hours of telescope time.
B. Ultra-deep survey: By integrating 50-100 hours per beam, it is possible to detect relatively high-mass
HI sources (> 10 9 M ) out to the bandpass limit of ALFA at z  0:15. To distinguish between alternative
evolutionary models, it will be necessary to detect approximately 40 galaxies. Based on the local density of
MHI > 10 9 M sources, this will require coverage of  0:4 sq. deg. The beam size will be  0:5 Mpc at the
2

limiting distance, so it is likely that galaxies in groups will be confused, but not isolated galaxies. This will be a
challenging survey as it will be a severe test of systematic noise sources and it will be ghting strong interference.
Coverage of 0.4 sq. deg. in single-beam spacing will require between 700 and 1400 hours of telescope time.
C. Medium-depth environmental survey: While the shallow survey will detect the largest number of
galaxies of all masses, it will only detect low-mass ( 10 7 M ) galaxies out to distances < 10 Mpc. This means
that the low-mass tail of the HI mass function will only be sampled well within the local supercluster. Existing
Arecibo HI surveys have already sampled low mass HI sources out to much greater distances, although with
such small numbers of sources detected that the inferred environmental dierences remain uncertain. A survey
covering 10-20 times the volume of these surveys at comparable sensitivity ( 1 mJy at 5 km/s resolution) would
be able to explore variations in the low-mass tail in a statistically robust way. Coverage of 2000 sq. deg. in
single-beam spacing, at 60 sec integration per beam, will require about 1200 hours of telescope time.
D. ZOA survey: Exploring galaxies and large scale structure within  5-10 ô of the Galactic plane is distinct
from the shallow survey for two reasons. First, Parkes is already surveying the ZOA at Arecibo declinations to
a greater depth than in the HIPASS project, so that the gain of a shallow survey would be in resolution, not
sensitivity. Second, there are special opportunities to \piggyback" with Galactic and pulsar surveys to achieve
high sensitivity levels. This would permit a much more detailed mapping of galaxies in the ZOA than would be
driven by extragalactic science goals alone. The parameters for a ZOA survey remain open to compromise with
other ALFA surveys' concerns.
E. Group and Cluster surveys: Several nearby regions and galaxies provide important targets for deeper
surveys to explore the low column density HI environment on the outskirts and between galaxies. For these
surveys it is useful to target regions that are well studied at other wavelengths to contrast properties of the
neutral gas with other populations. In particular, the Virgo Cluster is the nearest cluster and is well-situated
in the Arecibo declination range for a detailed study. Other high-density regions within the local supercluster
have also been identied that appear to have unusual properties in, for example, the number density of dwarf
systems, and several nearby galaxy groups provide an opportunity for exploring the intergalactic medium and
draw comparisons with the Local Group. Exploring the environment around nearby isolated galaxies at high
sensitivity would provide an important contrast for interpreting environmental eects. For example, a survey of
400 sq. deg. around a cluster or group, in single beam spacing at 300 sec integration per beam, will require about
1200 hours of telescope time. One important restriction on the scheduling of such a project is that it may require
a few to several hundred days of scheduling over a restricted LST range.
Table 1. Basic Survey Requirements
All-Arecibo Sky shallow survey 10 4 sq.deg. 10 s/beam 1000 hours
Ultra-deep survey 0.4 sq.deg.  300 ks/beam  1100 hours
Medium-deep environmental survey 2000 sq.deg. 60 s/beam 1200 hours
ZOA survey tbd tbd tbd
Group and cluster survey (400 sq. deg.) (300 s/beam) (1200 hours)
It should be noted that for most of the surveys proposed, a velocity resolution of  5 km/s is needed to allow for
the detection of the narrowest-line HI sources expected and to achieve good resolution of prole shapes. Some of
the surveys may be combined or overlapped. For example, the medium-depth survey can be achieved at the lower-
frequency end of the bandpass of group and cluster surveys, although this strategy would cover fewer independent
volumes than a xed-declination survey, and there may be problems with optical follow-up for background sources
in the vicinity of nearby, bright galaxies.
3. Follow-up Observations
It is expected that the brunt of follow{up activity after E{ALFA surveys will impact the Arecibo telescope.
Single{feed observations at 21 cm of detected sources will be required for a variety of reasons: (i) conrmation of
detection and signal{to{noise improvement; (ii) higher spectral resolution desired; (iii) higher spatial sampling of
3

resolved objects. Especially for the wide angle surveys, which will detect thousands of line sources, follow{up work
can potentially generate requests for telescope time commensurate with that assigned to the surveys themselves
| depending on how deep the signal extraction algorithms are trained to delve. Continuum and line follow{up
Arecibo observations at other wavelengths will likely request relatively small amounts of time.
Another major fall{out of E{ALFA surveys will be given by requests of HI aperture synthesis observations,
oriented mainly towards telescopes such as the VLA, WSRT, the GMRT and, in lesser measure, ATCA. At
other wavelengths, observations of E{ALFA detections will likely follow at mm telescopes and arrays, while
optical and UV photometry and spectroscopy will follow, especially to establish the characteristics of stellar
populations associated with the HI objects. The possibility of simultaneous operation of ALFA and SIRTF oers
very attractive options for mid{ and far{IR follow{up work.
A basic data set that would be highly desirable for cross{referencing with E{ALFA sources would be narrow-band
H information about all sources and at all conceivable redshifts covered by E{ALFA. Such data would allow us
to draw an immediate relation between the HI mass and the star formation rate. Narrow band lters are typically
about 50  A wide, so at least 15 lters would be required to cover much of the ALFA z range. Progress in the
area of tunable lters presents interesting possibilities in this area.
4. Synergies with Other Surveys
Synergies with non{ALFA Surveys: The HI line observations from ALFA surveys will be combined with
information at other wavelengths in orde to best characterize the nature of the detected objects. The surveys
that will have most relevant information available by the time the E{ALFA surveys will be executed are:
 1. Optical: the PSS and Sloan surveys. The latter is a ve-color imaging survey with the goal of mapping
about 10,000 square degrees of the sky to 22 mag at a few percent accuracy and provide spectroscopy
of all galaxies down to 19 mag. Sloan partly overlaps with the Arecibo sky. Future surveys with the
LSST (scheduled for operation after 2008) will also provide rich cross{comparison with the E-ALFA survey
archives.
 2. Ultraviolet: the GALEX (launched in 2003) all-sky survey will provide imaging and spectroscopy in two
bands to nearly the same depth as Sloan, depending on the spectral energy distribution of an object.
 3. Near IR: the 2MASS survey of the entire sky in three near-IR bands (J, H, and K{short), recently
released.
 4. Far IR: still the old and trusty IRAS, with supplements from ISO. The ASTRO-F mission is scheduled
for early 2004 and SIRTF is to be launched in August 2003. These telescopes will provide survey data much
deeper than IRAS.
 5. Radio Continuum: the VLA FIRST survey.
 6. X-ray: RASS (ROSAT All Sky Survey).
The requirement for eôcient use of these surveys in the process of source identication and follow{up is the
availability of HI source positions of the highest possible accuracy. In this context, eorts towards maintaining
ALFA to a high degree of pointing delity are urged. Wide eld ALFA surveys will benet from the natural
calibration provided by continuum sources in the mapped elds.
Given the large number of expected detections, the process of cross-correlating source positions with those of
dierent surveys should be automated. Since this is a problem generic to a number of surveys, existing tools
will be tapped (see for example the experience at JHU or NVO). Format compatibility of source lists is clearly
desired.
Synergies with other ALFA surveys: Multiple backends make commensal observing an exciting possibility
that will allow teh simultaneous achievement of various science goals while surveying the sky. Other consortia
4

are aimed at discovering pulsars (P{ALFA) and surveying the Galaxy, primarily in HI (G{ALFA). All consortia
have the common goal of surveying the Galactic Plane (b = 5 ô ), as well as the entire Arecibo sky to a shallower
level. The Galactic Plane is the highest priority for both P{ALFA and G{ALFA; in terms of observing strategy,
their exibility is thus likely to be higher for surveys out of the plane (i.e. commensal with an All-Arecibo-Sky
E{ALFA survey. The E{ALFA main goal in the Galactic Plane is a Zone of Avoidance (ZOA) survey, which
would favor drift mode observing to most accurately locate optically-hidden galaxies. P-ALFA favors a long dwell
time per position. The preferred G{ALFA observing method is drift mode, which is compatible with E{ALFA.
The E{ALFA and G{ALFA surveys would also like to plan the survey for multiple passes over each region of sky.
This approach would have the advantage of facilitating the discrimination between real sources and RFI, and
could be useful for the detection of continuum transients. On the other hand, piggy{backing on a Galactic Plane
P{ALFA survey would allow E{ALFA to sample a signicantly deeper volume than possible in drift mode. Both
options are thus attractive. Observing tests should be completed to determine the plausibility of using multiple
pass, drift scanning for part of the P{ALFA survey. G{ALFA and E{ALFA should also work together in doing
test observations to determine the best time of day to observe and the characteristics of the ALFA beam.
Both G{ALFA and E{ALFA are interested in online reduction of the data. Since these data reduction methods
will have many similarities, these two groups should communicate and hopefully ease the software development
burden. The G{ALFA group has the diôcult task of dealing with highly extended structures and is planning
preliminary observations to explore options for data reduction (e.g. frequency switching vs. position switching).
There will most likely be an overlap of velocity coverage for E{ALFA and G{ALFA at the low-velocity end (-500
to 500 km/s) which includes the Galaxy, the Local Group, and high-velocity clouds (HVCs). An understanding
will be formed between the two groups on the ndings in this overlap region. The E{ALFA and G{ALFA groups
desire a minimum velocity resolution of 5 km/s and 1 km/s, respectively. The dierent requirements of the surveys
require separate spectral backends: while E{ALFA wishes to maximize bandwidth coverage, G{ALFA requires
high spectral resolution. A parallel E{ALFA plus G{ALFA survey is very unlikely to satisfy both groups, by
using the WAPPs as a common set of backends.
5. Hardware Specications
While the development of hardware for data acquisition that will be employed by the E{ALFA science surveys is
the purview of NAIC, it is expected that NAIC and E{ALFA will interact closely on decisions impacting hardware
specication and use. Our requirements are summarized as follows:
 High sensitivity will be the key for E{ALFA surveys. Optimization of A eff /T sys across the entire ALFA
band to the lowest possible level, with implementation of plans for a tertiary skirt, is strongly desired.
 Several dierent E{ALFA surveys will be proposed, covering a wide range of desired backend characteristics.
While some of the surveys to be proposed may require only 50 MHz of dual polarization coverage for all
seven feed horns, other surveys will be impossible without access to 200 MHz of coverage. Table 2 below
presents a preliminary set of specications for an ideal backend system that will satisfy all the currently
foreseen E{ALFA survey demands.
 The success of E-ALFA science will strongly depend on our ability to combat RFI. RFI suppression, mit-
igation and excision both through vigilance and the development of hardware tools fall in the purview of
NAIC. Software excision and mitigation tools will be developed by the community and NAIC sta, perhaps
jointly. Modes to reduce or compensate for RFI must extend throughout the observing chain, from an IF
system with suôcient dynamical range to avoid gain compression in the nal signal, to insuring adequate
sampling in the backend to avoid spectral degradation, removal of RFI in the software, ltering of the signal,
placement of reference horns and providing for cross products between all horns in the ALFA system.
 As the development of backend instrumentation that will satisfy the requests of dierent sets of ALFA
users will be subject to the necessity for compromise, we emphasize the need for transparency in order to
maximize the spread of information among dierent groups. We urge NAIC to maintain an open forum on
this issue, facilitating cross{fertilization of ideas.
5

 The maintenance of state{of{the{art network connectivity at the Observatory and to the outside world is
urged for the ALFA era.
 NAIC should archive the raw data, possibly in collaboration with partners, and provide access to these data
by members of the E{ALFA science team via network transfer from a mainland location.
 Plans must also be put in place that provide adequate hardware for data processing, storage and archiving
both within NAIC, observing consortia and, in the case of the nal data product archive, the scientic
community at large.
 Compatibility with other national observatories is strongly encouraged.
Below we expand on the rationale for the above recommendations.
System Optimization: Continued eorts are encouraged to optimize system performance and to lower the
system temperature and spillover, in particular, through the deployment of a tertiary mirror skirt. Likewise,
optimization of the A eff /T sys across the entire ALFA frequency band and achievement of uniformity among the
outer beams is important to the E{ALFA surveys. A detailed understanding of the variations in beam structure
as a function of telescope conguration will be necessary in order to accurately correct for the large impact that
the \dirty" ALFA beams will have on the characterization of HI sources, especially those with larger angular
extent.
Backend: At the E{ALFA workshop, NAIC committed to delivering a minimum backend system (the expanded
WAPPs) for ALFA with bandwidth coverage of 100 MHz at 3-level sampling or 50 MHz at 9-level sampling, in
dual polarization mode for each of the 7 feeds. While this system may have acceptable specications for some
of the E{ALFA science projects, other surveys (e.g. the Ultra Deep Survey) will be severely hampered by such
a backend. Even an upgrade of such a system to a 2 times 100 MHz bandwidth coverage would fall short for a
very deep survey, which would require higher level sampling for RFI excision in an RFI{infested spectral range.
Surveys other than the Ultra Deep one may be succesfully carried out by the \minimum" backend, although the
opportunity of detection of the very largest HI masses would be reduced by a bandwidth restriction to 100 MHz.
Even in these cases, a backend system with high level sampling would be desirable.
Table 2 below outlines the requirements for the \ideal" E-ALFA backend, one that would allow all of the surveys
to achieve their science objectives with maximum eôciency. This list meets all currently foreseeable requirements
of the surveys identied in Section 1.
Table 2. The \Ideal" E-ALFA Backend
Spec. Loss for not Achieving Comments
200 MHz BW Ultra deep survey cannot be performed The deep surveys will push ALFA's limits
more than other proposed surveys
8+ bit sampling Reduction in RFI{excising capability RFI could also be alleviated with
Loss of useful BW (1)Narrow band lters &/or blankers
(2) Flexible scheduling
 8192 chan/pol Inability to achieve science objectives Max. channel sep. of 5 km/s over 200 MHz
2 Pols per Rx Loss of sensitivity
<1 ms data blanking Time/sensitivity/obs. eôciency loss <1 ms blanking allows eôcient RFI excision
Cross-products & Loss in useful BW and eôciency Require a min. of 2 ref. horns with all
reference horns X-products b/w ref. horns and ALFA feeds
For optimal limitation of the spectral spread of RFI features, multi-bit sampling is desirable. Preliminary studies
indicate that spectrally sharp, strong RFI signals can be constrained within < 3 spectral channels of a correlator
backend if  8 bit sampling of the ACF is allowed. It should be noted that backends with 12{16 bit sampling are
6

already commercially available. Thus obtaining a backend with suôcient sampling to severely limit the signal
degradation, in frequency/channel space, of extremely strong RFI sources is technically aordable.
RFI is also known to have an impulsive component with duration on the order of s { ms. It is still unclear
what the recurrence rate of such features is: were it to be high, high dump rates would be required in order to
optimally excise RFI in software; if, however, the statistical properties of impulsive RFI were such that the data
loss by deleting aected dumps at, e.g., 1 Hz dump rate were negligible, higher rates would not be necessary.
Ongoing RFI monitoring at high time resolution by the pulsar community will produce useful data that will help
resolve the dump rate question.
RFI Suppression and Mitigation: The presence of RFI is a strong concern because it might severely limit
or even prohibit some E{ALFA science. E-ALFA surveys will observe to 1230 MHz where many strong RFI
sources exist. Consequently, care needs to be taken to insure the maximum usable bandwidth is available over
the maximum amount of time. Appendix C lists the main, known sources of RFI in the ALFA spectral domain
at Arecibo.
NAIC must continue to practice due vigilance in minimizing the generation of RFI by on-site devices. We
commend the eorts of NAIC sta towards protection of the radio spectrum in national and international spectrum
management forums.
Below, we brie y discuss possible avenues in RFI mitigation.
 Notched Filters. These would be placed at the system front-end to prohibit certain frequencies from
entering the IF path. The main advantage of this approach is the prevention of gain compression. Ad-
ditionally, notched lters would ease the application of other RFI mitigation techniques discussed below.
Among the disadvantages of a notched lter system are (1) the expense of installing 7 notched lter systems
on ALFA, and (2) if additional burden of making them switchable, else they would prevent all observations
with ALFA at the notched{out frequencies.
 Radar Blankers. In the case where the characteristics of an RFI signal are well known, a tuned blanker
could be employed, much in the manner of the 1330/1350 MHz blanker now in place at Arecibo. The
advantage of a system like this is that it can be turned on and o with ease, and could also be employed
only on the E-ALFA backend path, simultaneously not restricting other ALFA surveys. Additionally, radar
blankers do not have to compete for space in the ALFA receiver structure, but would be placed downstairs
in the signal processing area. The main shortcoming is that the radar population in the ALFA band is large
and diverse; thus, total suspension of data taking for each of a majority of radar systems may signicantly
reduce the on{source duty cycle.
 Software. If an RFI signal is conned to a few channels (best if 8+ bit sampling is available), and the data
is dumped at a fairly rapid rate, the RFI signal could be removed from the data during the data reduction
process, allowing for considerable exibility in the RFI removal algorithms. Succesful application of this
technique is optimized with avoidance of gain compression, too slow data readout and sampling at too few
levels.
 Reference horn subtraction. This hardware addition would considerably aid in the elimination of RFI
at the software level. It could result in a fairly clean subtraction, but suers from the same constraints as
the software mitigation listed above.
Network Connectivity to Arecibo and to the data archive site: As in most aspects of astronomical
survey undertakings, E-ALFA will place strong demands on network connectivity. We understand that there is
currently some risk to Arecibo of the loss of its high-speed Internet connection when the current grant runs out
at the end of August 2003. As generally recognized, this would be a disaster, as nearly all aspects of E{ALFA
activities, from communication within E-ALFA and NAIC to observing to data transport to archive access, will
rely heavily on the Internet.
Because limitations in bandwidth and speed to the island of Puerto Rico are likely to be more stringent that
those placed on mainland sites, it is likely that at least the archive oering principal public access, but possibly
7

all archival storage, will be located o-site. It is critical that the highest performance access is available to the
archive wherever it might be located.
Data Storage, Processing, & Transport: An E{ALFA survey will produce raw data at a minimum rate
of > 1 GB/hr; this rate is likely to rise for next generation spectrometers. By many standards (e.g., LSST will
produce 18 TB/night), this is not a large number. Processed data will be reduced in volume considerably by time
averaging, but will have to be stored at a multitude of stages as it proceeds through the pipeline. We understand
that raw and Level I data (see next Section) will be archived and maintained by NAIC. While NAIC has been
acquiring hardware at suôcient pace to meet current needs, future planning for on-site monitoring, processing
and disk storage must accommodate E{ALFA needs both on and o site.
The size of E{ALFA data sets raises issues concerning the mode of data delivery. While NAIC may assume
responsibility for archival of level I data products, observing teams will require access to the raw data. For data
transfer rates, we assume a minimum data production rate of 0.5 MB/sec. Three principal options exist for
the transport of raw data: (a) network transfer, (b) magnetic tape and (c) hard disk. We recommend (a) that
network transfer must be enabled to allow real-time access and (b) a collaborative eort between the E{ALFA
participants and the NAIC needs to specify and develop a Linux-based hard disk storage/retrieval system that
meets survey requirements and is both cost eective and easy to maintain.
Processing Hardware: While data storage presents a manageable problem, data processing for E-ALFA
raises considerable challenges beyond the norm for current Arecibo extragalactic observers. Immediate and
detailed planning is needed to avoid a data reduction bottleneck due to lack of hardware (plus manpower and
software!) resources. NAIC will have to expand its current Linux cluster to accommodate Level I processing tasks.
Observing teams need to plan and acquire the necessary high performance systems with suôcient disk storage and
backup capability. Compatibility among systems must be insured if the processing task is distributed. Current
indications suggest that Linux clusters are preferred, as they can provide the needed capability and architectural
exibility at the lowest cost.
Public and long-term archive hardware: The location and requirements for long{term public access to the
E{ALFA data archive need to be specied.
6. Software Specications
The development of the software necessary to monitor, process, analyze, archive and access survey data and
data products presents a critical challenge to the success of E{ALFA. Of necessity, but also of desireability,
software development will be a shared responsibility of NAIC and the E{ALFA science teams. Many of their
tasks/requirements also apply to non E{ALFA surveys. NAIC, E{ALFA and the other ALFA concerns should
work together to develop a systems approach to these tasks, as appropriate and possible, that will minimize cost,
eort and the time to complete.
Software development tasks and their responsibilities can be broken into three distinct categories or \levels", as
shown in Table 3.
Table 3. Data & Software Processing Levels
Level Description Prime responsibility Access
I Observing and rst stage processing NAIC public
II Production of public data products E-ALFA public
III Science toolkit E-ALFA proprietary?
While the responsibility for the production of observing and calibrating software and lowest level data products
is expected to be assumed by NAIC, more elaborate stages of the data processing and associated software will be
deliverable by E{ALFA groups. While level I tasks and data products will be mainly the responsibility of NAIC,
substantial collaborative input from E{ALFA teams will be available. The basic understanding of who will be
responsible for what is summarized in Table 4.
8

Table 4. Software Responsibilities
NAIC  will provide the tools needed for telescope control, on-line data monitoring, trouble
shooting, raw data storage and record keeping needed by the science team:
ô to carry out the observations in an eective and eôcient manner and
ô to conduct real-time assessment of data quality.
 will archive the raw data and provide access to it.
 will conduct, evaluate and verify a rst level of data processing.
 will archive the Level I data products in standardized data formats
and provide the tools needed to access and use them.
 will provide access to the software needed for Level I processing.
 will provide the technical information fundamental to the Level II
processing (e.g. calibration, r identication and characterization,
pointing, beam characterization etc.).
 will independently verify the Level II data products.
 will archive the Level II data products and provide access to them.
 will provide access to the software needed for Level II processing.
E{ALFA team  will conduct, evaluate and verify a second level of data processing.
 will provide to NAIC the Level II data products in standardized data formats
and the tools needed to access and use them.
 will provide to NAIC the software needed for Level II processing.
NAIC + E{ALFA  will insure capability for portal to NVO.
Software tasks: The tables in Appendix D give a preliminary itemized listing of software required for E{ALFA.
Some portions may be of general use to other ALFA users. The most advanced elaboration, leading to level III
data products, might/will be proprietary; diusion will take place by publication through regular channels. As
for public products, access should be provided through NAIC. HI source catalogs might be produced not only
by groups associated with the E-ALFA consortium but also by others once public data access is made possible.
Because dierent detection algorithms might yield dierent catalogs, separate science analysis groups might
produce separate catalogs. Once published, they should all be accessible through the E{ALFA archive, with
appropriate documentation.
Data formats: Because of the ease of transport and availability of software tools/utilities, raw data should
be written and archived in SDFITS format. Further downstream processed data and data products of levels II
and III should also be exported in SDFITS whenever appropriate. Advance planning for the physical location
and maintenance of the entire archive should begin promptly. A coordinated eort with other ALFA concerns is
highly desirable. The development of access tools and the necessary links to the NVO should also be pursued.
Reduction platforms: There are several options for processing data, most notably Classic AIPS, AIPS++,
CLASS, GIPSY, IDL and MIRIAD. We anticipate that Level I processing at Arecibo will be conducted using
IDL. At this stage, there is no unanimous E{ALFA choice of a processing platform. It is however stressed that
whatever the platform, the processing pipeline should be well documented and exportable, and that output data
products conform to SDFITS standards and provide convenient access by a variety of software packages/tools.
7. Organization
The formation of a survey consortium should provide a framework within which extragalactic HI surveys with
ALFA will facilitate large science programs signicantly greater in scope and ambition that typical Arecibo ob-
serving programs. Such surveys will require a large investment of telescope time and hardware development,
manpower resources, planning and management. At the same time, such large projects can derive signicant
benet through coordination and standardization along the lines of the HST Treasury and SIRTF Legacy pro-
grams. As such, E{ALFA surveys carried out under the consortium paradigm should be governed in a manner
9

that balances both the investment of telescope time with the investment of individuals who carry out the surveys.
The nature of NAIC as a US NSF{sponsored national center plays a role in dening the paradigm. Both NAIC
and the science team will necessarily have to work together closely to insure the success of the science program.
Here, we present a summary of the principles, objectives and circumstances that might form the basis for the
organizational structure of an E{ALFA consortium. A detailed list of responsibility assignments between NAIC
and the E{ALFA consortium is also presented in Appendix E. In the end, the E{ALFA consortium structure
must accomodate the objectives and constraints while enabling the eective accomplishment of science goals. In
this draft, the proposed model is not rm, but rather an outline that the interested community and NAIC can
use for a proper denition.
The fundamental principles that should govern E{ALFA surveys are listed in the Introduction to this docu-
ment.
The objectives of the consortium or survey team should be:
 To achieve the maximum science goals
 To have the broadest possible scientic impact
 To make the best use of the telescope and available resources
 To engage the broadest community of researchers
 To provide the best education and training for students and their teachers
 To raise the broadest level of public interest
The operational principles which dene the function of a \consortium" are:
 To plan, execute and manage the survey observations on behalf of the larger community
 To coordinate the survey for maximum eôciency and maximum impact
 To contribute expertise, software and documentation to NAIC in concert with the Observatory's eorts on
behalf of the consortium.
 To deliver standardized, robust, high quality data products that enable further research
Competition with the other science \consortia" (P{ALFA and G{ALFA) will encourage optimization and com-
promise. In some cases, the choice to undertake parallel surveys would expedite the allocation of telescope time
but it may also introduce ineôciencies: it is the function of the consortia, in concert with NAIC, to argue in favor
of the best circumstance for the science output.
Tasks and Assignment of Responsibilities: The E-ALFA consortium and NAIC will work in a collaborative
partnership to achieve the success of E-ALFA science. To accomplish the stated science goals, NAIC and the
E-ALFA consortium/science teams will work together to plan, design and carry out the surveys, acquire and
develop the needed resources, archive the data and data products, and provide community access, educational
opportunities and public information and outreach. The assignment of responsibilities, which will be more
completely specied in a memorandum of understanding between NAIC and the consortium/science team, is
included as Appendix E.
Consortium Organization: As demonstrated by the attendence at the March 2003 workshop, the scientists
interested in E{ALFA are a broad and diverse group, driven by a variety of scientic motivations and well-
conditioned to multiwavelength studies. Several dierent E{ALFA surveys are likely to be conducted, which will
assemble dierent, perhaps partly overlapping, observing teams. These teams may develop their own internal
set of operational rules. The E{ALFA \consortium" will evolve into an organization which will include members
of the survey teams, as well as representatives of the community at large. A Coordinating Committee should
eventually be formed, which will provide the main forum within which will take place a well{tended interplay
between dierent observing teams, between dierent consortia and between consortia and NAIC.
At this stage, E{ALFA has constituted a preliminary structure which, by election, has chosen three coordinators
| two US based (Riccardo Giovanelli and Steven Schneider) plus one based abroad (Lister Staveley{Smith)
|, mainly charged with the responsibility of editing this draft and temporarily serving as an interface between
10

the E{ALFA community and NAIC. The evolution of this organization will take place at future gatherings of
E{ALFA interested scientists, and in uenced by NAIC and other ALFA consortia self{imposed guidelines. In the
meantime, ventilation of important issues in the community will be necessary. Among them: the denition of
consortium membership, the interplay between consortium members and observing team members, the role of the
consortium pre{ and during survey work, the denition of timescales for membership, levels of commitment, the
structure of a Coordinating Committee, its constitution, nomination/election process and duration of charges.
Preliminary ideas on the above matters have been assembled by members of the \Organizational Matters" Working
Group. Their questions and suggestions are enclosed in Appendix F.
8. Education and Outreach Activities
In accordance with the agency requirements of both NAIC and the U.S. NSF{funded participants, education and
outreach activities should be actively incorporated into E{ALFA activities. Other countries may encourage or
require similar development under grant awards. These eorts should be the joint responsibility of both NAIC
and the consortium members. As part of its mission, NAIC should continue to be a driving agent behind the
education and outreach initiatives associated with ALFA in general and should assist the E{ALFA consortium
in developing materials and programs specically for E{ALFA. Educational opportunities should be provided for
students at all stages of their college and graduate careers, including hands-on training at Arecibo as well as in
software development, data processing and analysis at their home institutions. Outreach activities should not
only focus on the outstanding Angel Ramos Foundation Visitor Center but also involve the institutions associated
with team members and in the various states and countries in which they are located. Examples of activities
associated with education and outreach appropriate to E{ALFA and their primary responsibility assignments are
included in Table 5.
Table 5. Example Education and Outreach Activities
NAIC Scientic workshops 1st one held Mar2003
NAIC E-ALFA Web site Technical information for astronomers about ALFA & EALFA
NAIC E-ALFA Web site Outreach materials, including E-ALFA science
NAIC Documentation website Repository for E-ALFA-specic documentation
NAIC Software website Repository for E-ALFA-specic software
NAIC Observing status website Updated during survey(s) (Need to specify access?)
NAIC+EALFA PPT presentation Set of slides that summarize E-ALFA for general use
NAIC+EALFA Scientic meeting(s) poster e.g. AAS June 2004/Jan 2005?
NAIC+EALFA E-ALFA "Ask an Astronomer" Possibly use http://curious.astro.cornell.edu?
NAIC E-ALFA AOVEF display E-ALFA-specic display (copies available to others)
NAIC NVO contact/exchange Coordinated with G-ALFA?
NAIC+EALFA Student involvement program At AO and home institutions
EALFA Undergrad problems/exercises To encourage incorporation of E-ALFA into courses
EALFA Graduate problems/exercises To encourage incorporation of E-ALFA into courses
EALFA Graduate seminar Cornell Astro 620 Spring 2004 (Cordes/Giovanelli/Haynes)
\Large Surveys in Radio Astronomy"
9. Funding Support
In order to insure success and broad participation in the E-ALFA surveys, participants need access to resources
covering all aspects of survey planning, software development, carrying out the observing program, data processing
and analysis, education and outreach. Current participants nd themselves in varying employment circumstances
with varying opportunities for funding.
11

Factors which complicate nancial support for participants of the E-ALFA surveys include:
 the multi-year duration of the surveys
 their legacy nature, that is, that they produce data products that enable further research by a broader
community
 the major manpower eort involved in their planning, execution software development and data processing
and verication
 the computer hardware resources necessary for data storage, data processing and scientic analysis
 the diverse, international and collaborative nature of the consortium/teams
 the national center character of NAIC and the Arecibo Observatory
While some of these complications are fairly new to extragalactic HI science at Arecibo, they are not new to other
large facilities, namely the NRAO, NOAO, 2MASS, SDSS, SIRTF, HST, Chandra, XMM, etc. How participants
in those large survey projects are supported and coordinated varies signicantly, and not all paradigms are equally
applicable to the circumstances that aect E-ALFA.
The E-ALFA surveys oer enormous potential to explore the nearby universe and to gain insight into how gas
disks assemble and evolve. They will produce unique data products that will enable and stimulate follow-up
research by a much greater community. Therefore it is in the interest of all parties { the survey teams, the NAIC
and the funding agencies - to secure suôcient resources to insure that the full potential of the E-ALFA surveys
is achieved.
US NSF Funding The 2000 Astronomy and Astrophysics Survey Committee (AASC) report \Astronomy and
Astrophysics in the New Millenium" urged that the funding agencies provide adequate funding for data analysis
along with facility support. In a recent experiment, NRAO has begun to oer support for instrumentation
development and graduate student thesis research associated with use of the Green Bank Telescope. The latter
program is being expanded to the VLA and VLBA within the limits of available funding.
While the NRAO/GBT programs are valuable rst steps, they (a) do not apply to software development and
(b) do not allow for student tuition payments. Therefore, they alone do not address the needs of faculty at
most US institutions. University faculty participants need support for summer salary, travel, capital equipment,
publications, student stipends and tuition, as well as applicable indirect costs; in the case of E-ALFA, they may
also need an augmentation for communications costs. For a truly level playing eld, sources of funding for all of
these costs must be identied.
Tenure-track faculty at many U.S. institutions must demonstrate their ability to raise adequate support for all
of their research initiatives, including student support. Thus, to enable junior faculty to participate, funding for
E-ALFA must be commensurate with that available from other possible sources (e.g., NASA).
The decoupling of E-ALFA support from the grants program funding process places the NSF-supported indi-
vidual investigator at risk of being granted telescope time but not having the funding resources to carry out
the observations, software development and subsequent analysis, or vice versa. In the case of the major survey
eorts involved in E-ALFA, covering all phases of the project from concept development through to scientic
analysis, the timescale for science results may exceed the grant duration. Because of the diverse nature of the
likely E-ALFA survey teams and the importance to the NAIC and the broader astronomical community of the
success of these surveys, possible mechanisms for long term continuity of funding, tied to project review, should
be explored. A collaborative proposal between NAIC and the consortium and its participants might be the best
t with the current NSF peer review process.
For full impact, NSF funding should also be provided, as suggested by the 2000 AASC report, for analysis of the
E-ALFA Level II data products by the broader community.
The broader impact, educational and outreach potential of the E-ALFA surveys are enormous. Likewise, funding
support for those activities should be approached collectively as a partnership by the E-ALFA teams, the other
X-ALFA consortia and the NAIC to both maximize and optimize their eectiveness.
EU Funding The international interest in E-ALFA is evident. Specically, concerning the interest shown from
within the European Union: among the 65 E-ALFA consortia members presently registered on its website, 22 are
12

from astronomical institutes in the European Union, representing a total of 11 Institutes in 5 countries (France,
Germany, Italy, Spain and the United Kingdom).
The European Commission, within its Sixth Framework Programme (FP6), covering the period 2004-2008, has
funds for various kinds of projects. The most appropriate for the E-ALFA project appears to be the inclusion of
research themes related to it in a Research Training Network of European Institutes.
The main purpose of such networks is the employment of postdocs, who do not necessarily need to be citizens
of a European Union country, thereby also providing potential job opportunities for young astronomers from the
US and elsewhere. The proposal for a Network can include other research themes as well. In practice, at most a
dozen Institutes can form a network and employ about one postdoc per institute in total.
There are in principle dierent options for the employment of postdocs under the FP6 Marie Curie Actions: host-
driven-actions (Marie Curie Research Training Networks, etc.); individual-driven actions (Marie Curie Individual
Fellowships); and excellence promotion and recognition (Marie Curie Excellence Grants, etc.).
The dierent options of benet to the E-ALFA Consortium will be investigated further. Information on the FP6
and the Marie Curie Actions can be found on the website (http://www.cordis.lu/fp6) of CORDIS, the Community
Research and Development Information Service.
13

APPENDICES
A. Scientic Justication: Wide Angle and Directed Surveys
In this section we elaborate on the science justication for the large area surveys.
Introduction: A shallow All-Arecibo Sky Survey will provide a more sensitive, higher resolution (spatial and
velocity resolution) look at the 21 cm sky than has been possible with HIPASS or HIJASS and will provide a
tremendous improvement in the volume of space surveyed over the Arecibo Dual-Beam Survey (Rosenberg &
Schneider 2000), the Arecibo HI Slice Survey (Zwaan et al. 1997), the Arecibo Survey (Spitzak & Schneider
1998), and all of the earlier surveys. The survey will provide much better statistics on the HI mass function
then has been obtainable with previous surveys as well as giving a look at the change in the mass function with
galaxy environment. The survey will also nicely combine with the data from the Galactic survey for studying high
velocity clouds up to high Galactic latitudes and will provide the basis for many other scientic investigations.
We discuss the basic parameters of the survey and some of the interesting science that we will be able to do with
it here.
Observing Strategy: Our present thought about the observing strategy for this survey would be to get 12
seconds of integration per point using 2 driven 6 second scans. The duplication of scans will greatly improve the
interference mitigation and data checking. The main concern with using 6 second scans is the eect of driving
the system on baseline stability and noise characteristics of the data. These issues will need to be tested to
determine whether this survey strategy will allow us to meet the scientic goals. Presently the Galactic group
is considering a double driftscan survey of the entire Arecibo sky which would give twice the observing time
discussed here. This increase in observing time would provide greater sensitivity to this survey if we were to
piggy-back the observations and would be desirable if it does not detract from the other science that is of interest
to the extragalactic consortium.
For some important pieces of the shallow survey science discussed here, including studies of large scale structure
beyond z = 0:1, searching for OH megamasers at z  0:25, and studying absorption line systems at high redshift,
the 200 MHz backend is a necessity or, for the case of the megamasers, would greatly improve the quality of the
science. Therefore, a 200 Mhz backend is very desirable for the shallow survey science goals.
Simulating Galaxy Detections: John Spitzak has put together a simulation of an ALFA survey with
adjustable parameters for the purpose of studying the expected detection statistics for a given survey strategy.
In particular we would like to investigate the statistics of galaxy detections as a function of mass and distance as
a way to understand how well we will be able to compute the HI mass function since that is a primary goal of
the survey. The simulation is designed to include a density map (not yet implemented) to study the statistics of
galaxy detections as a function of environment.
The basic outline of the simulation is as follows:
 The Arecibo sky will be divided into eqi-volume, equal velocity-depth cubes which will be populated with
galaxies.
 The cubes will be populated with galaxies based on the HI mass function that has been dened (the
parameters can be dened for any run of the simulation and can be tied to the density of the cube).
(M) = ln(10)
  M (+1) e M (A1)
 The scaling of the mass function (and possibly the input mass function slope) will be determined from
an estimate of the density in each cube. The major overdensities and underdensities that pass through
the Arecibo strip include the overdensities: Virgo Cluster, Local supercluster, Pisces-Perseus supercluster
- ZOA, Coma cluster, A1367 cluster, and the Coma Wall as well as the underdensities: Local Void - ZOA,
Monoceros Void - ZOA, Orion Void - ZOA, Taurus Void - ZOA, Microscopium Void, Virgo Void, and the
Coma Void. Those regions designated with "ZOA" appear to extend into the Zone of Avoidance. Right now
the simulation has a uniform density universe, but we will later include scaling based on these structures.
14

 The integrated ux can be determined for each galaxy:
MHI = 2:356  10 5 D 2
S
W 20 (A2)
 The measured line widths from the Arecibo Dual-Beam Survey are used to parameterize the distribution
of line widths for the galaxies. First, there is an upper limit to the line width for a given galaxy HI mass.
The upper limit was dened by a t to the data as:
W 20 = 211:765
logM 1524:706 (A3)
the lower limit to the line width is assumed to be 40 km s 1 but this is an adjustable parameter. The line
width distribution is dened by a Gaussian up to 100 km s 1 and by an exponential above 100 km s 1 as
follows:
P (x) = (A=
p
2)e (x x0 ) 2 =2 2
(A4)
P (x) = Ae a(x x0 ) (A5)
A=1,  = 100, x 0 = 100, a=0.007
 The S/N of the galaxy is determined from:
 = T sys [Jy]=
p
t
 (A6)
T sys =3.5 Jy (outer beams)
 = 5010 3 Hz (after Hanning smoothing)
The eective noise is:
N eff = v res (W 20 =v res ) 0:75 =(300=v res ) 0:25 (A7)
so the S/N is given by:
S=N = S
W 20 =N eff = (4:24  10 6 =)
MHID 2 (300=v res ) 0:25 W 0:75
20 (A8)
S=N = 2:7  10 4 t 1=2 MHID 2 (300=v res ) 0:25 W 0:75
20 (A9)
We use 8 as our galaxy detection limit for the simulations in the following sections. The mass function is dened
by:  =1.5,   = 0.005, and M  =7.610 9 .
Survey Sensitivity: We have looked at the sensitivity to galaxies in 12 seconds of observation considering
only the sensitivity of the outer beams (therefore there will be part of the survey that is deeper). We also have
not included the eect of averaging the polarizations in these calculations. Using this estimate of the sensitivities
we have found that we should get a S/N of 8 for:
 an MHI galaxy, width of 200 km s 1 at 180 Mpc (12,600 km s 1 )
 a 10 6 M  galaxy, width of 50 km s 1 at 3.8 Mpc (267 km s 1 )
 a 10 7 M  galaxy, width of 50 km s 1 at 12.1 Mpc (844 km s 1 )
 a 7x10 10 M  galaxy, width 200 km s 1 at 600 Mpc (42,000 km s 1 or the edge of a 200 MHz band)
 a 3.6x10 10 M  galaxy, width 200 km/s at z = 0:1 (429 Mpc, 30,000 km s 1 )
For a single driftscan (or two driven scans with half the time which would be preferable), interleaving scans to
obtain a Nyquist sampled map covering 34 ô of declination, these observations should take 350 days or 8400 hours.
This depth is probably useful for both the shallow survey and the ZOA work so at least for now we will combine
the two.
15

Galaxy Counts from Simulations: The simulation is presently for a uniform density universe, but it gives us
a look at the numbers of galaxies we might expect to detect. We detect 23,300 galaxies over the whole Arecibo
sky out to 40,000 km s 1 (assuming the mass function parameters from Rosenberg & Schneider 2002) and use
these results in the gured. Using the HIPASS values (Zwaan et al. 2003), the expected numbers are 35,300
due to a signicantly higher average density.
Fig. 1.| The distribution of expected galaxy detections as a function of HI mass. The numbers above the curve
on the low mass end of the histogram give the numbers of galaxies in the corresponding bins.
Figure 1 shows the distribution of the galaxies by their HI mass. The numbers above the curve on the low mass
end of the histogram give the numbers of galaxies in the corresponding bins. We expect to detect a handful of
source at around 10 6 M , but by 10 7 M , the statistics should be pretty good and above 10 7 M we will have
enough sources to study the shape of the HI mass function in dierent environments.
Fig. 2.| The distribution of expected galaxy velocities (distances where H 0 is assumed to be 70 km s 1 Mpc 1 )
for low mass sources in the sample.
Figure 2 shows the distribution of distances (expressed in velocity for H 0 = 70 km s 1 Mpc 1 ) for the low mass
sources. While the lowest mass sources will only be detected very near by, there will still be a signicant number
of 10 7 M sources detected beyond 20 Mpc (1400 km s 1 ) where velocity/distance confusion becomes somewhat
less signicant.
Figure 3 shows the full distribution of galaxy distances in the sample in order to demonstrate the population of
galaxies that we would expect to detect beyond z = 0:1. There are a signicant number of galaxies that should
be detectable at these large distances which will be important for mapping out large scale structure.
Comparison with HIPASS and HIJASS Surveys: An All-Arecibo sky survey provides 2 direct benets
over HIPASS and HIJASS with its improved spatial and velocity resolution. The higher spatial resolution will
be of benet because HIPASS and HIJASS science is complicated by the confusion of sources in the large beam.
The HIPASS follow-up needed at ATCA are enormous and therefore are limited to the higher ux sources. It
will be years before the sources are followed-up (if ever), so an Arecibo survey will not encounter as many issues
16

Fig. 3.| The distribution of expected galaxy velocities (distances where H 0 is assumed to be 70 km s 1 Mpc 1 )
for all sources.
with source confusion as HIPASS and will therefore be able to do science with the data without time consuming
interferometry follow-up. The higher velocity resolution of ALFA will be especially useful in detecting edge on
galaxies that have peak uxes near the noise limit - the edge of a double peak spectrum is much sharper with the
higher velocity resolution which should make it easier to automatically detect these sources (and detecting wide
things near the noise limit is not easy!).
The HIJASS survey faces similar problems to HIPASS in terms of source confusion and wider velocity channels.
HIJASS has an additional problem that there is very bad interference in the 4500-6000 km s 1 (within which
range a lot of the interesting large scale structure e.g., Pisces-Perseus). In addition, HIJASS is not scheduled to
observe any more in the Arecibo range (a 4x4 degree region in Virgo and a few other areas have been covered at
this point) for the next few years because they are concentrating on the region of the SDSS rst release.
Discovering Isolated HI Clouds: With any new survey there is always the chance of discovering new
objects. A blind survey like the one we are proposing has the potential to reveal new kinds of objects. One puzzle
for numerical modelers of galaxy formation is that the process has been so eôcient. Most of the gas seems to
reside in the big bright galaxies. Yet, QSO absorption line studies indicate many absorption line systems that do
not appear to be associated with bright galaxies. The Virgo survey will enable us to not only explore the faint
end of the mass function, but to also look for HI clouds not associated with optical sources.
HI Mass Function Studies: The shape of the HI mass function in the local universe and the cosmological
mass density of HI are important parameters for modeling the formation and evolution of galaxies. Blind 21
cm surveys are the only way to determine these parameters, but in the past these surveys have suered from
extremely poor statistics (50 { 300 sources) particularly relative to optical luminosity function surveys which
typically sample tens of thousands of galaxies. HIPASS was the rst 21 cm survey to have signicant galaxy
detection statistics ( 7000 galaxies), but there are limitations with the HIPASS data as discussed in x4.1. The
survey proposed here will detect  11; 000 galaxies between 20 Mpc and 145 Mpc where distance errors are
minimized and where galaxy detections are not limited to the most massive sources. With the mass function
well determined down to several times 10 6 M , we will be able to x the faint end much better than it has been
done up to this point. This survey will allow us to resolve the controversy over the faint-end slope of the HI mass
function (Zwaan et al. 2003, Rosenberg & Schneider 2002, Zwaan et al. 1997).
Previous studies of the HI mass function have concentrated on determining an \average" shape of the eld mass
function. However, we know that galaxy environment eects the galaxy content of galaxies through star formation,
tidal interactions, and merging. Because of these processes we would not expect the mass function to be the same
in all environments, but the relationship between the HI mass function and environment can not be studied with
the small samples that have been available. Only with the much larger number of galaxies that we expect to
detect in this ALFA shallow survey will we be able to look at the mass function in dierent environments and
begin to study how the processes which transform galaxies aects the HI distribution function. The combination
of the statistical study of galaxy populations with environment and the ability to map out the HI content of
galaxies with respect to high density regions like Virgo (see x4.6) will allow us to understand how gas in galaxies
17

is aected by environment in much greater detail.
Zone of Avoidance Studies: The obscuration due to dust and the high stellar density in our Galaxy varies
from place to place within the Milky Way. Overall, it blocks our optical view of the extragalactic universe over
about 20 percent of the sky, less in the infrared. This sky coverage limitation does not pose a problem for the
study of galaxies themselves, as there is no reason to believe that the population of obscured galaxies should
dier from those in optically unobscured regions. However, to understand the Local Group's motion requires
mapping the surrounding mass inhomogeneity, measured in practice by galaxy over- and under-densities, across
the entire sky, including the optically-obscured zone of avoidance (ZOA). Further, an accurate knowledge of the
mass distribution within our neighborhood is essential if we are to understand the dynamical evolution of the
Local Group from kinematic studies (e.g., Peebles et al. 2001), although previous surveys near the Galactic plane
indicate we are probably not missing any important local giant galaxies (Kraan-Korteweg, Henning, & Andernach
2000).
As rst demonstrated by Kerr & Henning (1987), galaxies that contain HI can be systematically found in the
regions of thickest obscuration and IR confusion, through detection of their 21-cm emission. Taking advantage of
the velocity information immediately available from the redshift of the 21-cm line, this technique also lls in the
ZOA in redshift space, valuable toward mapping the velocity eld in the ZOA.
The ZOA in the Arecibo sky cuts through some important known large-scale structures. The major known
overdensity probed will be the Pisces-Perseus supercluster. The main voids include the Local, Orion, and Taurus
voids, and the edge of the Monoceros void. At this sensitivity (see below) we anticipate uncovering many unknown
structures, as well.
Currently, we envision two possible basic survey strategies. First, we consider the style of the "shallow survey",
with 12 sec integration. Scientically, this is interesting, because it extends to the north the region of the ZOA
mapped by the Parkes multibeam ZOA project, and even with this "shallow", relatively short integration time,
goes deeper than Parkes ZOA (e.g. Parkes ZOA sees an M* galaxy to 70 Mpc, compared to the 180 Mpc
estimated for \shallow" ALFA) A shallow ALFA/ZOA survey would be enormously more sensitive than the
Dwingeloo Obscured Galaxies Survey, which covered this area but was sensitive only to very nearby objects, and
massive spirals to its survey limit of 4000 km s 1 (Henning et al. 1998, Rivers 2000). An integration time of 12
seconds with ALFA would uncover a galaxy with HI mass of only 610 8 M (linewidth of 100 km s 1 , with S/N
of 8) at the distance of the Pisces-Perseus supercluster (5000 km s 1 , or 71 Mpc, with H o =70 km s 1 Mpc 1 ).
This would allow for excellent mapping of the hidden structures in this rich area, including any backside infall,
very important for mapping the extent of over- and under-densities.
If ALFA/ZOA were conducted in the same style as the \shallow" all-Arecibo sky ALFA, then scheduling the
two surveys as one optimizes telescope scheduling. Also, we emphasize the desirability of the Nyquist-sampled
mapping envisioned for shallow ALFA, which is particularly important for accurately locating optically-hidden
sources in the ZOA. If ALFA ZOA is taken as the jbj  5 ô portion of the Arecibo sky, then the resulting 1000
square degrees would take 700 hours of the 8400 estimated for all-sky shallow ALFA.
Second, ALFA/ZOA could well take advantage of the very similar needs of the GALFA low-latitude survey. As
currently envisioned by the GALFA consortium, the Galactic HI Low Latitude Survey would cover the Galactic
plane region (jbj  5 ô ) in drift mode with Nyquist sampling, and 200 sec integration. Were ALFA ZOA to be
conducted simultaneously, we could reach sensitivities 4 times greater than the shallow survey, far surpassing any
other ZOA or all-sky HI survey. A 200 sec integration would nd galaxies at S/N = 8 with the low HI mass of
about 10 8 M at Pisces-Perseus, and 9  10 9 M at z=0.1.
Deep Surveying of Nearby Groups and Clusters: There are a handful of nearby groups and clusters
that we will be able to study in great detail with this survey, thereby providing the best information on gas-rich
galaxies in dense environments that we have to date. We discuss the data we will get in the Virgo Cluster below.
One of the most signicant groups that we will be able to study with this survey is the Canes Venatici I Group
at about 5 Mpc. Only the southern half of the group is in the Arecibo declination range, but it is extremely rich
in low mass HI systems. Given the relative orientations of the Arecibo strip and the Local Supercluster, there
is no other region where discovery rates would be expected to come close to what we'll nd in this region. The
group is near enough that we will be able to detect galaxies down to 10 6 M , thereby allowing us to study low
18

HI mass sources in high density regions.
In the same region of the sky as the Canes Venatici I Group are Canes Venatici II and Coma I at 10 { 20 Mpc.
At these distances we will not be able to detect the 10 6 M sources, but we will detect a large number of sources
at a few 10 7 M in these groups. We will be able to detect sources down to a similar level in the Virgo Cluster
as well (see next section).
Additionally, we will be surveying the regions of some other major overdensities including the Pisces-Perseus
supercluster, the Coma cluster, the A1367 cluster, and the Coma wall. There are also 20 additional groups
(smaller overdensities) at velocities less than 1000 km s 1 which we should be able to study in great detail.
Understanding the distribution of gaseous dwarfs in the Local Group is also an important question in under-
standing the formation and evolution of our own local overdensity. We will be able to survey the Local Group for
HI-rich dwarf galaxies.Several known Local Group galaxies, mainly dwarfs, will be observable with this survey
including: M33 (Sc; 850 kpc), Leo I (dSph; 205 kpc), Leo II (dSph; 270 kpc), And II (dSph; 680 kpc), And VI
(dSph; 775 kpc), LGS 3 (dIrr/dSph; 620 kpc), Peg DIG (dIrr/dSph; 760 kpc), IC 1613 (dIrrV; 715 kpc), and
Sextans B (dIrrIV-V; 1.3 Mpc; technically not part of the Local Group). The rst 4 of these dwarf galaxies
remain undetected in HI. Depending on their star formation history, we may expect to detect gas in the vicinity
of these galaxies, i.e. if the galaxy was recently forming stars we may nd its star formation fuel is still nearby
and only recently stripped from the dwarf. M33 and the latter dwarfs have already been found to contain HI, but
this survey will assess the full extent of the gas and determine if there are nearby high-velocity clouds accreting
onto these galaxies.
A 21cm Survey of the Virgo Cluster and its Surroundings: The Virgo Cluster is the largest nearby
grouping of galaxies and is the obvious place to study the individual and statistical properties of galaxies in
the cluster environment. The cluster is probably still assembling itself from a number of smaller sub-groupings
though in the central regions galaxy interactions will have been common because of the relatively short crossing
time  0:1 H 1
0 . Virgo is the nearest x-ray cluster with emission being detected as much as 6 deg from the
cluster center. This indicates that there is a substantial inter-galactic gas that the galaxies move through and
presumably this is what modies the gas content of cluster galaxies compared to those in the eld. Almost all
previously published work on the HI properties of the cluster galaxies has been obtained from pointed observations
of optically identied sources, here we propose to carry out a fully sampled HI survey of the cluster down to a
mass limit of  10 7 M . The entire cluster resides in the region observable by Arecibo.
 Cluster structure: Previous observations of Virgo indicate a complex velocity structure and a number
of possible infalling clouds (Binggeli et al. 1993). Given that the infalling population is likely to be gas-rich
compared to the established cluster population the best way of studying and distinguishing between the
dierent sub-groupings is to consider the velocity structure as a function of HI mass.
 Environmental dependence of the HI mass function: Optical luminosity functions show large
variations as a function of environment with steeper faint end slopes being in clusters (e.g., Lobo et al.
1997). Recently large numbers of faint cluster galaxies have been identied in the Virgo cluster (Sabatini et
al. 2003) leading to a cluster luminosity function which is consistent with standard CDM models, but very
dierent to the global luminosity function derived by 2df and Sloan. Galaxies evolve by converting gas into
stars and so an interesting way of determining the details of this evolution is to compare the luminosity
function and the HI mass function in dierent environments. Tantalizing evidence of an environmental
dependence comes from the Arecibo Dual-Beam Survey (Rosenberg & Schneider 2002) which indicates that
the HI mass function is much atter in the Virgo Cluster than in the eld. By considering the relationship
between the optical and the HI, we will be able to assess the relative importance of star formation and gas
stripping as means of consuming the gas.
 Beyond the cluster edge: An HI survey of the cluster would cover also its far edge, which is not well
dened, in particular the void behind Virgo. Recent high-resolution simulations in a at,  dominated
Universe (Gottlober et al. 2003 MNRAS, submitted) indicate that the voids in the L  galaxy distributions
contain low-mass halos. Their prediction is that a typical void with a diameter of  20h 1 Mpc would
contain  10 5 objects with mass greater than 10 7 M . Given that the Virgo survey would be limited in
sensitivity to  10 8 M for objects behind the Virgo cluster, we may expect to detect a few hundred such
19

objects. Their presence in the void, the HI mass function, and dynamical properties would pose signicant
constraints on modern cosmological models.
Mapping the the Distribution of Luminous and Dark Matter to z = 0:1: We will be able to detect
410 10 M galaxies out to z = 0:1 with this survey (if the hardware allows us to observe to z = 0:1). While this
is the high mass end of the HI mass function, these galaxies will allow us to trace out the large scale structure
at these distances inside and outside of the Zone of Avoidance providing an excellent picture of the large scale
features in luminous and dark matter out to z = 0:1.
High Velocity Cloud Searches: An All-Sky ALFA HI Survey will greatly improve our understanding of how
gas is accreted onto galaxies. Previous surveys of the high-velocity clouds (HVCs) of neutral hydrogen around
our Galaxy have been of substantially lower resolution (15.5 0 at best) and/or were unable to trace the connection
between HVCs and Galactic HI emission (Putman et al. 2002; Wakker & van Woerden 1991). An ALFA HVC
survey will trace important high-velocity structures, such as the Magellanic Stream and Complex C, at 5 times the
resolution of previous surveys. It will also be 8 times more sensitive to unresolved small clouds, or ultra-compact
HVCs (if any exist with central neutral column density above  10 20 cm 2 ). Several key questions with regard
to high-velocity clouds (HVCs) will be addressed with an ALFA survey, including:
 What is the origin of HVCs? Is there a link between HVCs and Galactic emission? How does our Galaxy
obtain fresh star formation fuel? Are HVCs infalling onto the Galaxy (e.g. Tripp et al. 2003), possibly
as the remnants of satellites, or does the gas appear to be out owing in some cases (de Avillez 2000)?
Previous surveys have either not been suited to studying both Galactic emission and HVCs, or the spatial
and velocity resolution of the survey was insuôcient.
 Do HVCs show evidence of interaction with a halo medium? Is there an interface between the cold clouds
and a hot halo medium? This question has become especially important of late with the nding of OVI,
OVII, and OVIII absorption that appears to be associated with a hot halo medium (Sembach et al. 2003;
Rasmussen, Kahn, & Paerels 2003). The OVI absorption often directly traces HVCs detected in HI and
appears to be produced via collisional processes (Sembach et al. 2003). The resolution and sensitivity
of this survey will allow for a close examination of the head-tail HVCs (e.g., Bruns et al. 2000) and the
environment of a large percentage of the OVI absorbers. Recent numerical simulations (Quilis & Moore
2001; Konz, Bruns & Birk 2002) demonstrate that interacting HVCs show a low-intensity tail of diuse HI
gas and a cold compression front at the leading edge. This compression front shows structures on arcminute
scales, well observable with Arecibo.
 How many ultra-compact HVCs are there? Previously detected CHVCs and mini-HVCs exhibit a strong
correlation between peak column density and angular size such that clouds as small as the Arecibo beam
would have peak column densities < 10 18 cm 2 (Homan & Salpeter 2002) and therefore would not be
detectable in a shallow survey. However, the large number of "missing satellites" might be extremely small
angular-size, low mass, gaseous objects with detectable peak column density (Moore et al. 1999; Klypin et
al. 1999). An 8 detection of an ultra-compact HVC at 10 kpc, 100 kpc, and 1 Mpc would be 3.5, 350,
and 3:5  10 4 M , respectively. If there is a substantial "missing satellite" population, we would expect to
detect a large number of these objects in the general direction of M31/M33. A population of M33 satellites
would have a typical angular diameter of about 5 arcmin, if the known (Galactic) CHVCs have distances
of the order 100 to 200 kpc. Arecibo is the perfect instrument to detect such M33-CHVCs.
An All-sky ALFA survey will observe several specic high-velocity complexes which are of interest for answering
the above questions. The tail of the Magellanic Stream, a high-velocity cloud which is unique in that its origin
is clearly know as originating from the Magellanic Clouds, is included in the survey. The tail of the 100 ô long
Magellanic Stream fans out and appears to break up into many small clumps, possibly where it is being over-
whelmed by the Galaxy's halo (e.g., Stanimirovic et al. 2002; Putman et al. in preparation). The structure of
the Magellanic tail may provide important clues to how satellites lose their gas as they spiral into the Galaxy.
The Southern edge of the giant high-velocity cloud, Complex C, is also included in the survey. Since we will be
observing Galactic velocities as well, we can trace if and how this complex is feeding star formation fuel into our
Galaxy (e.g., Tripp et al. 2003). In addition, the anti-center high-velocity complex, which may represent the
gaseous trail of the Sgr dwarf galaxy (Putman et al. 2003), will be observed.
20

High-velocity clouds have a typical linewidth of 25 km s 1 (e.g., de Heij, Braun & Burton 2002). Many HVCs
show a two-component prole, with cores with very small linewidths ( 2-10 km s 1 ) and extended envelopes
( 20-30 km s 1 ) (Wakker & van Woerden 1997). The velocity resolution of the E-ALFA survey will be 10 km s 1
after Hanning smoothing. This resolution will be suôcient to investigate the kinematic structure of the clouds'
extended envelopes, especially of the larger complexes. The data from the Galactic ALFA survey will have 1
km s 1 resolution. This will allow us to look at the detailed kinematics of some of the brighter clouds and cores.
Searches for Galaxies Near Low Redshift Lyman- Absorbers: The spectra of QSOs and AGN
are strewn with absorption features from intervening low column density HI along the line of sight, yet their
relationship to galaxies is still uncertain. Are these systems related to small, low surface brightness (LSB)
galaxies (Impey & Bothun 1997) or to the extended halos of high surface brightness galaxies (Lin et al. 2000;
Lanzetta et al. 1995)? Do the absorbers trace laments of primordial material which are correlated with galaxies
because they are overdensities in the same large scale structures (Dave et al. 1999; Stocke et al. 2000)?
At present there are only a handful of low redshift Lyman- absorbers known and the regions around the higher
column density systems (NHI >  10 14 ) have now been surveyed using Arecibo, the VLA, Parkes, and the Australia
Telescope Compact Array. However, HST (the only instrument capable of identifying low redshift Lyman
absorbers), has been extremely restricted in the QSOs it can study because virtually all of them are too faint
to allow detailed HST spectroscopy in moderate observing times. This situation is about to change dramatically.
In mid-2004 the Cosmic Origins Spectrograph (COS) will be placed aboard HST, allowing high SNR (R=22,000)
spectra of 17th mag QSOs in only a few orbits; i.e., 10-30 times the throughput of the present spectrograph at
comparable resolution. This new spectrograph will dramatically expand the available sample of QSOs that can
be observed, thereby producing a new list of nearby absorption line systems. This survey will be an excellent
starting point for determining whether there are gas-rich galaxies near the lines of sight to these absorbers.
Absorption Line Studies at High Redshift: The background source counts for this work at 1.4 GHz gives:
 190 sources/steradian (665 sources in the Arecibo range) brighter than 1 Jy
 840 sources/steradian (2940 sources in the Arecibo range) brighter than 0.4 Jy
 5600 sources/steradian (19600 sources in the Arecibo range) brighter than 0.1 Jy
We still have to determine what column densities this will allow us to get to. However, it does seem that by
looking in absorption, this survey has the potential to detect galaxies that are not detectable by other means.
Find Rare OH Megamasers near z = 0:25 OH Megamasers (OHM) are powerful line sources observed
in the L band, arising from the nuclear molecular regions in merging galaxy systems. Approximately 100 such
sources are known to date. Several of them are observed to have spectral features variable in strength, an eect
which allows keen insights in the source structure and physics. Observations of OHMs hold the potential to being
able to trace the merger history of the Universe. ALFA surveys would have an important impact in the detection
of such systems at intermediate redshifts. We estimate below the detection rates of OHMs for three possible
ALFA surveys.
For a 12-second integration, the expected rms noise in a 150 km s 1 channel (the average OHM line width) is
roughly 0.67 mJy, and a 5 detection would be 3.4 mJy. This corresponds to an OH luminosity range of roughly
log LOH = 2:85{3.26 L for observations spanning 1320{1420 MHz (z = 0:263{0.174). A survey of the entire
Arecibo sky with an upper log LOH cuto of 4.4 should detect about 130 OH megamasers (we use the OH
luminosity function determined by Darling & Giovanelli 2002). Note that this estimate is optimistic because it
does not include the tendency for luminous OH megamasers to be broad.
OH detection rates would benet from extended spectral coverage. An additional 100{200 MHz would enhance
detection rates signicantly, despite the higher redshifts and dimming of sources.
A Survey for the Community: While the above science will guarantee exciting results from this survey,
much of the exciting science that will be done can not be summed up here. There will be many more uses for this
survey by the community as a complement to ongoing work. This will be the deepest, highest resolution survey
over this area of the sky. By virtue of its unique depth and resolution it will be a major resource. The survey will
21

be a starting point for anyone wanting to know the HI properties of a galaxy in this velocity range in this part
of the sky, it will be a place to start a search for galaxies around low redshift Ly- absorbers, and much more.
As the rst resource for this kind of work, it may also increase the interest in the larger community in the use of
Arecibo for follow-up work on regions where the survey is not deep enough to answer the questions at hand.
REFERENCES
Binggeli, B., Popescu, C. C., Tammann, G. A. 1993, A&AS, 98, 275.
Bruns, C., Kerp, J., Kalberla, P. M. W., & Mebold, U. 2000, A&A, 357, 120.
Darling, J. & Giovanelli, R. 2002, ApJ, 572, 810.
Dave, R., Hernquist, L., Katz, N., & Weinberg, D. H. 1999, ApJ, 511, 521.
de Avillez, M. A. 2000, Ap&SS, 272, 23.
de Heij, V., Braun, R., & Burton, W. B. 2002, A&A, 392, 417.
Grebel, E. K., Gallagher III, J. S., & Harbeck, D. 2003, AJ, 125, 1926.
Henning, P.A., Kraan-Korteweg, R.C., Rivers, A.J., Loan, A.J., Lahav, O., & Burton, W.B. 1998, AJ, 115, 584
Homan, G.L., & Salpeter, E.E. 2002, in The Outer Edges of Dwarf Irregular Galax-
ies, 2002 Lowell Workshop On-line Proceedings, eds. D. Hunter and S. Oey
(www.lowell.edu/Workshop/Lowell02/Proceedings/poster/homan.htm).
Impey, C. D. & Bothun, G. D. 1997, AR&A, 35, 267.
Kerr, F.J., & Henning, P.A. 1987, ApJ, 320, L99.
Klypin, A., Kravtsov, A. V., Valenzuela, O., & Prada, F. 1999, ApJ, 522, 82.
Konz, C., Bruns, C. & Birk, G.T. 2002, A&A, 391, 713.
Kraan-Korteweg, R.C., Henning, P.A., & Andernach, H. 2000, eds. ASP Conf. Ser. 218, Mapping the Hidden
Universe: The Universe Behind the Milky Way - The Universe in HI (San Francisco: ASP)
Lanzetta, K. M., Bowen, D. B., Tytler, D., & Webb, J. K. 1995, ApJ, 442, 538.
Lin, W. P., Borner, G., & Mo, H. J. 2000, MNRAS, 319, 517.
Lobo et al. 1997, A&A 317, 385.
Moore, B. et al. 1999, ApJ, 524, L19.
Peebles, P.J.E., Phelps, S.D., Shaya, E.J., & Tully, R.B. 2001 ApJ, 554, 104
Putman, M.E., Staveley-Smith, L., Freeman, K.C., Gibson, B.K., & Barnes, D.G. 2003, ApJ, 586, 170.
Putman, M.E., et al. 2002, AJ, 123, 873.
Quilis, V. & Moore, B. 2001, ApJ, 555, 95.
Rasmussen, A., Kahn, S.M., & Paerels, F. 2003, in The IGM/Galaxy Connection, eds. J. Rosenberg & M. Putman,
Kluwer Academic Publishers, 109.
Rivers, A.J. 2000, Ph.D. thesis, Univ. of New Mexico
Rosenberg, J. L. & Schneider, S. E. 2000, ApJS, 130, 177.
Rosenberg, J. L. & Schneider, S. E. 2002, ApJ, 568, 1.
22

Sabatini, S., Davies, J., Scaramella, R., Smith, R., Baes, M., Linder, S. M., Roberts, S., Testa, V. 2003, MNRAS,
341, 981.
Solanes, J. M., Manrique, A., Garc  ia-Gomez, C., Gonzalez-Casado, G., Giovanelli, R., & Haynes, M. P. 2001,
548, 97.
Stanimirovic, S., Dickey, J. M., Krco, M., & Brooks, A. M. 2002, ApJ, 576, 773.
Sembach, K., et al. 2003, ApJS, 146, 165.
Spitzak, J. G. & Schneider, S. E. 1998, ApJS, 119, 159.
Stocke, J. T., Shull, J. M., Penton, S. V., Gibson, B. K., Giroux, M., L., McLin, K. M. 2000, astro-ph/0009190.
Tripp, T. et al. 2003, AJ, 125, 3122.
Wakker, B.P. & van Woerden, H. 1997, ARA&A, 35, 217.
Wakker, B.P. & van Woerden, H. 1991. A&A, 250, 509.
Zwaan, M., Briggs, F. H., Sprayberry, D. & Sorar, E. 1997, ApJ, 490, 173.
Zwaan, M. et al. 2003, AJ, 125, 2842.
This preprint was prepared with the AAS L A T E X macros v5.0.
23

B. Scientic Justication: Ultra- Deep Pencil{Beam Survey
Introduction: An ultra-deep survey of an very small patch of sky is proposed. The purpose of this is to probe
the gas content and evolution of the most distant galaxies detectable in HI at Arecibo and to investigate the
properties of nearby galaxies, the IGM and the Cosmic Web at extremely low column densities.
Galaxy Evolution across Cosmic Time: The rate of conversion of gas into stars is one of the fundamental
quantities which describe the evolution of galaxies and remains a diôcult measurement to make for all but the
most nearby galaxy populations (Madau et al. 1996, MNRAS 283, 1388; Haarsma et al. 2000, ApJ, 544, 641).
At redshifts between zero and unity, the global comoving star-formation rate appears to increase by an order of
magnitude. Models which attempt to explain the global evolution of galaxies (e.g. Pei & Fall 1999, ApJ, 522,
604) over this redshift range, predict a corresponding sharp increase in the cosmic comoving gas density, with

HI / (1 + z) 3:2 . However, the limited data we have on the gas density at these redshifts from ultra-violet
observations of damped Lyman- (DLA) systems (e.g. Storrie-Lombardi & Wolfe 2000, ApJ, 543, 552) suggests
that the redshift density of absorbers with NH > 2  10 20 cm 2 only increases as dNDLA=dz = 0:05(1 + z) 1:1 ,
which is consistent with constant comoving gas density, if the absorbers do not evolve. This discrepancy may be
a limitation of the observations or with the models, or with both. Lyman- observations in particular are diôcult
to make at z < 1:6 when the line is no longer shifted into the optical regime, making suitably bright background
probes harder to nd. In addition, the expansion of the Universe makes DLA systems rarer at low redshift.
Furthermore, extinction of the background probe by a foreground galaxy associated with a DLA system is a
source of possible bias in the measurement of gas density. A much better method, where possible, is to observe HI
in emission and to thus directly measure the change in cosmic gas density, without appealing to model-dependent
interpretations of the data.
ALFA makes it feasible, in principle, to observe down to frequencies of 1225 MHz, corresponding to a redshift of
0.16, corresponding to a lookback time of  15% of the age of the Universe. Whilst this is a fairly limited redshift
in a cosmological context, it is suôcient to be able to distinguish between the evolutionary scenarios above. Indeed
other relatively shallow surveys, such as IRAS and 2dF, have found detected signicant evolution of luminosity
and comoving number density over similarly modest redshift ranges (for IRAS galaxies, n(z) / (1 + z) 3 , Takeuchi
et al. 2003, ApJ, 587, L89). However, signicant integration times, careful control of observational parameters,
a low system temperature, and mitigation of interference are required.
The Cosmic Web: Evidence from QSO absorption line studies shows that the intergalactic medium is popu-
lated with low-density ionized gas, forming an important, possibly dominant reservoir of the Universe's baryons
(Shull et al. 1996, AJ, 111, 72; Penton et al. 2000, ApJ, 544, 150). The structure of the clouds is still a topic
of debate: theoretical modeling suggests that the absorption arises in lamentary structure (`the cosmic web'),
whereas some observers believe that the clouds are associated with the extreme outer halos of individual galaxies
(Chen et al. 2001, ApJ, 556, 118). The largest cross sections would be provided by highly ionized halos character-
ized by CIV absorption, followed by mainly neutral Lyman limit/MgII absorbers of HI column densities greater
than a few 10 17 cm 2 .
The distribution of the number of absorbers along a redshift path as a function of column density (the f(N)
distribution) is nearly a power-law with slope 1:5 over 10 orders of magnitude. Deviations from the power-law
occur at column densities where clouds become optically thin to ionizing radiation (around a few times 10 17 cm 2 ).
This region of column densities is poorly understood in the optical because the Lyman- line lies on the at part
of the curve-of-growth, and in the radio because typical 21-cm observations do not reach this column density
sensitivity. The deviation from the power-law might be related to the ionization edge in the gaseous disks of
galaxies (Maloney 1993, ApJ, 414, 41). Our proposed observations reach a 6  column density sensitivity of
2  10 16 cm 2 per 5 km s 1 for gas lling the beam. The Doppler parameters of the clouds are typically 30 km
s 1 (Shull et al. 2000, ApJ, 538, 13), which means that they would be resolved with the proposed 5 km s 1
resolution. An ultra-deep `blind' ALFA survey would therefore be ideal for exploring this column density regime.
A deep survey (HIDEEP) has previously been undertaken with the Parkes multibeam system (Minchin et al.
2003, MNRAS, submitted). The ALFA survey has four signicant advantages:
1. Depth. The proposed ALFA ultradeep survey has an integration time per beam which is a factor of 30
greater than HIDEEP, giving a column density sensitivity, for a source lling the beam, more than ve
24

times better.
2. Higher resolution. The ultra-deep survey will have much better (> 4) spatial resolution than HIDEEP,
allowing the same column density limit to be applied to sources at > 4 the distance. It also allows the
positions of sources to be found much more accurately.
3. Higher sensitivity to unresolved sources. HIDEEP runs out of sources at around 10 8 M , and has nothing
to say about low column-density sources below this limit. For the same volume, the ALFA ultra-deep
survey will detect objects with 250 lower column density, in the case of unresolved sources with diameters
estimated by other (optical, IR) methods..
4. Higher velocity resolution. The ultra-deep survey will be able to detect sources with velocity widths similar
to QSO absorption-line systems, which would only contribute to a single channel in HIDEEP.
Frozen Disks: The cosmic ultra-violet background is believed to give rise to ionization edges in the low
column-density outer regions of galaxies (as mentioned above). However, an alternative explanation due to
Disney & Minchin (2003, in The IGM/Galaxy Connect, eds. Rosenberg & Putman, publ. Kluwer Academica,
NL, p. 305) is due to the low column-density Hi cooling to the temperature of the CMB. Such gas in `frozen
disks' would still be visible in absorption, thus contributing to the Lyman limit systems and the Lyman- forest,
but would not be seen in emission. If Lyman-limit or Lyman- systems are detected in the ultra-deep survey, it
would demonstrate that the cut-o seen in galaxies is due to ionization, not `frozen disks'. If no such systems are
seen, then the survey would lend weight to the `frozen disk' hypothesis.
Observational Goals: A fundamental requirement of the ultra-deep survey is the ability to survey a large
volume to the highest practical redshift (z  0:16). For galaxy evolution studies, and in order to sample galaxies
which contribute most to the mass function at z = 0, we need to be sensitive to M  and sub-M  galaxies in the
range 10 9 10 M . Simple arguments suggest that a sample of  40 galaxies would be appropriate to distinguish,
at the 99% condence level, between models of the
form
HI / (1 + z) p , where p = 0; 3. With a linewidth of 200
km s 1 , a galaxy with MHI = 10 9:5 M has an integrated ux of 0.04 Jy km s 1 and a mean ux density of only
0.2 mJy.
Observing Parameters: The detection of nearby low column density objects and distant galaxies requires
long integration times. A possible survey that could achieve the goal of 40 galaxies at the highest emission
redshifts possible with ALFA will take 70 hr/beam, would have an rms of 0.05 mJy/beam (260 times deeper than
HIPASS) and would survey 0.36 deg 2 , corresponding to 8000 Mpc 3 , in 1000 hrs. The observations would be taken
at night in driftscan mode.
Hardware Issues:
 Wide bandwidths (200 MHz) are required for this survey to be eective. A wide band will give us a
simultaneous view of galaxies from z =0 to 0.16. Coverage at high-z is necessary to study galaxy evolution.
Coverage at low-z is necessary to obtain the highest column-density sensitivities for objects of a xed size. It
will also to provide a zero-point for the low-z HI mass function, though it is likely that other ALFA surveys
could do this more eectively. The 200 MHz band is not required to be in a single contiguous band. In
fact, in the case where limited sampler resolution (i.e. the number of sampler levels) is available, a separate
low-frequency band (where much of the interference lies) might be desireable.
 A resolution of 5 km s 1 is desired. More might be useful for interference mitigation. This corresponds to
8192 channels, or greater, over 200 MHz.
 Observations at high redshifts will lie near strong interference lines from radar installations. To mitigate
against such interference, this project needs a spectrometer with good dynamic range. Currently, we are
informed by NAIC sta that between 8 and 12 bits are required of the sampler and spectrometer in order
that emissions from strong interferers do not disrupt data at other frequencies.
 The primary interference lines may be able to be removed from the data before or after correlation using
a reference horn. We'd like the use of such techniques to be investigated, and we may be able to carry
out proof-of-concept observations at Parkes. However, the spectrometer will need two extra inputs and the
ability to correlate between 3 and 6 times as much data as for normal autocorrelation mode.
25

 Radar emissions are impulsive, and an extra degree of immunity against them is possible using blanking
techniques. So we would like blanking to be possible at the 1 ms level. Blanking could be automated in the
spectrometer hardware, or enforced by applying the phase and period of known radar transmitters.
 Knowledge of accurate calibration factors and beam shapes over a range of zenith angles will be required.
Although we propose a drift scan survey, the observing mode will involve chasing the eld over a range of
hour angles.
26

C. Information on RFI at Arecibo in the 1230-1430MHz Bandwidth
As emphasized previously, E-ALFA science will be dependent on and possibly limited by our ability to suppress,
mitigate and excise RFI. The need for considerable eort in the realm of RFI mitigation techniques and hardware
is strongly emphasized.
In this section, we include some relevant information on the RFI environment at Arecibo in the range of frequencies
covered by ALFA. This information will be updated as modications are required and as information becomes
available.
Table C.1 US Frequency Allocations
Frequency Primary Secondary
1215{1240 MHz Radiolocation { Government
Radionavigation Satellite { Government
1240{1300 Radiolocation { Government only Amateur { Nongovernement
1300{1350 Aeronautical Radionavigation { Mixed Radiolocation { Government
1350{1390 Fixed, Mobile, Radiolocation { Government
1390{1400 Fixed, Mobile, Radiolocation { Government
1400{1427 Radio Astronomy { Mixed
Earth Exploration Satellite { Mixed
Space Research (Passive) { Mixed
1427{1429 Space Operation { Mixed Fixed { Nongovernment
Fixed, Mobile { Government Land Mobile { Nongovernment only
1429{1432 Fixed, Mobile { Government Fixed, land mobile { Nongovernment
1432{1435 Fixed, Mobile { Government Fixed, land Mobile { Nongovernment
1435{1530 Mobile (Aeronautical Telemetering) { Nongov.
Future RFI at Arecibo
 The Galileo Project: (ESA's next generation global navigation system). This will be operating in
the ranges of 1164-1300 MHz & 1559-1610 MHz. The current plan is to begin deployment of the satellites
in 2006 and have al 30 satellites (full global coverage) by 2008. Further information is available online at
http://europa.eu.int/comm/dgs/energy transport/galileo/programme/phases en.htm
 GPS{L2: GPS L2 is at 1227.60 MHz, just outside the ALFA band. It is switching its usage from military
to military + civilian over the next eight years. The GPS L2 signal has a 1.023MHz clock rate which will repeat
every 20ms or 1.5s (depending upon the signal being sent). GPS satellites that will transmit the new L2-signal will
be launched from 2003 onward. By 2008, there will be 18 modernized so called Block IIR satellites in orbit. Full
operational capability with about 28 satellites will be reached three years later. Further information is available
online at www.navcen.uscg.gov/gps/modernization/ModernizedL2CivilSignal.pdf
or www.geoinformatics.com/issueonline/issues/2002/september 2002/pdf september/06 07 dejong.pdf
 Ultra-Wide Band Devices: UWB devices are now permitted by the FCC. None of the devices currently
allowed will fall within the ALFA bandwidth. More information is available in the FCC ruling document available
online at http://www.naic.edu/astro/RXstatus/Lnarrow/fcc UWB.pdf
27

Common, known RFI seen at Arecibo in the frequency range 1220{1520 MHz
Frequency [MHz] Occupancy Strength 1 Timing BW Identication
1217, 1227.5,
1265.0, 1313.0 { { { { Pico del Este
1222.32, 1231.28,
1240.24,
1249.2,1258.16,
1267.12, 1276.08,
1285.04, 1294.00,
1302.96, 1311.92,
1320.88, 1329.84,
1338.80, 1347.76,
1356.72, 1365.68,
1374.64, 1383.60,
1392.56
Always in some
mode 2
{ 400s+51s { Puntas Salinas Frequency
Hopping Radar
1241.7, 1246.2,
1256.7
{ { 400s+51s { Aerostat
1270.9/1289.9
Occasional (and
only one of the two
freq. at a time)
{ 5s, 12s
rotation
{ FPS20-93a radar located in
Ramey
1324,1340 Occasional (only
during war games) { 5s, 12s
rotation
{ Naval "landing system"
1330 & 1350 Always { 1s pulse,
12s rotation
{ FAA Radars (Pico del Este)
1366.3,1382.7 Occasional (only
during war games) { { { Naval "landing" system
1370.1,1387.3 Occasional (only
during war games) { { { Naval "landing" system
1381.1 { { { { GPS L3 satellites
28

NOTES
(1) Many of the military radars may no longer be active due to the shutdown of military facilities. A list of the
RFI sources which may be removed from the spectrum will be added as it becomes clear.
(2) The PRANG (Puntas Salinas) Frequency Hopping Radar: This radar runs in the range from 1220 - 1400 MHz
and has the potential for completely killing most HI observations at Arecibo. Fortunately, through considerable
work by the RFI folks at Arecibo, an agreement has been made between us and the folks at Puntas Salinas. If a
request is put in in advance, Puntas Salinas can restrict the frequencies used to just two of a given pair (to be
listed later).
(3) Columns designated by a dash ({) will be lled in.
29

D. Detailed Lists of Data Products and Relevant Software Tasks
Table D.1 Level I Data & Software Tasks
Item Requirements/Specs needed
Data Acquisition:
User Interface (GUI) Observing modes, displays, remote capabilities
Telescope Control Motions, rates
Backend Control Spectrometer setup, dump times, blanking
Radar Blanking Multiple?
Data Recording Dump times, formats, headers
Bandpass Calibration Schemes per observing mode
Position Registration Encoders, catalog Xreference, pointing info
Beam Pattern Sidelobe/comalobe per beam as fn (AZ,ZA,freq)
Record Keeping What info?, access?
Raw data Archive Capacity, format, location
Basic documentation E-ALFA specic + general
On-Line Monitoring:
Monitor display
Statistics tools
RFI id & removal Cross id with hilltop, known r
NVSS cross reference Click to get NVSS id
1beam bandpass Click to get a bandpass
1beam spectra Click to get a spectrum
1beam coordinates Click to get coordinates
Quick-look T-F maps Time-frequency domain
Quick-look X-Y maps Sky maps
Level I Processing:
Bandpass subtraction
Continuum subtraction
Flux calibration K to Jy scale
Higher level RFI excision Pipeline processing (auto ag)
Astrometry Real positions on the sky
Gridding Scheme development
Level I CLEANing First pass continuum removal
Level I spectra Individual spectra in Jy-km/s per beam
Level I maps & cubes "Dirty" maps
Continuum catalog Flux in Jy/beam at position
Data validation process
Level I access tools
Level I documentation
Level I software export available for small project ALFA users
Level I archive Contents? location? access?
30

Table D.2 Level II Data & Software Tasks
Item Requirements/Specs needed
Level II Processing:
Level II CLEANing sidelobe/comalobe deconvolution?
Level II spectra calibrated, at any position/nearest grid point
Level II maps
Level II cubes
Data validation statistics/reality checks
Level II archive archived at/through NAIC
Level II access tools access through NAIC
Level II documentation provided by E-ALFA to NAIC
Level II software export available for small project ALFA users
Portal to NVO in collaboration with NVO
Table D.3 Level III Data & Software Tasks
Item Requirements/Specs needed
Detection toolkit:
Baseline uctuations
Noise characteristics
Sidelobe "cleaning" schemes
Object nding algorithms SExtractor, Xcorrelation, etc.
Template tting
Injection of "fake sources"
Extended sources vs. point sources
Detection documentation
Data validation process
HI source catalog Made available to the public upon publication
General science toolkit:
Survey simulator (RG already has some)
ALFA simulator Simulates what ALFA will see, including multibeaming aspects
Overlay tool e.g. NVSS, DSS, SDSS, 2MASS
Catalog Xreference tool ditto
Toolkit documentation ditto
31

E. Assignment of Responsibilities
In this section, we detail a proposed assignment of responsibilities as summarized in Section 7. This list, when
nalized, could serve as a basis for a memorandum of understanding between NAIC and the survey team(s).
NAIC responsibilities:
1. establishing the processes that will govern other activities, including but not limited to:
 Proposal guidelines and schedule
 Proposal review process and criteria
 Project readiness review
 Observing time allocation and schedule
 Survey progress review and criteria
2. Facilitating the formation and activities of survey groups, including but not limited to:
 Supporting ALFA web-based information dissemination
 Organizing consortia workshops
 Facilitating cross-consortia (X-ALFA) discussions
 Providing on-site educational opportunities
 Organizing X-ALFA outreach activities
3. Providing the hardware and software needed to acquire data, including but not limited to:
 Telescope control and data aquisition software, record keeping tools etc.
 On-line monitoring displays, statistical tools, quick-look tools
 Beam characterization, calibration tools and information
 RFI suppression/mitigation, radar blanking schemes, real-time RFI removal
 Standardized data formats, software for export to standard SDFITS
4. Providing the archive for the raw data
5. Undertaking the rst phase of data processing, including but not limited to:
 Bandpass calibration and subtraction
 Astrometric registration and gridding tools
 Level I (rst pass) spectra, at but \dirty" maps and data cubes
 Continuum catalog
 Level I data archive
6. Verifying Level I data quality validation
7. Establishing the process for Level II data quality verication
E-ALFA responsibilities:
1. Specifying the hardware and software needs and how they will be contributed by the consortium
2. Specifying the optimal observing strategies
3. Planning and overseeing the observations (with assistance of NAIC sta, by agreement)
4. Assisting NAIC in Level I data processing
5. Developing, and providing to NAIC, the necessary software for Level II data processing
6. Undertaking the second phase of data processing, including but not limited to:
 Fully continuum cleaned maps
 Deconvolved spectral line maps (when appropriate)
 Robust, gridded data cubes
 Calibrated spectra at nearest grid point
 HI source catalog
 Data product access tools
32

7. Verifying Level II data quality
8. Providing Level II data products to NAIC (or designate institution)
9. Prompt publication of scientic results, HI catalogs
Shared NAIC and E-ALFA responsibilities:
1. Outlining a mutually agreeable framework for the collaboration with a clearly delineated assignment of
tasks, delivery schedule and evaluation criteria
2. Insuring that the observations are carried out eectively and eôciently
3. Obtaining NSF support for the surveys for U.S. participants
4. Identifying mechanisms for individuals at non U.S. institutions to participate
5. Specifying the format of standardized data products
6. Archiving the Level II data products
7. Developing a portal to the NVO
8. Establishing mechanisms for graduate thesis work
9. Establishing educational opportunities for undergrads, grads, postdocs
10. Developing educational materials for K-12, undergrads, grads
11. Providing public information and outreach materials
33

F. Organizational Matters, Working Questions/Proposals
In establishing a framework for the E{ALFA consortium, the following items need to be considered carefully in
light of the principles and objectives discussion in Sections 1 and 6.
A. Consortium denition:
1. What constitutes a \consortium"?
 One large, open group interested in all E-ALFA surveys
{ What about piggyback surveys with other X-ALFA groups?
{ How is eort allocated/rewarded?
 Hierarchical structure
{ Each E-ALFA survey is undertaken by a separate team
{ Membership in a team constitutes membership in \consortium", but rules specied by teams.
B. Consortium membership:
1. What constitutes \participation in a consortium"?
 Personal eort
 Commitment of additional resources
 Timescales for commitment
 Return on commitment
2. What are rules for \joining"?
 Pre-survey
 Post-survey
 Data/software access groundrules
 \Buy-out" options
3. What are rules for \continuing membership"?
 Can people be kicked out?
 On what grounds?
 Who decides?
 What mechanism for con ict resolution
C. Consortium/team leadership
1. Function of leadership
 Survey design and planning
 Resource planning and identication
 Survey management
 Oversight and accountability
 Return on commitment
2. Rules for electing leadership
 Nomination process
 Election process
Some (hypothetical) questions whose answers must be clear from the proposed framework include:
1. Suppose a consortium forms that expresses its intent to propose to undertake a specic ALFA survey, but
another group, not associated with the consortium for whatever reason, proposes to undertake the same
survey. Is the rst group guaranteed the telescope time and/or access to NAIC resources? Is the second
group precluded from submitting a proposal? Who decides who gets the time, and based on what criteria?
34

2. Suppose an individual joins the consortium at the outset and plays a major role in the survey and proposal
planning and execution (for example, by running simulations, developing deconvolution software, providing
computer hardware, executing the observing program etc) but then, for whatever reason, does not partic-
ipate in nal data product delivery or science analysis and is not involved in the consortium later. What
commitments are made to this individual by the consortium for early involvement in the project? Who
decides? Who insures that commitments are carried forward?
3. The early phases of the project (from now until late 2005) are unlikely to produce scientic results. It
thus seems reasonable that it might be unwise for some parties such as untenured faculty, current senior
graduate students and current postdocs, to spend signicant fractions of their time in the tasks of software
and algorithm development and other \investment" activities. How will be the balance be struck between
adequate return for those who do undertake the groundwork and the later involvement of those who do not?
4. Suppose a survey project starts but after six months, several participants have not fullled their commit-
ments, thereby putting the entire consortium at risk of not delivering data/software products on schedule.
How is this handled? By whom?
5. Faculty at undergraduate colleges and other primarily-teaching institutions may not be able to commit large
amounts of time, particularly during the academic year. How can they and their students be involved given
such limitations in time commitment, and possibly, hardware resources and nancial support?
6. There can be signicant imbalance, particularly in the United States, between scientists at institutions
supported on umbrella grants and those who receive support through individual investigator grants with
regard to the availability of funds for travel, hardware resources and system support, student support,
overhead and fringe benet costs, etc. Are there mechanisms available through NAIC and/or the consortium
concept that can alleviate or minimize the potential imbalance, particularly for the full duration of a survey
(planning to analysis)?
F.1. Proposed Organizational Structure
A proposal for a possible organizational structure has been put forward, summarized in the accompanying diagram
and governed by the rules below. At this stage, it does not answer many of the questions posed above but serves
as a strawman for the purpose of discussion.
NAIC
Science Initial
Data Products
Detection
Deconvolution
&
Strategies
&
Science
Follow-up Funding
Justification
Strategies
Manpower Details
Synergies Observing details
First look
Data format
Calibration
Archiving
Reduction software
Related science
Confirmation
Steering Committee
ALFA Project
Manager: TBD
Coordinator: Desh
ALFA Scientific
POC:
Chris Salter (Interim)
Outreach
ALAC: Riccardo Giovanelli, co-chair
Daniel Altschuler, co-chair
Miller Goss
Rich Kron
Joel Weisberg
1. Steering committee comprised of three people who are voted in by all active consortium members. The
chair of the steering commitee rotates every six months. After serving as chair, that member steps down
and a new commitee member is elected. (This results in the rst two chairs serving abbreviated terms, but
allows for continuity within the commitee itself.)
2. All other commitees are open to volunteers. The leader of each comittee is elected by vote from the active
members of that comittee. Comittees are free to form subcomittees according to their needs.
35

3. An \active" member is a person whom is, or has been, actively working on some aspect of E{ALFA within
the last six months.
4. The steering commitee, with NAIC, is responsible for insuring the planned consortium meetings take place,
and for organizing SOC and LOC for those meetings.
5. Nominations for replacement members of the steering comittee & committe leaders will be given at each
meeting and will continue for two weeks after the meeting (via a web-based submission form). Elections
will then occur via a secure web form at the end of theose two weeks. All active members will be allowed
to vote.
36