7/29/2019 1118466 Hewitt

1/122

COVER SHEET FOR PROPOSAL TO THE NATIONAL SCIENCE FOUNDATION

FOR NSF USE ONLY

NSF PROPOSAL NUMBER

DATE RECEIVED NUMBER OF COPIES DIVISION ASSIGNED FUND CODE DUNS# (Data Universal Numbering System) FILE LOCATION

FOR CONSIDERATION BY NSF ORGANIZATION UNIT(S) (Indicate the most specific unit known, i.e. program, division, etc.)

PROGRAM ANNOUNCEMENT/SOLICITATION NO./CLOSING DATE/if not in response to a program announcement/solicitation enter NSF 10-1

EMPLOYER IDENTIFICATION NUMBER (EIN) ORTAXPAYER IDENTIFICATION NUMBER (TIN)

SHOW PREVIOUS AWARD NO. IF THIS ISA RENEWALAN ACCOMPLISHMENT-BASED RENEWAL

IS THIS PROPOSAL BEING SUBMITTED TO ANOTHER FEDERALAGENCY? YES NO IF YES, LIST ACRONYM(S)

NAME OF ORGANIZATION TO WHICH AWARD SHOULD BE MADE ADDRESS OF AWARDEE ORGANIZATION, INCLUDING 9 DIGIT ZIP CODE

AWARDEE ORGANIZATION CODE (IF KNOWN)

IS AWARDEE ORGANIZATION (Check All That Apply) SMALL BUSINESS MINORITY BUSINESS IF THIS IS A PRELIMINARY PROPOSAL(See GPG II.C For Definitions) FOR-PROFIT ORGANIZATION WOMAN-OWNED BUSINESS THEN CHECK HERE

NAME OF PERFORMING ORGANIZATION, IF DIFFERENT FROM ABOVE ADDRESS OF PERFORMING ORGANIZATION, IF DIFFERENT, INCLUDING 9 DIGIT ZIP CODE

PERFORMING ORGANIZATION CODE (IF KNOWN)

TITLE OF PROPOSED PROJECT

REQUESTED AMOUNT

$

PROPOSED DURATION (1-60 MONTHS)

months

REQUESTED STARTING DATE SHOW RELATED PRELIMINARY PROPOSAL NO.IF APPLICABLE

CHECK APPROPRIATE BOX(ES) IF THIS PROPOSAL INCLUDES ANY OF THE ITEMS LISTED BELOWBEGINNING INVESTIGATOR (GPG I.G.2)

DISCLOSURE OF LOBBYING ACTIVITIES (GPG II.C.1.e)

PROPRIETARY & PRIVILEGED INFORMATION (GPG I.D, II.C.1.d)

HISTORIC PLACES (GPG II.C.2.j)EAGER* (GPG II.D.2) RAPID** (GPG II.D.1)

VERTEBRATE ANIMALS (GPG II.D.6) IACUC App. Date

PHS Animal Welfare Assurance Number

HUMAN SUBJECTS (GPG II.D.7) Human Subjects Assurance Number

Exemption Subsection or IRB App. Date

INTERNATIONAL COOPERATIVE ACTIVITIES: COUNTRY/COUNTRIES INVOLVED

(GPG II.C.2.j)

HIGH RESOLUTION GRAPHICS/OTHER GRAPHICS WHERE EXACT COLORREPRESENTATION IS REQUIRED FOR PROPER INTERPRETATION (GPG I.G.1)

PI/PD DEPARTMENT PI/PD POSTAL ADDRESS

PI/PD FAX NUMBER

NAMES (TYPED) High Degree Yr of Degree Telephone Number Electronic Mail Address

PI/PD NAME

CO-PI/PD

CO-PI/PD

CO-PI/PD

CO-PI/PD

Page 1 of 2

1118466AST - MID-SCALE INSTRUMENTATIONNSF 10-1

042103594

Massachusetts Institute of Technology

0021782000

Massachusetts Institute of Technology77 Massachusetts AvenueCambridge, MA. 021394307

The Murchison Widefield Array: Buildout to 512 Antenna Elements andFirst Investigation of the Epoch of Reionization

12,450,352 48 01/01/11

Center for Space Research

617-253-0861

77 Massachusetts AvenueRoom 37-607Cambridge, MA 021394307United States

Jacqueline N Hewitt PhD 1986 617-253-3071 jhewitt@mit.edu

Judd D Bowman PhD 2007 480-965-8880 judd.bowman@asu.edu

Colin J Lonsdale Ph.D 1981 781-981-5542 clonsdale@haystack.mit.edu

Miguel F Morales DPhil 2002 617-253-2354 mmorales@phys.washington.edu

James M Moran PhD 1968 617-495-7477 jmoran@cfa.harvard.edu

001425594

Electronic Signature

12/28/2010 3 03020000 AST 1257 12/28/2010 7:36pm

7/29/2019 1118466 Hewitt

2/122

CERTIFICATION PAGE

Certification for Authorized Organizational Representative or Individual Applicant:

By signing and submitting this proposal, the Authorized Organizational Representative or Individual Applicant is: (1) certifying that statements made herein are true and complete to the bestof his/her knowledge; and (2) agreeing to accept the obligation to comply with NSF award terms and conditions if an award is made as a result of this application. Further, the applicant ishereby providing certifications regarding debarment and suspension, drug-free workplace, lobbying activities (see below), responsible conduct of research, nondiscrimination, and flood hazardinsurance (when applicable) as set forth in the NSF Proposal & Award Policies & Procedures Guide, Part I: the Grant Proposal Guide (GPG) (NSF 10-1). Willful provision of false informationin this application and its supporting documents or in reports required under an ensuing award is a criminal offense (U. S. Code, Title 18, Section 1001).

Conflict of Interest Certification

In addition, if the applicant institution employs more than fifty persons, by electronically signing the NSF Proposal Cover Sheet, the Authorized Organizational Representative of the applicantinstitution is certifying that the institution has implemented a written and enforced conflict of interest policy that is consistent with the provisions of the NSF Proposal & Award Policies &Procedures Guide, Part II, Award & Administration Guide (AAG) Chapter IV.A; that to the best of his/her knowledge, all financial disclosures required by that conflict of interest policy havebeen made; and that all identified conflicts of interest will have been satisfactorily managed, reduced or eliminated prior to the institutions expenditure of any funds under the award, inaccordance with the institutions conflict of interest policy. Conflicts which cannot be satisfactorily managed, reduced or eliminated must be disclosed to NSF.

Drug Free Work Place Certification

By electronically signing the NSF Proposal Cover Sheet, the Authorized Organizational Representative or Individual Applicant is providing the DrugFree Work Place Certification contained in Exhibit II-3 of the Grant Proposal Guide.

Debarment and Suspension Certification (If answer "yes", please provide explanation.)

Is the organization or its principals presently debarred, suspended, proposed for debarment, declared ineligible, or voluntarily excludedfrom covered transactions by any Federal department or agency? Yes No

By electronically signing the NSF Proposal Cover Sheet, the Authorized Organizational Representative or Individual Applicant is providing theDebarment and Suspension Certification contained in Exhibit II-4 of the Grant Proposal Guide.

Certification Regarding LobbyingThe following certification is required for an award of a Federal contract, grant, or cooperative agreement exceeding $100,000 and for an award of a Federal loan or a commitment providingfor the United States to insure or guarantee a loan exceeding $150,000.

Certification for Contracts, Grants, Loans and Cooperative Agreements

The undersigned certifies, to the best of his or her knowledge and belief, that:(1) No federal appropriated funds have been paid or will be paid, by or on behalf of the undersigned, to any person for influencing or attempting to influence an officer or employee of anyagency, a Member of Congress, an officer or employee of Congress, or an employee of a Member of Congress in connection with the awarding of any federal contract, the making of anyFederal grant, the making of any Federal loan, the entering into of any cooperative agreement, and the extension, continuation, renewal, amendment, or modification of any Federalcontract, grant, loan, or cooperative agreement.(2) If any funds other than Federal appropriated funds have been paid or will be paid to any person for influencing or attempting to influence an officer or employee of any agency, aMember of Congress, an officer or employee of Congress, or an employee of a Member of Congress in connection with this Federal contract, grant, loan, or cooperative agreement, theundersigned shall complete and submit Standard Form-LLL, Disclosure of Lobbying Activities, in accordance with its instructions.(3) The undersigned shall require that the language of this certification be included in the award documents for all subawards at all tiers including subcontracts, subgrants, and contractsunder grants, loans, and cooperative agreements and that all subrecipients shall certify and disclose accordingly.

This certification is a material representation of fact upon which reliance was placed when this transaction was made or entered into. Submission of this certification is a prerequisite formaking or entering into this transaction imposed by section 1352, Title 31, U.S. Code. Any person who fails to file the required certification shall be subject to a c ivil penalty of not lessthan $10,000 and not more than $100,000 for each such failure.

Certification Regarding Nondiscrimination

By electronically signing the NSF Proposal Cover Sheet, the Authorized Organizational Representative is providing the Certification RegardingNondiscrimination contained in Exhibit II-6 of the Grant Proposal Guide.

Certification Regarding Flood Hazard Insurance

Two sections of the National Flood Insurance Act of 1968 (42 USC 4012a and 4106) bar Federal agencies from giving financial assistance for acquisition orconstruction purposes in any area identified by the Federal Emergency Management Agency (FEMA) as having special flood hazards unless the:(1) community in which that area is located participates in the national flood insurance program; and(2) building (and any related equipment) is covered by adequate flood insurance.

By electronically signing the NSF Proposal Cover Sheet, the Authorized Organizational Representative or Individual Applicant located in FEMA-designated special flood hazard areas iscertifying that adequate flood insurance has been or will be obtained in the following situations:(1) for NSF grants for the construction of a building or facility, regardless of the dollar amount of the grant; and(2) for other NSF Grants when more than $25,000 has been budgeted in the proposal for repair, alteration or improvement (construction) of a building or facility.

Certification Regarding Responsible Conduct of Research (RCR)

(This certification is not applicable to proposals for conferences, symposia, and workshops.)By electronically signing the NSF Proposal Cover Sheet, the Authorized Organizational Representative of the applicant institution is certifying that, in accordance with the NSF Proposal& Award Policies & Procedures Guide, Part II, Award & Administration Guide (AAG) Chapter IV.B., the institution has a plan in place to provide appropriate training and oversight in theresponsible and ethical conduct of research to undergraduates, graduate students and postdoctoral researchers who will be supported by NSF to conduct research.The undersigned shall require that the language of this certification be included in any award documents for all subawards at all tiers.

AUTHORIZED ORGANIZATIONAL REPRESENTATIVE SIGNATURE DATE

NAME

TELEPHONE NUMBER ELECTRONIC MAIL ADDRESS FAX NUMBER

fm1207rrs-07

* EAGER - EArly-concept Grants for Exploratory Research** RAPID - Grants for Rapid Response Research

Page 2 of 2

Laureen Horton Dec 28 2010 3:34PMElectronic Signature

617-253-3922 laureena@mit.edu 617-253-4734

7/29/2019 1118466 Hewitt

3/122

The Murchison Widefield Array: Buildout to 512 Antennas and

First Investigation of the Epoch of Reionization

Intellectual Merit

The study of the origin and evolution of our universe draws upon observations that range in

redshift from z=1000 to the present day, connecting the initial conditions of the Cosmic

Microwave Background to the current structure of galaxies and clusters of galaxies. However, a

major chapter of this history has yet to be examined the times from redshift z=1000 to z=6,

commonly known as the Dark Ages and the Epoch of Reionization (EoR). The EoR, in

particular, marks a major milestone when the first stars formed from primordial dark matter and

gas, and reionized the intergalactic medium. Remarkably, theoretical calculations predict that the

neutral hydrogen from the time of the EoR should be detectable with modern-day low-frequency

radio telescopes through observations of the redshifted 21cm spectral line. It is expected that a

radio telescope with approximately a square kilometer of collecting area (a Square Kilometer

Array (SKA)) will be needed to detect individual hydrogen features. Before the SKA, however,

a statisticaldetection of density, temperature, and ionization fluctuations in the hydrogen should

be possible with a much smaller precursor array, one with an area of about 1% of the SKA.

Building upon work previously supported by the NSF, the construction and operation of such an

array, the Murchison Widefield Array (MWA) sited in the radio-quiet environment of the

Murchison Radio Observatory (MRO) in Western Australia, will enable such a detection.

Phased-array dipole antenna elements (tiles), wideband digital receivers, technology for a

massive correlator, and a real-time calibration and imaging system have all been designed and

protoyped. A 32-tile prototype system is operating at MRO, demonstrating the viability of the

technology developed. The next step is to build out the array to 512 tiles, and to operate it during

two six-months observing seasons of two fields selected for EoR study. Analysis of the data will

complete an EoR experiment that has the goal of detecting and characterizing neutral hydrogen

fluctuations. The data will be made available to the MWA EoR collaboration and to the

astronomical community for further studies of the EoR and other topics, including space weather

and astronomical radio transients. Time on the array will also be made available to othercollaborations for separately-funded investigations.

This research is part of the Hydrogen Epoch of Reionization Arrays (HERA) program, a series of

radio arrays of increasing collecting area aimed ultimately at the full characterization of the EoR.

The HERA program was favorably reviewed by the Astro2010 Decadal Survey Study.

Broader Impact

The proposed research develops and implements new technology that takes advantage of recent

advances in digital signal processing, computation, and software. The technology is important for

future large telescope arrays, in radio astronomy and in other fields of remote sensing. The

signal and data processing techniques have broad application in modern data-intensive

environments, which might include topics as diverse as earth resource assessment, surveillance,financial market analysis, and social network data mining, to name some examples. Techniques

of radio frequency excision and spectrum management are necessary to carry out sensitive radio

astronomy observations in the modern world permeated by radio communications. Such

spectrum management techniques are increasingly required by society as the available spectrum

becomes crowded by many applications. A large number of graduate and postdoctoral students

will be involved in the hands-on work of building, commissioning and running a large radio

array, receiving training in these technical areas that have broad application in our modern

information-driven economies.

1

7/29/2019 1118466 Hewitt

4/122

TABLE OF CONTENTS

For font size and page formatting specifications, see GPG section II.B.2.

Total No. of Page No.*Pages (Optional)*

Cover Sheet for Proposal to the National Science Foundation

Project Summary (not to exceed 1 page)

Table of Contents

Project Description (Including Results from PriorNSF Support) (not to exceed 15 pages) (Exceed only if allowed by a

specific program announcement/solicitation or if approved in

advance by the appropriate NSF Assistant Director or designee)

References Cited

Biographical Sketches (Not to exceed 2 pages each)

Budget(Plus up to 3 pages of budget justification)

Current and Pending Support

Facilities, Equipment and Other Resources

Special Information/Other Supplementary Docs/Mentoring Plan

Appendix (List below. )(Include only if allowed by a specific program announcement/

solicitation or if approved in advance by the appropriate NSF

Assistant Director or designee)

Appendix Items:

*Proposers may select any numbering mechanism for the proposal. The entire proposal however, must be paginated.Complete both columns only if the proposal is numbered consecutively.

1

1

40

4

16

38

11

2

7

7/29/2019 1118466 Hewitt

5/122

1

1. SCIENTIFIC JUSTIFICATION

1.1 The Epoch of Reionization, the Dark Ages, and Cosmology Remarkably, the study of the originand evolution of our universe draws upon observations that range in redshift from z=1000 to the presentday, connecting the initial condition of the Cosmic Microwave Background (CMB) to the currentstructures of galaxies and clusters of galaxies. However, a major chapter of this history has yet to beexamined the times from redshift z=1000 to z=6, commonly known as the Dark Ages and the Epoch ofReionization (EoR). The EoR, in particular, marks a major milestone when the first stars and galaxiesreionized the intergalactic medium (IGM), which was dominated by neutral hydrogen followingrecombination. In order to understand the process of structure formation, the role of the first generationsof stars in enriching the IGM and subsequent generations of stars, and the impact of small halos and theirobserved underabundance today, we must understand reionization. Recording the history of reionizationis a major goal of the astrophysics community, with the James Webb Space Telescope, 10- and 30-mclass ground-based telescopes, and low-frequency radio observatories all aimed at studying differentsignatures of the reionization transition. In fact, the study of the Cosmic Dawn is one of three primaryscience objectives for the next decade identified by the Astro2010 Decadal Survey study (NationalResearch Council 2010).

Observations of the Lyman- forest in high redshift quasar spectra tell us that reionization must have beencomplete by a redshift of 6. The high ultraviolet opacity of HI makes the interpretation of higher redshiftspectra more uncertain (Bolton & Haehnelt 2007, Lidz et al. 2007, Becker, Rauch & Sargent 2007), andthus little is known about the history of reionization beyond a redshift of 6. Thompson scattering of theCMB provides a complementary measure of the integrated amount of ionized hydrogen between us andthe surface of last scattering, and it indicates a reionization redshift of zreion = 10.9 1.4 (Komatsu et al.2009) if instantaneous reionization is assumed. Deep Hubble Space Telescope observations provide thefirst measure of the galaxy luminosity function at z=7-8 and imply a radiation density sufficient forreionization produced by galaxies at those redshifts. These recent results therefore support a latereionization scenario (for a review see Robertson et al. 2010).

The theoretical expectation is for reionization to be extended in redshift and patchy over a wide range of

scales. Figure 1 is a simulation (Alvarez, private communication; see also Alvarez and Abel 2010)showing the large HII bubbles that trace structure formation during the middle of the EoR. It is believedreionization was driven by star-forming galaxies, which ionize nearby regions of the IGM. Due to theclustering of galaxies, these ionized bubbles of gas overlap to form large ionized regions, whichthemselves eventually overlap and merge until essentially all the IGM is reionized.

Figure 1: Predicted structure of reionization in a

1 Gpc h-1 simulation volume at redshift z~9, for a

model in which the universe is about half ionized

at this redshift. Ionized regions are blue andtranslucent, ionization fronts are red and white,

and neutral regions are dark and opaque.

Reionization is still quite inhomogeneous on these

large scales, with some large regions ionizing long

before others (M. Alvarez, personal

communication).

7/29/2019 1118466 Hewitt

6/122

2

The details of early structure formation are encoded in the sizes, distribution, and evolution of the ionizedbubbles, and in the structure of cold and hot (relative to the CMB) gas. The first generation of stars canenhance the formation of subsequent stars through the creation of heavier elements and shocks, or hindersubsequent star formation by heating and ionizing the gas; i.e., by promoting or preventing gas collapse.Similarly the mean free ultraviolet photon path, related to the maximum bubble size for an individualgalaxy, is determined by the density and longevity of HI in small dark matter halos. These are halos that

might one day become dwarf satellites, and thus processes during the EoR likely influence the dwarfpopulation today. Meanwhile X-rays from early black hole emission, or shock-induced heating, canintroduce fluctuations in the kinetic temperature of the hydrogen gas, and these fluctuations are coupled tothe spin temperature by the Wouthuysen-Field effect when conditions are appropriate. Finally, quasarsmay ionize large regions near the densest areas of structure formation. All of these effects are recorded inthe distribution and growth of HII regions during the EoR. For a more complete description, the reader isencouraged to refer to the recent review papers by Furlanetto, Oh, & Briggs (2006) and Morales &Wyithe (2010).

While images of high redshift galaxies and absorption spectroscopy will provide valuable information onthe EoR, only 21 cm observations of the HI have the promise of revealing directly the process ofreionization and the detailed effects of the first luminous objects on the IGM. To the best of our

knowledge, the first suggestion that the EoR could be probed by redshifted 21cm observations was madeby Sunyaev and Zeldovich in 1972. A number of quantitative predictions of observational signatureswere made in subsequent years, including those by Hogan & Rees (1979), Scott & Rees (1990), Madau,Meiksin & Rees (1997) and Tozzi et al. (2000). Figure 2 shows a simulation of 21cm radio brightnesstemperatures for nine different times during the EoR (Furlanetto, Sokasian & Hernquist 2004). Themagnitude and scale of the features in Figure 2 imply that long integrations with a low-frequency radiotelescope with collecting area of order one square kilometer are required for detection of individualfeatures; in fact such observations are one of the main motivations for the proposed Square KilometerArray (SKA; Carilli & Rawlings 2004). However,statisticalinvestigations, such as measuring the powerspectrum of 21cm fluctuations averaged over a large area of the sky, could in principle be carried out witha much smaller radio telescope (Morales & Hewitt 2004). This statistical detection of 21cm fluctutationsduring the EoR is the primary goal of the Murchison Widefield Array (MWA), the subject of this

proposal.

Figure 2: The radio brightness temperature, the

target of our proposed observations, simulated at

nine different redshifts by Furlanetto, Sokasian

and Hernquist (2004). Each panel corresponds to

the same slice of the universe seen at a different

time, and has a depth that corresponds to a

bandwidth of 0.1 MHz. The redshifts range from

12.1 to 7.6.

7/29/2019 1118466 Hewitt

7/122

3

While the MWA is focused on detecting and measuring the EoR power spectrum, there is also significantinterest in the 21 cm cosmology observations that would characterize the dark ages before reionization,improve the measurement of cosmological parameters (Bowman, Morales & Hewitt 2007; Mao et al.2008), and constrain the properties of dark energy and the neutrino mass (Loeb & Wyithe 2008, Wyithe& Loeb 2009, Seo et al. 2010). Because the 21 cm signal reveals the fluctuations within a volume and notjust on the surface of last scattering, it provides a way to measure many more modes of spatial

fluctuations than does the CMB (Loeb & Zaldarriaga 2004). By extending 21cm measurements to highredshift with more collecting area than the proposed MWA, or even with new instruments, the study ofthese modes can include the otherwise inaccessible dark ages, providing direct measurements ofgravitational collapse and the temperature and coupling of the primordial gas. It is also possible to extendthe EoR power spectrum techniques to the post-reionization universe at lower redshift by tracing theneutral hydrogen in galaxies. This provides a powerful tracer of dark energy through the baryon acousticoscillations (BAO), and may enable three-dimensional BAO measurements (including Alcock-Paczynskieffects) over redshifts of 14 with relatively modest instruments. Depending on how our understanding ofdark energy evolves, this might provide a cost-effective approach for measuring the size of the universeover large portions of cosmic time. Thus, while these measurements are not in themselves possible withthe MWA as proposed here, the MWA is a crucial pathfinder for future 21 cm cosmology instruments.The MWAs anticipated advances in foreground subtraction, large-N correlators, and precision calibration

will provide a foundation for the future instruments. Recognizing the important role the MWA will playin developing low-frequency capabilities, the International SKA organization has named the MWA aPrecursor for the SKA; i.e., a pathfinder located at a candidate SKA site.

1.2 Other Science Detecting the redshifted 21cm signal from the EoR is the most technicallydemanding science goal for the MWA, and has driven most of the technical specifications for the array.This yields an array design suitable first for EoR studies, and also for other science. Therefore, theversatility of the array creates opportunities to seek funding for operating the array from diverse sources.The investment to date into the MWA, by NSF and other US agencies, has been made with heliosphericscience and transient science as explicit goals. Non-EoR science topics are also identified in theStatement of Collaboration between the international partners. The buildout to the full collecting area ofthe array must therefore be done in a manner consistent with these prior US financial investments andinternational partner commitments, including some features that are not strictly necessary for the EoRgoal addressed by this proposal. This applies to outlier tiles, and tied array beamforming capability, aswell as other less significant features. While we do not seek funding in this proposal to carry out otherscience, for completeness we summarize its status and prospects.

Scientific collaborations have been formed to exploit non-EoR MWA science in three broad areas:

1. Solar, heliospheric and ionospheric (SHI) science. The unifying theme for this science collaboration isspace weather, its causes and effects. Geomagnetic storms caused by coronal mass ejections (CMEs)have impacts of economic significance, affecting both space-based and ground-based assets. A betterunderstanding of such events is needed, and MWA measurements of various kinds can make important,and in some cases unique, contributions. Of particular note is the measurement of heliospheric Faradayrotation signatures of CMEs, which are diagnostic of magnetic field orientation within the CME plasma.This orientation, in turn, determines the degree of coupling of CME energy into the Earths

magnetosphere, and the severity of the ensuing geomagnetic storm event. The MWA will greatlyenhance our capabilities for remote sensing of the heliosphere, including both Faraday rotation andinterplanetary scintillation measurements, and offers the potential for significantly improved spaceweather prediction. Support is being sought from the Atmospheric and Geospace Sciences Division ofthe NSF for these studies.

2. Transient radio sources. The exploration of transient and time-variable phenomena is one of the lastfrontiers in astronomy, especially over the various domains of the radio frequency band. Radio transientsvary on time scales of seconds to years. The potential sources of radio transient emission fall into five

7/29/2019 1118466 Hewitt

8/122

4

broad categories: coronal emission from nearby stars, emission from compact objects such as neutronstars and accreting black holes, explosive events such as gamma-ray bursts and radio supernovae,emission associated with planetary magnetospheres, and new phenomena discovered through thedevelopment of a new observational capability. The first two are known sources of transients, andconstitute guaranteed science that the MWA should probe. For the last three categories there aretheoretical expectations that transient emission should exist, but its properties are not well constrained by

current observations; these constitute serendipitous science. Through its combination of largecollecting area, large fields of view, and long observing times, the case the MWA will have a sensitivityto transient events dramatically better than any previous instrument in this frequency range, and highlycompetitive with (and complementary to) other new or planned instruments. The MWA observingprogram for radio transients includes several backend instruments that will process the data in differentways, making the MWA sensitive to different timescales of variability. These programs are being carriedout by the Transient Collaboration of the MWA, with data analysis support secured from a variety ofsources by different members of the collaboration.

3. Galactic and extragalactic (GEG) science. This category encompasses all MWA astronomical sciencethat cannot be classified as EoR or transients. It can be broadly divided into galactic science topics andextragalactic science topics, hence the name (GEG). In addition to the EoR observations supported by thisproposal (useful for GEG science as well), daytime observing time and nighttime observing time whenthe Galaxy is overhead will enable GEG-specific observations across the sky and over a range offrequencies. We list several GEG science topics for which the MWA is particularly well suited:

Study of the interstellar medium (ISM) via Faraday tomography. The ISM is visible insilhouette against the bright polarized galactic synchrotron emission, through the Faradayrotation effects of the turbulent and inhomogeneous magnetoionic component of the medium.

Constraining the distribution of relativistic electrons in the ISM by using low-frequency free-freeabsorption in HII regions at known distances to block galactic synchrotron emission behind theHII region, thereby measuring the synchrotron emissivity in frontof the HII region

Taking a census of galactic supernova remnants. This information is needed to understand theproduction and energy density of cosmic rays, for the study of turbulence and triggered starformation, and for an understanding of ISM energetics.

A survey for old, low-luminosity and nearby pulsars, providing a robust estimate of the birth rate,luminosity function and total size of the overall Galactic pulsar population.

Imaging the cosmic web at low frequencies. The intergalactic medium is laced with relativisticparticles and magnetic fields from shocks and AGN activity. The MWA has the observationalcharacteristics to make the first maps of the resulting low brightness synchrotron structures,tracing the particle acceleration, and diffusion away from radio galaxies and clusters.

Broadband spectral decomposition of extended extragalactic radio sources. The detailed spectralshape of various structures in the extended lobes of such sources are governed by multipleabsorption processes, particle acceleration, and synchrotron aging. The MWA will provide high-precision continuous spatially resolved spectra over the full 80-300 MHz range, leading to abetter understanding of the radio loud AGN phenomenon, and interactions with the IGM.

We anticipate that financial support for operations and analysis associated with these studies will besecured by other groups through other proposals.

2. THE EoR: EXPERIMENTAL APPROACH

2.1 Power Spectrum Measurements The redshifted 21 cm emission is faint, and it is masked byforeground radio emission that is about five orders of magnitude brighter than the EoR-induced brightness

7/29/2019 1118466 Hewitt

9/122

5

temperature fluctuations. Imaging individual features will require the sensitivity of a telescope with largecollecting area, such as the proposed SKA. The cost of a low-frequency SKA is estimated to be at leastseveral hundred million dollars (Backer et al. 2010). As was first shown by Morales & Hewitt (2004), astatisticaldetection of EOR 21cm fluctuations can in principle be carried out with a telescope with amuch smaller collecting area. For this reason, several first-generation EoR instruments have beenproposed: the MWA (Lonsdale et al. 2009), the Precision Array for Probing the Epoch of Reionization

(PAPER; Parsons et al. 2009), the core of the Low Frequency Array (LOFAR; Butcher 2004), and thecore of the Giant Metrewave Radio Telescope (GMRT; Pen et al. 2008). None of the first-generationEoR instruments will have the sensitivity to image individual ionized regions during the EoR (with thepossible exception of targeted long integrations on the very largest ionized regions around known quasars;Wyithe 2008). Instead, all are focusing first on observing the three-dimensional EoR power spectrum todetect the faint EoR signal and to distinguish it from the galactic and extragalactic foregrounds.

Because the 21cm spectral line is optically thin, radio measurements probe a three-dimensional volume.The three-dimensional EoR power spectrum is analogous to the more familiar two-dimensional CMBpower spectrum. While the MWA will be capable of a wide variety of scientific studies, the design of theMWA has been optimized for the EoR power spectrum measurement. The sensitivity to the powerspectrum is maximized by high survey speed, short baselines, and high dwell time (Morales 2005). Byfully cross-correlating 512 small antennas, the MWA obtains a ~900 deg2 field of view at 150 MHz. Theantennas are dominated by sky noise, so their being electronically steerable allows the MWA to track coldregions of the sky, minimizing the Galactic noise contribution and achieving long dwell times to enhancepower spectrum sensitivity (sensitivity is degraded in proportion to the square root of the number offields, given a fixed total integration time; Halverson 2002). Short baselines have the highest sensitivityto the expected signal, leading to choice of a highly centrally condensed antenna distribution.

Because of the design choices, the MWA has the best power spectrum sensitivity of the first generationEoR observatories. However, raw sensitivity is only part of the challenge. Subtracting the brightforegrounds requires extraordinary measurement precision, and the MWA has been further optimized forexcellent instrumental calibration. Some of these design innovations include: nearly complete snapshotcoverage of the Fourier u-v plane (yielding an excellent point spread function enabled by a correlatorprocessing an unprecedented number of baselines), the capability for continuous holographic antenna

measurements for calibration, a horizon-to-horizon correlator field of view (for imaging confusingsources in far sidelobes), and a site in Western Australia with very low levels of radio frequencyinterference.

The three-dimensional image volume of the 21 cm observations can be described by the coordinates ofangle and frequency (x , y, f) that map into the cosmological volume (x, y, z) through the angulardiameter distance and redshift relationships. Radio interferometers directly measure angular Fouriermodes {ux , uy} in a large number of narrow frequency channels, forming a (ux, uy, f) measurementvolume. Through a simple scaling, these measurements can be interpreted as cosmological wavenumbersin the angular direction (with units of inverse distance) and line-of-sight distance in the frequencydirection (kx, ky, z). Then the interferometric measurements can either be Fourier transformed in theangular directions to form the image volume, or they can be Fourier transformed in the line-of-sight(frequency) direction to form a three-dimensional wavenumber cube (kx, ky, kz) (Morales & Hewitt 2004).

In the absence of cosmic evolution and velocity distortions, the power spectrum signal would bespherically symmetric in the three-dimensional k-space due to the rotational invariance of space.Measurements of the one-dimensional power spectrum can be performed by measuring the variance of allthe measurements within a spherical shell in k-space, taking velocity effects taken into account.Astrophysical foregrounds can be identified, and separated from the EoR signal of interest, by theirspecific and generally nonspherical shapes in three-dimensional k-space. This is the same rotationalsymmetry that leads to averaging over angular m modes in CMB analyses, translated from the two-dimensional surface of the CMB to the three-dimensional volume of HI cosmology measurements.

7/29/2019 1118466 Hewitt

10/122

6

Figure 3 shows the sensitivity of the MWA to the reionization power spectrum at different redshifts,assuming the foregrounds can be perfectly removed.

Figure 3: The sensitivity of the MWA to the EoR power spectrum for redshifts 6, 8, 10, and 12

after 360 hours of integration. The solid line is a fiducial model of the 21cm fluctuations that

represent the fluctuations due just to gravitationally-enhanced primordial density fluctuations,

without including effects of reionization. Reionization bubbles can make the fluctuations either

smaller or larger, typically by a factor of a few. The dark shaded region shows the errors due to

sample variance; the light shaded region shows the combined errors of sample variance and

thermal (sky-dominated) noise. These calculations assume perfect subtraction of foregrounds and

show just the impact of thermal noise and MWA design choices (eg, the number, size and layout of

the antennas). Adapted from Bowman, Morales & Hewitt (2006).

To illustrate the potential of the MWA, Figure 4 shows the sensitivity of the MWA at redshift z=8compared to different models of reionization. Each model power spectrum plotted in Figure 4 is for adifferent neutral fraction; the wide range of neutral fractions depicted at this redshift is consistent with thecurrent weak constraints on reionization. The most critical determinant of the magnitude of the powerspectrum at any redshift is the ionization fraction at that redshift, so these plots capture the essentialbehavior that one might expect at a single redshift. In addition to having the sensitivity to detect thepower spectrum, the MWA also has the capability to record its dynamics as a function of redshift as theionized regions expand and overlap. Lidz et al. (2008) showed that it is the dynamics of the powerspectrum that provide many of the strongest constraints on reionization. Thus the MWA 512T will have

the sensitivity to detect plausible EoR scenarios, if a significant neutral fraction remains at theredshifts accessible to the MWA, roughly z=6 through z=11. While a positive detection of EoRfluctuations would be the most exciting, even a meaningful null result would reveal much about thehistory of reionization, forcing us to accept very early reionization models.

7/29/2019 1118466 Hewitt

11/122

7

Figure 4: Sensitivity of the MWA at z=8 compared to

different reionization models. This is the same plot as the

second panel of Figure 3, but dashed lines have been added

that show different reionization models. From top to

bottom, the ionization fraction for each model is 0.51, 0.0

(solid line), 0.43, 0.38, 0.25, and 0.13. In general, fluctuations

are largest when the ionization fraction is near one half. Themodels are those of Furlanetto et al. (2004a, 2004b) and the

sensitivity calculations are those of Bowman, Morales and

Hewitt (2006) and, to illustrate theoretical sensitivity, assume

perfect foreground subtraction.

The need to separate the bright foregrounds from the faint cosmological signal is what makes the EoRpower spectrum measurement difficult. This separation is possible in principle because most of theastrophysical foregrounds are spectrally very smooth (being due primarily to synchrotron and free-freeemission), whereas the HI signal fluctuates both spatially and spectrally (i.e., as a function of redshift).This separation is readily apparent if one separates the k-vectors into their components perpendicular to

the line-of-sight (k) and parallel to the line-of-sight (k). Nearly all of the foreground emission iscontained within the first few k-modes in the line-of-sight direction as shown in Figure 5 (left panel). Forexample, at the MWAs observing frequency the sky is confused with a sea of extragalactic radiogalaxies. Every pixel contains many galaxies, with the brightness, spectral slope, and spectral curvature

varying from pixel to pixel due to Poisson noise and large scale clustering. In the 3D k-space of Figure 5,this foreground is very bright and nearly constant along k (white Poisson noise slightly reddened by

large scale clustering), but it falls very rapidly in the k direction due to the spectral smoothness. DiMatteo et al. (2002) and Oh & Mack (2003) showed this extragalactic emission precludes a power

spectrum measurement in angular directions only (k = 0 in Figure 5). However the spectral smoothness

allows us to measure the EoR power spectrum along the line-of-sight away from the k axis. All of theidentified foregrounds share this spectral smoothness or are at fixed observing frequencies (such as RFIand galactic radio recombination lines), and it is now believed that foregrounds do not form aninsurmountable impediment to observing the EoR power spectrum (see, for example, di Matteo, Ciardi &Miniati 2004; Gnedin & Shaver 2004; Morales & Hewitt 2004; Zaldarriaga, Furlanetto & Hernquist 2004;Santos, Cooray & Knox 2005; McQuinn et al. 2007; Liu, Tegmark & Zaldarriaga 2009).

7/29/2019 1118466 Hewitt

12/122

8

Figure 5: These three panels depict the wavenumbers of the 21 cm signal signal in the sky (k) and

line-of-sight (k ) directions. The left-hand panel (adapted from Morales & Wyithe 2010) is acartoon showing the spectrally smooth foregrounds being confined to very low k , but dominatingin the sky plane due to Poisson noise (foreground radio galaxies) and other angular fluctuations.

The 1-D power spectrum signal is created by averaging the power within the spherical rings. The

central panel then shows the expected level of the EoR signal with its near spherical symmetry

(Datta et al. 2010, note the spherical rings appear square due to log-log axes). The right-hand panel

is also from Datta et al. 2010, and shows the foreground contamination due to small 0.1" rms errors

in the location of subtracted bright sources. While the fundamental contamination from this

foreground is entirely in the sky plane (k), the chromatic sidelobes of the instrument couple with

the subtraction errors to throw power in the line-of-sight direction (k ). The wedge of power seen inthe right-hand panel is fairly generic, and is related to the Fourier transform of the instrumental

point spread function. The center and right-hand panels are on the same color scale (though the

axes are slightly different). It is clear the EoR signal is stronger than this foreground over much of

the space, and dominates along the line-of-sight at low k. The key to foreground subtraction is to

keep these mode-mixing effects low enough to preserve a region of wavenumber space where the

signal is stronger than the residual foregrounds.

In practice the separation of EoR signal and foregrounds is complicated by mode mixing, whereinstrumental effects and calibration errors mix foreground emission into the spectral domain. Forexample, the point spread function (PSF; synthesized array beam) of a low frequency interferometer isinherently chromatic, and the position on the sky of sidelobes shifts as a function of frequency. If a sourceis mis-subtracted (either in amplitude or location), a residual source will remain along with its PSF. Alonglines of sight away from the residual source the sidelobes will sweep through as a function of frequency,creating artificial spectral ripples. The third panel of Figure 5 shows the mis-subtraction of spectrally

smooth bright sources throwing power that was inherently along the k axis into the line-of-sight

direction. In this example the residual foreground contamination is less than the EoR signal within thewedge of k-space to the upper left. Common mode-mixing effects include Faraday rotated galacticemission and polarization leakage, calibration errors and bright astrophysical sources, chromatic pointspread functions and fields of view with confusion level sources, and ionospheric refraction andscintillation with bright sources (see Morales & Wyithe 2010 and references therein). Each of these caseshas its own characteristic k-space contamination template (Morales, Bowman, Hewitt 2006), and must beconsidered and subtracted separately.

7/29/2019 1118466 Hewitt

13/122

9

This problem of mode-mixing is the reason for the MWAs emphasis on precision calibration. In general,the magnitude of the product of the calibration error level and the mixing foreground needs to be belowthat of the expected EoR signal. This places stringent requirements on instrumental calibration andforeground subtraction. The MWA calibration and the foreground subtraction pipeline described beloware designed with a view to providing the dynamic range needed to separate the EoR signal from thebright foregrounds.

To summarize, the scientific goals of this proposal are centered on the EoR science program describedabove. Specifically the goals are to (1) characterize the astronomical foregrounds, (2) separate thedifferent foreground components from the residual EoR signal, and (3) detect the power spectrum of theEoR or place meaningful limits on its magnitude at the redshifts accessible to the MWA. If these threegoals are achieved, then we also add a fourth: (4) record the evolution of the power spectrum of the EoRover the redshift range accessible to the MWA. Formally, this redshift range is z=5 to 16, and there is noexpectation for detectable neutral hydrogen below z=6 or above z=11 (see Figure 3). The absence ofsignal from z=5 to 6 is a prediction that is testable with MWA data. Above z=11 MWA detectability islikely to be limited by sensitivity.

2.2 Other EoR Measurements While the scope of this proposal is limited to foreground subtraction andpower spectrum measurement, other probes of the EoR have been proposed. We anticipate that the

archived MWA EoR data will be used, first by the MWA EoR collaboration and later by the community,to use these probes to extract further information about the reionization process. Such investigationsmight include:

Non-Gaussian statistics - there is interest in going beyond studies of the power spectrum andinvestigating the full statistical properties of the 21-cm brightness distribution. The fluctuations due toreionization are inherently non-Gaussian, and the degree to which the non-Gaussianity is manifestedincreases as reionization proceeds (Furlanetto, Zaldarriaga & Herniquist 2004b; Wyithe & Morales 2007;Harker et al 2009; Ichikawa et al. 2010).

Cross-correlation with optical observations of galaxies a probe of the connection between reionizationand the sources of ionizing photons (Wyithe & Loeb 2007, Furlanetto & Lidz 2007, Lidz et al. 2009). Forexample, if over-dense regions are ionized earlier, 21cm emission and galaxies should be anti-correlated.

Observations of ionized regions around quasars first detections of individual ionized regions are likelyto be around known quasars, and the very largest ones may be detectable with first-generation 21cmarrays (Valdes et al 2006, Geil et al. 2008, Wyithe 2008).

The planned MWA observations will yield an extraordinarily rich dataset. Open access to these data, firstby the MWA EoR Collaboration and later by the community at large, is likely to lead to innovativeinvestigations of the EoR (and other science topics) that cannot now be predicted.

2.3 The HERA Concept As noted in section 1.1, fully exploiting the scientific potential of 21cmtomography will require a telescope with very large collecting area and dynamic range. Investigation ofthe EoR and the Dark Ages is in fact one of the primary scientific goals of the international SquareKilometer Array (SKA) project, and it is anticipated that a low-frequency component, SKA-low, willultimately be needed to fully explore this area of science. The optimal design for such an SKA-low has

not yet been determined; developing the design requires near-term investment in scientific andtechnological demonstrators that will guide design choices.

As part of the discussion informing the Decadal Survey process in the United States, a collaboration hasdeveloped a concept and roadmap for a sequence of Hydrogen Epoch of Reionization Arrays (HERA;Backer et al. 2010). HERA was scientifically the most highly ranked major new initiative of theRadio, Millimeter, and Submillimeter (RMS) Paneland was identified as the U.S. development path forparticipation in the international low-frequency Square Kilometer Array. The Panel recommendedsupport for the two pathfinders currently under way, MWA and PAPER, and further recommended a

7/29/2019 1118466 Hewitt

14/122

10

review mid-decade for a decision on whether or not to proceed to the next stage. The #2 large-scaleground-based initiative of Astro2010 is a mid-scale program at the NSF, and the committee identified

a second phase of HERA as among the most promising such initiatives for the next decade.

The HERA program consists of three phases: (1) HERA I, with the goal of detecting the power spectrumof the 21cm line emission structures at several redshifts, a first probe of galaxy formation; (2) HERA II,

with the goal of characterizing these structures and pioneering the extraction of detailed information aboutgalaxy formation astrophysics and cosmological physics; and (3) HERA III, with the goal of detailedimaging of individual structures across a large fractional step in cosmic time, fully exploring firststructure formation and the accessible fundamental physics. The scientific goals and the scale of HERAIII are aligned with those of the international SKA-low project, and a merging of the projects may occurin the future as appropriate. The HERA I arrays are those currently under development, the MWA andPAPER. The MWA and PAPER collaborations have formed a HERA Coordination Group withrepresentatives from both experiments who communicate via regular telecons. A HERA webpage hasbeen established (http://reionization.org), and will serve as a communications medium for the two groupsand as a repository for scientific and technical results. We intend to nurture and expand this collaborationas the two projects mature.

The HERA concept is based on arrays that feature wide-field capability, large-N architecture, high

precision polarimetric calibration, large fractional bandwidths, and high sensitivity. While it is prematureto specify the detailed design of the future arrays, the basic scenario is summarized in Table 1, reproducedfrom the Decadal Survey white paper.

Instrument Phase Aeff

(km2)

No. of Elements Field of View(degrees)

Scientific Goal1

MWA 2010-2014 0.01 512 20-40 PS detection

PAPER 2010-2014 0.003 128-512 60 PS detection

HERA II 2015-2019 0.1 50002 20-602 PS

characterizationHERA III 2020- O(1) O(50,000)2 (20-60)

2 Structureimaging

Notes: 1PS = Power Spectrum 2Design parameters to be determined based on HERA I and HERA II results.

Table 1: The proposed arrays comprising the HERA roadmap, as described by Backer et al. (2010).

A fundamental premise of the HERA concept is that a period of scientific and technological developmentstudy must be carried out before the design of HERA II is finalized. These studies (see section 2.4) are tobe carried out with the current pathfinder arrays and related technology development programs. The goalof the work proposed here is to realize the MWA component of the HERA I effort.

2.4 The MWA as a HERA Testbed The specific questions to be addressed by the MWA, PAPER, andother experiments during the HERA-I phase are:

a. Antenna calibration: How to achieve the necessary dynamic range is a central question in the designof HERA-II. A key to high dynamic-range is precision wide-field, frequency-dependent calibration of thecomplex gain and polarization responses of the antennas that comprise the HERA instrument. Possiblecalibration strategies range from using static, maximally predictable hardware that can be wellcharacterized through simulations or laboratory measurements, to relying on an overconstrained system of

7/29/2019 1118466 Hewitt

15/122

11

many redundant measurements of antenna properties using astronomical calibrator sources across theentire sky during observations.

Two strategies are being explored by PAPER and the MWA to achieving the calibration precision

required. The PAPER antennas are passive, static systems designed to have very predictable and stableantenna patterns as a function of both time and frequency, while MWA's higher-gain, tracking phasedarray antenna systems are inherently less uniform and change while tracking sky rotation, but provide

more collecting area at a fixed cost. If the calibration challenges presented by the MWA design can bemet, there will be significant implications for the design and costing of large-area arrays, such as HERA-

II and the SKA. The MWA team will develop and quantify software calibration techniques that exploitmassively redundant measurements of each antenna's wide-field, frequency-dependent gain responseevery 8 seconds using astronomical sources.

b. Drift scans vs. field tracking: For 21 cm power spectrum measurements, field-tracking observationsare desirable because they can provide a factor of 2-3 increase in sensitivity and improved synthesisrotation coverage (for enhanced spatial dynamic range) since the tracked target fields can be placed incold Galactic windows. This is to be balanced against the well-known advantages of drift scan observingstrategies for overcoming a variety of systematic errors inherent in high dynamic-range radioobservations. Assessing the relative merits and drawbacks of each approach in an experimental setting is

an important, unresolved consideration for minimizing the cost and complexity of HERA-II.The MWA will provide a platform to compare directly both observing strategies. The Baseline EoRExperiment for the MWA outlined in this proposal will acquire very deep integrations on two EoR fieldsby tracking the fields using the MWAs electronically steerable antenna tiles. Complementary

observations with the MWA, led and funded by collaborators at RRI, will acquire zenith-locked driftscans. Therefore, both approaches will be investigated with the same instrument. Furthermore, PAPERwill employ a drift scan strategy and will provide another point of comparison.

c. Array layout: The layout of array antennas has strong implications for EoR sensitivity, instrumentalcalibration, and foreground subtraction. The power spectrum and imaging sensitivity are enhanced withmany short baselines, whereas antenna calibration and foreground subtraction are enhanced with longerbaselines and a well-filled UV-plane. Finding the optimum balance between compact and sparse arraydistributions requires actual end-to-end tests of calibration and foreground subtraction performance.

Another issue is the degree to which baselines should be redundant to overconstrain calibration solutions.

The MWAs large number of elements arranged in a compact core surrounded by a power-law fall-off ofelements will exploit the advantages of a well-filled uv-plane and explore the possible advantages ofbaseline redundancy for calibration, while providing relatively long baselines to facilitate calibration andforeground subtraction. The chosen configuration thus represents an effective testbed for assessing theconfiguration requirements and projected performance of a HERA-II array. The MWA has some

flexibility in its configuration in that the receiver-tile cables are not trenched and might be moved; thiswill be particularly useful for exploring baseline redundancy.

d. Ionosphere: The degree to which ionospheric fluctuations affect EoR power spectrum measurementsremains unclear. Under most conditions, the shifts in angular positions from the rapid ionosphericfluctuations are on smaller scales than those expected to be relevant for HERA-II power spectrum

measurements. However, their effect on foreground subtraction, particularly for the bright point sources,may be significant. Resolving these issues requires an understanding of the spatial and temporalproperties of ionospheric fluctuations, and the degree to which different analysis techniques are sensitiveto the fluctuations.

EoR observations with the MWA will extend over six months each observing season, sampling the highlyvariable ionospheric conditions that we expect. The baseline EoR analysis and other planned analyseswill have to contend with these ionospheric effects. Attention will be paid to characterizing the types ofvariation we see and their severity, and investigating their impact on calibration and foreground

subtraction is an important part of the study. We will also learn how residual ionospheric errors affect

7/29/2019 1118466 Hewitt

16/122

12

the extraction of cosmological information from the data. An interesting issue to address as well iswhether ultimately space-based or lunar farside experiments will be of interest.

e. Foregrounds: The anticipated spectral smoothness of low-frequency diffuse Galactic emission andconfused continuum sources needs to be confirmed at the level of 1 part in 104 in order to ensure thatplanned foreground mitigation strategies will be capable of revealing the redshifted 21 cm powerspectrum. The development of optimum matched filters for the separation of diffuse foregrounds, pointsources, and the EoR signal requires measurements of the spatial and spectral correlations of the differentcomponents.

As described in other sections, characterizing the foregrounds and developing subtraction techniques are

a major focus of the MWA EoR investigation.

f. Automatic data evaluation and editing: The data rates that will be produced by HERA instrumentsare enormous. Some of the data will be corrupted by radio frequency interference, extreme ionosphereconditions, and equipment failures, and will have to be removed. Such editing is only practical withautomatic data evaluation and editing techniques. These cannot be developed effectively until we havegained an understanding of the ways in which bad data manifest themselves in practice.

The MWA data collection and analysis plans include both on-line and off-line data assessment, andpreliminary work has already been done in the execution of the 32-tile engineering and science program.

On-line data assessment will be used to veto data before it is included in averages. Off-line dataassessment tools will be made available to individual investigators to apply to the data according to their

needs and judgment.

g. Trade-offs between data storage and real-time processing: For the purposes of minimizing effectsof systematic error and cross-checking, archiving minimally processed data is desirable. This has to bebalanced against the cost of data storage. A rational evaluation of this trade requires the data processingtechniques to be established and understood.

For the period of time addressed by this proposal, MWA data will be transferred via fiber to the $A80MPawsey HPC Centre for SKA Science in Perth. These data will consist of samples of raw visibilities forengineering evaluation and software testing, and data images from the RTS system for science. Funding

has already been secured for 500-Tbyte archives for EoR data at MIT and the University of Melbourne,

and with the ever-decreasing cost of storage, larger archives are already becoming feasible. Dealingwith the MWA data will give us practical experience that will be critical for guiding future experiments.

h. Environment and RFI mitigation: Low-frequency radio measurements are inevitably plagued byRFI. The proposed HERA sites in Western Australia and South Africa have been chosen because theyare relatively radio-quiet locations, and placement of radio transmitters and sources of interference at thesites will be strictly controlled through the SKA site preparation process. However, instrusion fromnearby external transmitters and interferers, airborne sources, and atmospheric propagation of far sourceswill inevitably occur. While RFI studies are useful, the true suitability of the two sites will not be knownuntil intensive science observations are undertaken with very deep multi-hundred hour integrations.

This is an opportunity to demonstrate that interference at the MWA can be kept at an acceptable levelthrough technical and regulatory controls; work in this area is funded by the CAASTRO program for

astronomy investigation and is, of course, a major component of Australian pre-SKA activities fromwhich the MWA benefits. Similarly, PAPER will test the viability of the South African SKA site. Inaddition to the science data, the design of the MWA EoR database includes extensive information about

the environment collected from sensors and the calibration system (these metadata are comparable involume to the science data themselves). Future experiment design efforts will be able to draw from all

these resources for trade studies and design optimization.

i. MWA as a HERA-II testbed: Whether or not the MWA reaches its theoretical sensitivity to the EoRsignal, and whatever degree of power spectrum characterization is achievable with the 512-tile array, wewill have at our disposal a large, sensitive, well calibrated and thoroughly understood low-frequency array

7/29/2019 1118466 Hewitt

17/122

13

supported by extensive infrastructure in a radio quiet location. The MWA 512-tile array (512T) designrepresents a major departure from that of conventional imaging radio astronomy interferometers, andsimultaneously incorporates multiple innovative concepts and technologies, in hardware, software, andsystem design. As such, the MWA will constitute a critically important and highly capable testbed forcandidate HERA-II devices, subsystems, observing strategies and analysis approaches. A wide variety ofcandidate antenna designs can be hooked directly into the data path and comprehensively evaluated with

high precision, using a fully debugged and understood antenna pattern calibration system. Digitalbeamforming can be explored in detail compared to the precisely characterized analog systems of theMWA. New correlation concepts can be tested, and results compared with reference MWA dataproducts. Calibration algorithms designed for HERA-II can be assessed for robustness and accuracyusing large volumes of real-world data.

3. DESIGN OF THE MURCHISON WIDEFIELD ARRAY

Our scientific and technical goals have driven the design of the MWA. As such, it is a focusedexperiment that will enable pathfinding measurements of the EoR, and other areas of science, at relativelylow cost.

3.1 Conceptual Basis for the 512T Design. The 512-tile (512T) MWA design concept is guided bytwo central considerations. First, there is a compelling scientific need for a very wide instantaneous fieldof view. This is the case of all four science areas, and is especially true for the goal of detecting the EoRpower spectrum, and for measurement of heliospheric Faraday rotation. To avoid imposing scale sizes ondetected structures due to instrumental imperfections, approaches involving the stitching together of manysmaller fields of view have been rejected, in favor of maintaining a seamless, wide field of viewthroughout the data path, from antenna to final data product.

Second, it is recognized that instrumental and ionospheric calibration at low frequencies and for a widefield of view requires the accurate determination of a large number of direction-dependent parameters. Inorder to solve quickly, robustly and uniquely for these parameters, we require a large, orthogonal set ofhigh-SNR independent measurements. The elemental independent measurement provided by the system

is the complex visibility obtained on a baseline between a single tile and the rest of the array, in thedirection of a single calibrator source on the sky. As the number of tiles increases, the sensitivityincreases which rapidly increases the number of suitable calibrator sources on the sky, including locationsof weak tile response which strongly constrain tile models. Simultaneously, array sidelobe confusiondrops, improving the accuracy of calibrator measurements. The combination of these effects leads tostrong and nonlinear dependence of calibration efficiency on the number of tiles. In fact, simulationsindicate that at least 360 tiles are required for acceptable calibration solutions (Greenhill et al. 2009).

Both of these considerations lead to a design featuring a large number of physically small antennas withinherently large fields of view, and a full cross-correlation architecture that preserves information as farinto the data processing chain as is computationally feasible. The array must also comprise sufficientcollecting area and bandwidth to generate high SNR information on timescales comparable to thoseassociated with ionospheric fluctuations, thereby rendering the calibration problem tractable.

Consequently, an array scope was developed involving ~500 physically small, dipole phased arrays(tiles), with ~10,000 m2 of aggregate collecting area, and full cross-correlation of all tile pairs. From thisfoundation, the rest of the MWA system design flowed. Extensive subsequent analysis and simulationvalidated the above scope parameters, and showed that in addition, the sensitivity of such an array waswell matched to the key science goals. An account of the design goals, tradeoffs and justifications can befound in Lonsdale et al. (2009).

Finally, the need for a radio-quiet site is paramount, and has driven the selection of the Murchison RadioObservatory (MRO) as the MWA site. The MRO is located in Western Australia (WA) on land owned by

7/29/2019 1118466 Hewitt

18/122

14

the Commonwealth Scientific and Industrial Research Organization (CSIRO), and is part of anIndigenous Land Use Agreement between the WA State Government, CSIRO, the Yamatji MarlpaAboriginal Corporation and the Wajarri Yamatji people. It is protected for radio astronomy by Section 19of the WA Mining Act, and comprises an inner restricted zone and an outer coordinated zone where radiotransmitter licensing is regulated. The protected frequency band is from 100 MHz to 25 GHz.

3.2 Array Architecture and Subsystems The MWA system comprises four major subsystems, and anumber of supporting subsystems (see Figure 6). The major subsystems are:

The antenna tile and beamformer, delivering a dual-polarization 80-300 MHz analog RF signal ontwin coaxial cables from each of 512 tiles spread over a 3km extent. The tiles comprise 16 dual-polarization broadband dipoles in a 4x4 array over a ground screen. The tiles are configured assmall phased arrays by analog beamforming, and are electronically steerable down to 30 degreeselevation.

The digital receiver, which is responsible for digitally sampling the incoming RF, applying apolyphase filterbank to generate 256 channels of 1.28 MHz each, from DC to 327.68 MHz,selecting 24 of those channels from the 80-300 MHz range for downstream processing, andformatting the resulting data onto serial streams for transmission to the correlator. Each of the 64digital receivers serves 8 antenna tiles, and resides in the field within an environmentally

controlled and RF-tight enclosure. The correlator, located at the central processing facility within the CSIRO building on site. This

subsystem is responsible for receiving digital data from optical fibers from the digital receivers,filtering the data to high spectral resolution, performing the cross-correlation of all antenna tilepairs and all polarizations, accumulating the results in time, and transmitting the resulting dataover gigabit Ethernet connections.

The real-time system (RTS) calibration software, which runs on an on-site Graphics ProcessingUnit- (GPU-) accelerated computer cluster (the real-time computer, or RTC). The RTS receivesdata from the correlator, and solves for instrumental and ionospheric calibration parameters on an8-second cadence. The primary output data product is calibrated snapshot images as a function offrequency on a standardized sky projection grid, suitable for subsequent scientific analyses.

Supporting subsystems include the data transport network, the monitor and control system, the tied-arraydigital beamformer (using correlator subsystem hardware), the distributed sampling clock system, plusdata storage and general computing in the central facility.

Systems and resources specific to science topics and experiments are not considered to be part of theMWA system. Supporting infrastructure, such as power generation and reticulation, roads, buildings, andfiber connectivity to the outside world, are also not considered to be part of the MWA system.Infrastructure will be provided by CSIRO through cooperative agreements.

7/29/2019 1118466 Hewitt

19/122

15

Figure 6: A flow-chart showing the path of the

signal from antenna front end to delivery of data to

the science groups. The four major subsystems are

identified.

3.3 Antenna and Beamformer The antenna tile is designed for: (a) Zenith collecting area > 10 m2; (b)Field of view from zenith to 60 ZA; (c) Minimal antenna gain at the horizon, to reduce terrestrial RFI;(d) System temperature dominated by sky noise; (e) full polarization response; and (f) Low manufacturing

cost. In addition, the response needs to vary smoothly in time, frequency and angle, for calibration.

The basic antenna element is a dual-polarization active dipole employing vertical bowtie elements, withan integrated LNA/balun at the juncture between each of two orthogonal arms of the bowtie. The bowtiegives a broad antenna pattern and antenna impedance adequately matched to the LNA across the band.The output signal from each LNA is carried via a 7-m coax cable to the analog beamformer. Thetemperature of the sky in cold regions exceeds the contribution from the LNA and beamformer by afactor of a few over most of the frequency range, declining toward 1 at the ends of the range.

An antenna tile consists of 16 dual-polarization dipoles in a square 4x4 configuration with 1.1 metercenter-to-center spacing. The spacing corresponds to /2 at 136 MHz, and was chosen to optimizesensitivity to the EOR signal. The dipoles are attached to a 5m x 5m area of galvanized steel meshserving as a common ground plane for the tile. A tile deployed in Western Australia is shown in Figure

7.The analog beamformer sits beside the antenna tile in a metal enclosure, receives dual polarization signalsfrom all 16 crossed dipoles in a tile, and applies independent delays to each signal to form a tile beam in achosen direction on the sky. True delay steering is employed, with five switchable delay linesindependently controllable for each polarization of each dipole, and each differing in delay by a factor oftwo, implemented as coplanar waveguide on printed-circuit boards. The delayed signals are combined,amplified, Walsh-modulated, and sent over coaxial cable to the receiver for digitization. SubsequentWalsh decoding occurs digitally. The beamformer components and the delay line architecture have been

7/29/2019 1118466 Hewitt

20/122

16

designed for insensitivity to thermal changes, and for minimal RFI generation.

Figure 7: Photo of a MWA tile deployed in Western Australia, consisting of 16 dual-polarization

dipole antennas over a steel mesh ground plane. The analog beamformer that forms and steers the

phased-array tile beam is visible to the upper left. The inset shows the contents of the beamformer

box. The front panel has 16 SMA inputs for each polarization, and one output coaxial connector

for each polarization on the left side. Each dipole signal is routed through switched delay lines on

the PCB. The output power-combined RF signals, as well as power, pointing control commands,

acknowledgment signals and Walsh function waveforms are all simultaneously transmitted on the

two coaxial lines connected to the digital receiver enclosure.

3.4 Digital Receiver The receiver is housed in a cooled enclosure of dimensions 1.6 x 1.1 x 0.6 m,situated centrally with respect to the set of 8 tiles it services. It is supplied with power and fiberconnectivity via cables in a trench network leading back to the CSIRO building some 4km distant.

The incoming RF signals are subjected to anti-aliasing filtering and low frequency rejection, signal leveladjustments, and equalization. The band-limited signals are fed in pairs to dual 8-bit analog-to-digitalconverter chips running at 655.36 Msample/sec. The dual outputs of each ADC chip are fed into apolyphase filterbank implemented in FPGA hardware, yielding channel widths of 1.28 MHz. 24 of thesechannels, comprising 30.72 MHz of RF bandwidth, are selected for further processing, formatted, andtransmitted to the central facility via 3 physical fibers, each carrying 10.24 MHz of bandwidth for all tilesand polarizations. Data are transmitted in the form of 5+5 bit complex samples.

Additional fibers provide Ethernet communications for monitor and control functions and distribute a

centralized clock signal for the samplers, a clock for driving the FPGA logic, and timing signals for arraysynchronization. A single board computer controls the receiver node functions and services M&C needs.

The aggregate data rate for the MWA, with 30.72 MHz of processed bandwidth from 512 tiles servicedby 64 digital receivers, is ~315 Gbits/sec carried on 192 fibers.

3.5 Correlator Cross-correlation for the MWA is a formidable computational challenge, requiring524,800 signal pairs to be multiplied together, each at 30.72 MHz bandwidth, corresponding to 1.61013

7/29/2019 1118466 Hewitt

21/122

17

complex multiply and accumulate (CMAC) operations per second. Furthermore, computationallyexpensive spectral filtering (to 10 kHz spectral resolution) precedes the multiplication step.

The spectral filtering and cross-multiplication are implemented on two different FPGA-based board types,referred to as the polyphase filterbank board (PFB) and the correlator board (CB; see Figure 8). The fullsystem comprises 16 PFBs and 72 CBs, mounted in ATCA card cages, along with a slot-1 controller

board, a full-mesh backplane, and small physical interface boards, known as rear transition modules(RTMs). The full system occupies 8 card cages in 4 standard size racks, and the power consumption is~15 kW.

Figure 8: The MWA correlator board. This photo shows the board with a monolithic heat sinkattached, to keep the 15 FPGA chips cool. The full correlator system will consist of 72 boards of the

type shown, 16 similarly sized boards responsible for spectral filtering, and a number of smaller

interface boards. The system occupies 8 ATCA card cages in 4 standard size racks, with a total

power dissipation of ~15 kW.

The PFB boards execute a 128-channel operation on each 1.28 MHz data stream, yielding a spectralresolution of 10 kHz, and a set of 3,145,728 data streams. These streams are re-ordered to facilitate thecross-multiply operation, and transmitted to the CBs over the backplane.

The cross-multiply and accumulate operation is done on the 10 kHz data streams, using multiplier units inFPGA devices that have typical clock speeds on the order of 250 MHz. The design employs extensive

data buffering and multiplexing in time and frequency for each FPGA multiplier. The system comprises72 correlator boards hosting a total of 576 SX35 FPGA chips containing 110,592 multipliers, and 1.6billion complex cross products are thereby formed within the 100 sec sample interval of the datastreams. These cross products are accumulated for 0.5 seconds before being packaged and transmittedover gigabit Ethernet to the real time computer, which is physically co-located with the correlator.

For the MWA, the delay compensation and fringe stopping operations traditionally used in radioastronomy have been eliminated, and the correlator output is coherent across the entire visible hemispherethanks to small array extent, low observing frequency, and high time/frequency resolution in thecorrelator. All geometric calculations are instead handled by the real time software, simplifying thesystem and adding flexibility, but resulting in high data volumes.

3.6 Real-Time Computer and Calibration Data flow from the correlator to the real-time computer(RTC) at up to 160 Gbits/sec over gigabit Ethernet connections. In the initial RTC implementationrelevant to this proposal, additional frequency averaging in the correlator reduces the rate to 40 Gbits/sec.The data are processed by the RTC to produce high-precision instrumental and ionospheric calibrationsolutions at a rate set by ionospheric conditions and the requirements of the experiment. Currently, theRTC cadence produces solutions once every 8 seconds; this is an important parameter to be investigatedin the initial years of MWA operation as it is a significant cost driver for the MWA and future HERAinstruments.

Software known as the Real-Time System (RTS) has been developed that consists of a visibilityintegrator (time and frequency), the calibration measurement subsystem, and an imaging pipeline

7/29/2019 1118466 Hewitt

22/122

18

(Mitchell et al 2008). The imaging pipeline incorporates gridding, Fast Fourier Transform computation ofimages, and the application of the ionospheric solutions. Integration of the images is performed byregridding them onto a standard HEALPix grid and accumulating the images for a user-specified periodof time. The key elements of the RTS have been coded and have been tested with simulated data.Remaining near-term tasks are implementation of the code for real-time operation and tolerance testing todetermine optimum operating parameters. A long-term task, representing a research effort as described in

Section 5.3.1, is to implement holographic beam correction for real (not simulated) data by characterizingthe antenna primary beams, determining an appropriate scheme for the parameterization of the beams, andexercising and optimizing the gridding step that applies the correction.

The computation described above will require the ability to carry out approximately 2.5 TFLOP persecond (Edgar et al 2010). This requirement will be met by a cluster consisting 32 worker nodes, eachwith two GPUs. Each node will accept 24 frequency channels from the correlator. A master node willconstruct the ionospheric solution from data provided by the worker nodes and carry out other globalcomputations. The currently estimated power consumption of the RTC is 26 kWatts.

4. RESULTS FROM PRIOR NSF SUPPORT: CURRENT STATUS OF THE

MWASupport for the U.S. effort in MWA construction and development has come primarily from the awardAST-0457585, Mileura Widefield Array Science and Technology Demonstrator, by C. Lonsdale (PI)and L. Greenhill, J. Hewitt, and J. Salah (Co-PIs); $5,890,245 (including supplements); June 2006 May2011.

4.1 Subsystem and Prototype Development

4.1.1 Antenna and beamformer: The antenna tiles have undergone multiple design refinements throughlaboratory, anechoic chamber, and field testing. The dipole mechanical and electrical structure has beenfrozen in its third and final version since May 2008. The beamformer delay-line boards reached maturityin March 2008, with minor revisions a year later, while the beamformer communications circuitry hasundergone one major field-tested revision, culminating in the present and final DOC design, verified onsite in September 2010 following prototype testing in July 2009. The antenna LNA circuit design hasbeen unchanged for over five years, but the printed circuit board layout has been revised four times toaccommodate different mounting and input wiring schemes and to simplify maintenance; the versionpresently undergoing testing is designed to give better performance repeatability between dipoles.

As a result of these activities, the following design specifications have been achieved and verified for theMWA antenna and beamformer sub-systems: 1) the receiver noise temperature is less than thetemperature of coldest regions of the sky between 80 and 210 MHz, by factors of a few at prime EoRfrequencies, 2) antenna tile primary beam patterns are consistent with detailed electromagneticsimulations (using FICA, WIPL, and NEC), 3) the antenna tile power response varies smoothly with time,frequency, and pointing angle, and 4) relative delay and gain performance of beamformer transmissionpaths were demonstrated at all delay settings over the full 80-300 MHz range. The first 32 antenna tiles

and beamformers have been manufactured, delivered to site, and assembled all within planned budgetallocations needed for the full 512T construction.

4.1.2. Digital Receiver and Correlator: The basic quantization scheme (8 bit real samples, 5 bits real persample from the receiver in coarse channels, 4 bits complex per fine channel internally within thecorrelator) has been demonstrated, not only in the lab, but also in the presence of strong interferingsatellites and strongly varying system temperatures associated with changing sky brightness. Thehierarchical digital filtering, first to coarse channels in the receiver, then by down-selection, and finally by

7/29/2019 1118466 Hewitt

23/122

19

fine-channelization at the correlator has also been demonstrated. All the basic correlator elements (FXarchitecture, fine channelization, 16x16 high speed correlation cell) that will allow us to process in excessof 500K signal pairs for 512T have been verified. Flexible mapping of the visibility streams tocomputing resources has been implemented via Ethernet switching of the correlator output. (Receiverdevelopment was also supported by Indian and Australian funding sources.)

4.1.3 Monitor and Control: The prototype M&C system was deployed in March 2009, and has steadilyincreased in its capability on each subsequent data taking expedition. The design for the full MWA M&Csystem was finalized in May 2010, and implementation is well underway. The current version of theM&C system is regularly used for all data taking with the 32T prototype array. Current capabilitiesinclude: full remote and/or unattended data taking, control of all available hardware settings of theantennas and receivers (all dipole, delay, pointing, tracking, gain, frequency, and data output modes),control of data acquisition (including hardware correlator, software correlator, and broad-band raw burstdata), dynamic array configuration mapping, dynamic web reporting of the real-time status of the array,database logs of hardware and software performance, and quick look tools for verifying the data quality.(M&C development was also supported by Australian funding sources.)

4.1.4 Real-Time System: The RTS repository presently contains over 50,000 lines of code in .c files;

8,000 lines in .h files; and 11,000 lines in .cu files (NVIDIA CUDA code for GPU acceleration). Thereare currently two branches of the repository, one branch was tuned for 32T imaging and was frozen inearly 2010, and the second branch has sustained active development toward the full 512T GPU-enabledimplementation.

The 32T version of the RTS routinely ingests data from the 32T correlator to produce real-time images. Itused the fully-polarized calibration system, including peeling, and a simplified version of the 512Timaging pipeline. The simplified imaging pipeline consists of gridding using "traditional" anti-aliasingkernels, 2D FFT snapshots, resampling to weighted polarization maps in the HEALPix coordinate frame,and real-time integration of image maps to several minutes. Image maps have then been further integratedoff-line over many MHz of bandwidth and several hours of observations.

The 512T GPU version of the RTS is optimized for modern NVIDIA GPUs, but is similar to the 32Tversion in most respects. A single hardware compute node running the latest version of the 512T RTSsoftware has been shown to achieve the required 8-second calibration and imaging cadence on simulateddata. This includes visibility-dependent kernels large enough to account for primary beam effects (about20x20 pixels), 21x21 degree images, 12 frequency channels (i.e., comparable to the full RTS using 64compute nodes in total), 50 sources peeled (with 2x2 Jones matrix self-calibration and ionosphericmeasurements for each), and the core imaging pipeline.

Through the combined 32T operation and 512T simulation, the RTS engine has been demonstrated towork on real data in real time, and it has been shown to be capable of delivering the required performanceneeded to meet the MWA computational requirements with the planned computing hardware. (RTSdevelopment was also funded in part by the Smithsonian Astrophysical Observatory.)

4.2 The 32-tile Prototype Array (32T)

Fourteen campaigns to the MRO have been conducted between 2006 and 2010 to develop siteinfrastructure and deploy and test MWA prototype hardware, firmware, and software. These activitieshave resulted in the construction and operation of a complete 32-tile prototype array (32T) on site. TheMWA 32T demonstrator has achieved the major milestone of end-to-end operation, including four rack-mounted prototype digital receivers, a hardware correlator, and a real-time calibration and imaging

7/29/2019 1118466 Hewitt

24/122

20

pipeline (RTS). An intensive test and verification (T&V) program has been completed, meeting everygoal set for it and described in the 04 September 2009 memo, MWA 32 T Objectives and QualityAssurance Evaluation Criteria. The full report from February 2010 on T&V measurements,methodology, and results is available. Here we summarize the results.

Tile beam measurements Using autocorrelation spectrum measurements of Orbcomm satellitebeacons, the MWA antenna beams were shown to vary smoothly with angle, and to be wellrepresented by a simple analytic model to better than 15 to 20 dB. Allowing the complex gains ofthe individual dipoles to be free parameters improved agreement with the models to better than 20dB. The beams of different tiles are similar down to 20-30 dB below the peak response. Theseresults support the initial choice of antenna beam models to be used by the RTS in calibration.

RFI characterization An initial description of the RFI spectrum at the MWA site was achievedwith the EDGES (Bowman & Rogers 2010) experiment. Identified signals include FM radio,satellites, aviation, analog TV, and under rare propagation circumstances, digital TV. Effectiveexcision techniques were demonstrated, and are appropriate for similar implementation in theMWA.

Measurements of diurnal variation i