Dear Michael, Dimitri, and Dennis,

Noting that the initial set of SciDac projects are coming to the end of
their 5-year funding period, we anticipate that DOE will issue a call
for new SciDac proposals in the near future.  For reasons outlined
below, we believe that a proposal focusing on computational solutions
to fundamental problems in nuclear structure and reactions is both very
important and very timely. This research will address some of the most
fundamental properties of nuclei, namely how they are constructed and
how they interact with each other and with external probes.  It will
also advance our ability to use the nucleus as a probe for physics as
diverse as nuclear astrophysics, neutrino physics and fundamental
symmetries, and to understand and characterize nuclear reactions of
importance to national needs in energy research, stockpile stewardship,
and threat reduction.

Important progress has been made with recent advances in computing
and computational techniques, but  increased computational resources,
algorithmic development and a coherent effort spanning all of
computational nuclear structure and reactions is important for the
future development  of nuclear physics.  The proposal we envision will
tie together important goals of the Office of Science Nuclear
Physics Program (SC/NP), the SciDac program for advancing computational
capabilities, and to the ASC program for Science Based Stockpile
Stewardship within NNSA.

The undersigned are all involved in applying high-performance computing
to problems within low-energy nuclear physics.  We have begun
discussions on collaborating in an effort to make substantial advances
in the scope and accuracy of these calculations.  We outline here what
will likely be our major thrusts within the proposal.

We envision a multipronged and overlapping effort that will
address difficult problems in areas of nuclear structure and
reactions relevant to the various missions of SC/NP and NNSA.
We anticipate an expansion of the computational techniques and
methods we currently employ to better utilize a variety of
machine architectures that exist or are planned in the future.
Broadly speaking, there are two major thrusts within nuclear
structure theory.  The first involves ab initio approaches that
seek to determine the properties of nuclei starting
from our best knowledge of internucleon interactions.  They are
currently being used to study lighter nuclei.  The second thrust,
based on mean-field theory, is necessarily more phenomenological,
but provides an ability to calculate basic properties of medium
mass and heavy nuclei.

Of the ab initio methods, we single out several that are at the
computational frontier.  Green's Function Monte Carlo (GFMC),
has achieved a number of important milestones, including the
demonstration that realistic two-nucleon (NN) interactions
and plausible three-nucleon (3N) interactions can reproduce
the binding energy and excitation structure of light nuclei up
to A=10, and that the 3N interactions contribute significantly
to the spin-orbit field and the stability of the Borromean nuclei.
The no-core shell model (NCSM), has confirmed many of the GFMC
findings with a broader range of interactions and for slightly
larger nuclei, achieving an impressive spectroscopy of excited
states in light nuclei.  The coupled-cluster method (CCM), offers
substantial promise for extension to larger nuclei like 16O and 40Ca.
We are also developing auxiliary-field diffusion Monte Carlo (AFDMC)
techniques, which are currently able to handle large numbers of
neutrons (N=114) in a box, as a model for dense matter in neutron
stars.

By considering these theories under a common collaborative
umbrella, we could validate the results and assess their
accuracy by comparing one method to another as well as to
experiment.  Each of these methods faces substantial
computational demands and requires significant algorithmic
developments to fully take advantage of current and planned
architectures.  The full proposal will outline the specific
problems to be addressed and the computational resources
required to achieve designated milestones.

For heavier nuclei, methods anchored in mean-field theory will explain
many features of nuclear structure across a wide range in mass. As
currently used, these approaches are very similar to
density-functional theory (DFT) for electronic systems, requiring as
input an energy functional and providing the ground state structure of
any nucleus.  It is only recently that calculations and optimization
of density functional parameters have been carried out across the
entire chart of nuclides.  Much more can be done to improve the
functionals if more computational resources are applied to the
problem.  As an example, it is known from fundamental considerations
and from electron DFT that an exact treatment of exchange needs to be
included.  More flexibility in the functionals would allow connections
to ab-initio approaches, which would improve and validate the
functionals.

Several promising extensions of DFT exist, and all require extensive
computational resources.  The time-dependent DFT (TDDFT) permits
computation of collective excitations and transitions, and its
real-time implementation from nuclear physics has been applied in
other fields.  The TDDFT in imaginary time is the most fundamental
theory available for some dynamic processes like spontaneous fission.
Another dynamical extension of DFT, the generator coordinate method,
is at present the leading approach for several structure problems
ranging from nonshell (intruder) states in medium-mass nuclei to
low-energy spectroscopy in heavy nuclei.  The systematic application
of these techniques requires computational resources well beyond what
is currently available.

For medium and heavy nuclei, the auxiliary-field shell model Monte
Carlo (SMMC) method is very effective.  It can treat extremely large
configuration spaces and has produced successful microscopic
calculations of level densities and nuclear partition functions in the
presence of correlations.  However, the interactions used
are necessarily phenomenological and specific to mass regions.  We
hope that the collaboration would lead to the derivation of nuclear
interactions that are based on ab initio input, both for the DFT and
the SMMC. Furthermore, we plan to merge the mean field and shell model
approaches to nuclear structure by developing a mapping of the DFT
energy functional onto effective shell model Hamiltonians.  We can
then combine the advantages of both approaches: the global
parameterizations of the DFT across the table of nuclides and the
higher accuracy of the correlated shell model calculations.  Because
in heavy nuclei the minimal truncated spaces required for a successful
mapping are prohibitively large for conventional shell model
calculations, the SMMC approach is indispensable.

Tying nuclear structure directly to nuclear reactions within a
coherent framework applicable throughout the nuclear landscape is an
important goal. For light nuclei, ab initio methods hold the promise
of direct calculation of low-energy processes, including reactions
occurring in thermonuclear environments, such as stars, and tests of
fundamental symmetries. For heavier nuclei, the continuum shell model
and modern mean-field theories allow the consistent treatment of open
channels, linking the description of bound states, unbound states, and
direct reactions. In reaction theory, a better treatment of nuclear
structure is equally crucial. The battleground in this task is the
newly opening territory of weakly bound nuclei where the structure and
reaction aspects are interwoven and the interpretation of future data
will require advances in the understanding of reaction mechanisms.

At present, reaction theory has many approximate methods for different
types of reactions; a comprehensive picture based on the microscopic
nature of nuclei is lacking.  Present practice relies heavily on
approximate theories and empirical tuning to known data.  While
practical for nuclei where some information is available, this
approach has limited predictive power so that there are significant
uncertainties in many applications relevant for the SC/NP and SBSS
programs. The primary sources of these limitations are the complexity
and computational demands of a proper treatment of all relevant
degrees of freedom.  Indeed, the complexity of the reaction problem
has led to this topic being considerably less mature than many
approaches for nuclear structure. Because of recent advances in
computer technology, however, it is time to take a fresh look at the
theoretical program for low-energy nuclear reactions and the
integration of nuclear structure into reaction theory.
Computational methods range from Faddeev-Yakubovsky and Correlated Hyperspherical
Harmonics for few-nucleon systems to extensions of ab-initio methods
above to mean-field and coupled-channel methods for larger systems.
The complexity of the problem demands nothing less than a concerted effort to
coordinate research efforts between several subfields in nuclear
physics and the computational sciences.

In summary, our proposal will bring together the top
computational nuclear theorists with top computational
scientists to take the study of nuclei to a new level of
sophistication.  This research incorporates the breakthroughs of
the last five years with new techniques and algorithmic
developments to take advantage of current and planned parallel
architectures. We believe that this research will advance
quantum many-body theory across disciplines, and produce
techniques and codes valuable in areas as diverse as
cold atoms, quantum dots, and quantum liquids, as
well as in both realistic and effective interaction calculations
of nuclei.

Our proposal will require substantial new
large-scale computing resources. This effort is relevant to
SC/NP, and NNSA-SBSS, and will require collaboration and
integration of computational scientists from the various ISICS
components of SciDAC (including TOPS and CCA).
We expect to hold a meeting on this collaboration in the near
future which will include the undersigned as well as others
interested in these goals.  In the meantime please consider
David Dean below as our contact person.

Sincerely,

Yoram Alhassid (Yale)
George Bertsch (INT, University of Washington)
Joe Carlson (LANL)
David Dean (ORNL)
Erich Ormand (LLNL)
Steven Pieper (ANL)
Kevin Schmidt (Arizona State)
Rocco Schiavilla (TJNAF)
James Vary (Iowa State)