GRAPE newletter vol.4 ( May 26 2003)

Dear Colleagues:

This is the fourth issue of our GRAPE newsletter, a brief summary of
recent developments regarding the GRAPE special-purpose computer for
stellar dynamics.  For further information: "http://www.astrogrape.org".

       +----------------------- CONTENTS:-----------------------------+
       | 1) Baby GRAPE Development Status
       | 2) GRAPE-6 Result on Black Holes in Globular Clusters
       | 3) MUV (Mitaka Underground Vineyard) Current Status
       | 4) Gordon-Bell Performance Prize 2002
       | 5) Reports from the field: GRAPEs in Tokyo
       +--------------------------------------------------------------+

1---------1---------1---------1---------1---------1---------1---------1

 BABY GRAPE-6 DEVELOPMENT STATUS

For those of you who did not have funds to buy a full board version of
the GRAPE-6, we will soon have a cheaper smaller version, the so-called
Baby GRAPE-6, with four chips on a PCI card, which you can install
directly in the computer on your desk, instead of adding a special box.
Surprise your friends with your mystery machine doing stellar dynamics
at 120 Gflops with the humble appearance of a desk-top PC!

Toshi Fukushige has produced a prototype system that has been operational
since January 2003, and we hope Hamamatsu Metrix will be able to ship the
commercial version soon.  When ready, it can be installed on standard
x86 PCs running Linux.

2---------2---------2---------2---------2---------2---------2---------2

 GRAPE-6 ON BLACK HOLES IN GLOBULAR CLUSTERS

On January 10, NASA together with a group of GRAPE users (Holger
Baumgardt, Jun Makino, Piet Hut, Steve McMillan and Simon Portegies
Zwart) offered a joint press release on simulations of the globular
cluster M15, performed on the GRAPE-6 system in Tokyo, under the
title of `Researchers Seek "Heart" of Black Hole Mystery'; see
"ftp://ftp.hq.nasa.gov/pub/pao/pressrel/2003/03-009.txt" .
It refers to an article by the above authors, published in 2003,
Astrophys. J. Lett. 582, L21-L24.  In response to an earlier claim
based on HST observations of a central black hole in M15 of 4,000
solar masses, later corrected to 2,000 solar masses, this letter
showed how a standard core collapse simulation without central
black hole was also perfectly consistent with the observations.

Another publication in 2003, in Astrophys. J. Lett. 589, L25-L28
by the same group showed that the giant globular cluster G1 in M31
can also be modeled very well without recourse to a central black
hole, as had been claimed to have been observed as having a mass
of 20,000 solar masses, based on HST observations.

3---------3---------3---------3---------3---------3---------3---------3

 MUV (Mitaka Underground Vineyard) Current Status

Almost one year has passed since the Astronomical Data Analysis Center
(ADAC) of the National Astronomical Observatory of Japan (NAOJ) started
the open use of the GRAPE-6 system at Mitaka Underground Vineyard (MUV).
As of May 2003, the MUV system consists of 16 GRAPE-5s with 833MHz
Alpha 21264 front-ends and 8 GRAPE-6s with 2GHz Pentium 4 front ends.
All these front-end host computers are connected via the fast network
Myrinet2000.  You can find further information on the English MUV web
pages at "http://www.cc.nao.ac.jp/muv/index_e.html" .

Twice a year ADAC puts out a call for proposals for simulation projects
using the MUV GRAPEs.  Currently, according to ADAC open use policy,
only researchers studying astronomy at Japanese institutions can be
principal investigators (PIs) of projects, unless otherwise permitted
by the director of ADAC/NAOJ.  Astrophysics researchers at institutions
outside Japan can submit a proposal as co-investigators.  In that case,
they need a Japanese collaborator as a PI.  If you are interested in
using the MUV system, please check the online application form posted at
"http://www.cc.nao.ac.jp/cc/public/documents-e.html" .

MUV Administrator, Eiichiro Kokubo (NAOJ), email: kokubo@yso.mtk.nao.ac.jp

4---------4---------4---------4---------4---------4---------4---------4

 GORDON-BELL PERFORMANCE PRIZE 2002

After the GRAPE family of computers has won five Gordon Bell prizes,
over the last seven years, our entry did not win the Performance Prize
in November 2002.  Interestingly, our GRAPE-6 entry of a simulation
of the Kuiper Belt region using 1.8M particles showed a sustained
performance was 29.5 Tflops, on a 64-board, 16-host parallel
GRAPE-6 system, and thus presented the fastest-ever actual simulation.
In contrast, the prize was given to a climate model simulation on the
Earth simulator, with a best performance (for 10 timesteps) of 26.6
Tflops and sustained performance (for 1-day calculation) of 23.9 Tflops.
Equally interestingly, the Earth simulator application was mentioned as
a new accomplishment, which implies that the GRAPE family of computers
is by now so robust and established that we do not need any new type of
accomplishment, while still retaining the first place in terms of speed.

5---------5---------5---------5---------5---------5---------5---------5

 REPORTS FROM THE FIELD: GRAPES IN OOKAYAMA, TOKYO

 by Shigeru Ida (email: ida@geo.titech.ac.jp)

 web site (in Japanese): http://www.geo.titech.ac.jp/lab/ida/ida/top.htm

GRAPE computers have been used at the Department of Earth and Planetary
Sciences, Tokyo Institute of Technology, since 1993, when Shigeru Ida
moved from the University of Tokyo at Komaba (the original GRAPE yard).

Collaborating with Eiichiro Kokubo, who was a grad student at the
University of Tokyo at Komaba, and now is at the National Astronomical
Observatory, Japan, we started N-body simulations of planetesimal
accretion with GRAPE-2 and GRAPE-4.  We first performed simulations in
a radially local region (a ring-like region) with radial width 0.02 AU
of a planetesimal disk to find that only one body in this region grows
predominantly, which is called "runaway growth".  Although runaway
growth had been predicted by statistical calculations of coagulation
equations, our N-body simulations with wider radial ranges revealed a
"non-linear" stage of runaway growth.  When runaway growth proceeds to
some degree, the growth slows down.  Other bodies run away and catch
up with the first one, resulting in a two-component system consisting
of a small number of runaway bodies and a large number of small
planetesimals.  We also found that orbital distances between the
runaway bodies are almost equal (about 10 Roche-lobe radii of the
bodies).  This is "oligarchic growth".  We also apply our simulations
to planetesimal disks with different masses to explain the diversity
of discovered extrasolar planets.

We have been also working on dynamics of planetary rings and evolution
of proto-moon disks generated by giant impacts between protoplanets.
Because planetary rings and most parts of proto-moon disks are within
the Roche limits of host planets, collisions of the particles do not
result in gravitational binding but in rebounds by tidal forces of the
host planets.  Through N-body simulations with GRAPE-4 and GRAPE-5, we
found that when the particles are distributed densely, the collisions
and rebounds generate density waves in a ring or a disk and angular
momentum is relatively rapidly transferred.  In the case of a
proto-lunar disk, this rapid angular momentum transfer pushes
particles out from the Roche limit, allowing accretion of moons
outside the Roche limit.  We found that the number of moons decreases
and their sizes increase with mass of a disk relative to the host
planet's mass.  In the case of Earth's moon, the disk would be so
massive that one big moon would have been formed.  On the other hand,
the disks around Jovian planets would not be massive enough compared
with the host planets, so that several moons were formed.  Planetary
rings may be remnants of the disks.

0---------0---------0---------0---------0---------0---------0---------0

           Piet Hut and Jun Makino
           (submissions to: piet@astrogrape.org or grape@astrogrape.org)


HOW TO (UN)SUBSCRIBE:  send an email message to grape@astrogrape.org
                       with Subject: (un)subscribe

00--------00--------00--------00--------00--------00--------00--------00