RESEARCHERS MAKE IMPORTANT STEP IN MODELING PLASMA BEHAVIOR         12.08.95
  SCIENCE & ENGINEERING NEWS                                            HPCwire
  =============================================================================
    Ithaca, NY -- Using Cornell Theory Center's 512-processor IBM RS/6000
  PowerParallel System (SP) Professor John Dawson and colleagues at the
  University of California at Los Angeles, have made an important step in
  modeling the behavior of plasma turbulence in magnetic confinement devices
  for fusion such as the Tokamak Fusion Test Reactor (TFTR) at Princeton
  University. Dawson's virtual reactor will help guide changes in the
  engineering design of test reactors such as TFTR and brings the country an
  important step closer to a working fusion energy plant.
  
    "CTC's SP is currently the best machine in the country where we can do the
  kinds of things we want to do," noted Dawson. "Using the SP has allowed us
  to perform the largest production run simulation in the Numerical Tokamak
  Grand Challenge project. This has also been the most stable operating
  environment for production simulation."
  
    The Numerical Tokamak Project (NTP) is a multidisciplinary program to
  develop the computational resources, software, hardware, and communications
  tools needed to simulate the complex features of the dynamics within the
  chamber of a tokamak reactor. NTP is a joint effort among more than a dozen
  academic institutions and government laboratories.
  
    The tokamak (an acronym created from the Russian words TOroidalnaya KAmera
  ee MAgnitnaya Katushka, which means toroidal chamber and magnetic coil) is
  a doughnut-shaped reactor designed to confine plasma within a complex and
  powerful magnetic field. Plasma is considered a fourth state of matter in
  which many of the atoms or molecules are ionized. Current tokamak efforts
  focus on the fusion of the isotopes deuterium and tritium.
  
    To begin the fusion reaction, practical reactors are thought to require an
  internal plasma temperature of 100 million degrees Celsius (six time the
  internal temperature of the Sun). Once the reaction is underway, the
  containment system must hold up and allow the reaction's behavior to be
  controlled.
  
    Dawson's group has been trying to understand the behavior of turbulence in
  the plasma that causes the hot core plasma to mix with the cooler outer
  layer near the wall. If mixing occurs too quickly, the heat of the core
  falls below the critical temperature needed to sustain the reaction.
  
    As plasma courses through the tokamak, it is constrained into a torus shape
  in response to magnetic and electrical forces. Fusion occurs as the ionized
  nuclei collide in the turbulent cloud within the core of the plasma. High
  energy neutrons, a by-product of the fusion reaction, shoot off in straight
  lines through the reactor walls where they heat a surrounding layer of
  lithium. This heat powers a conventional steam generator that produces
  electricity.
  
    Dawson's breakthrough came after his team parallelized and optimized his
  simulation code and moved it from the Cray-C90 to the SP. "This was the
  most dramatic change I've ever witnessed," said Rick Sydora, a member of
  Dawson's research group. "An order of magnitude increase in memory and
  power. It has allowed us to do breakthrough science."
  
    Previously Dawson's group had simulated plasma behavior for a tokamak at
  one-fifth the scale of the TFTR, which is the largest operating magnetic
  fusion experiment in the United States. When they simulated the turbulent
  behavior for a device size on the order of the TFTR at Princeton, they
  observed new features in the development of the turbulence and rates of
  thermal diffusion in much better agreement with experiments. "We also found
  that if we reverse the pitch of the field lines in the core region of the
  plasma our full nonlinear model leads to improved particle and thermal
  confinement which has recently been observed in the TFTR and other
  tokamaks" said Sydora. "These new results are a consequence of the better
  resolution possible on the SP," added Viktor Decyk, another member of the
  research group.
  
    This result is also an important step in the ability to predict the
  behavior of proposed tokamaks and other magnetic confinement schemes. The
  virtual tokamak holds the promise of optimizing and assisting in the
  engineering design of real reactors. Simulating the effect of design
  changes will be far more effective than spending millions of dollars
  implementing physical modifications to billion dollar test reactors that
  don't work as expected. The cost-savings will be even more dramatic when
  engineers move from test reactors to the first working fusion plant.
  
    The goal of the test reactors in the U.S., Europe, Japan and Russia is to
  develop systems that can draw off enough energy to sustain their own
  reaction along with enough additional energy left to generate electricity
  for a very short time. TFTR sustained a reaction that produced 10.7 million
  watts of energy for a few seconds in 1994. This amount of energy was a
  record 100 million times greater than what was possible 20 years ago.