Plasma State Pitfalls

Document Actions

Here are some open holes that I (and others) have recently fallen into. Nota bene:

1) Defining coordinate grids in Plasma State. Coordinate grids have different properties and requirements in Plasma State than other arrays. Quoting Doug McCune:

To completely define a grid, there are 3 steps:

1) define its size:

         ps%nrho_icrf = inum

2) allocate the grid and associated profile arrays:

         call ps_alloc_plasma_state(ierr, state=ps)

      (state argument optional).

3) fill in the grid's individual values. E.g. for an evenly spaced grid:

         do ii=1,inum
            ps%rho_icrf(ii) = (ii-1)*1.0_rspec/(inum-1)
         enddo

ALL THREE STEPS must be completed before calling ps_store_plasma_state.

Our mistake was to call ps_store_plasma_state before filling the grid with data and attempting to fill it later. Whereas other arrays evolve in time, grids do not. They are defined for the whole simulation and must be initialized with the grid values before being stored. Different grids can be defined at different times with multiple calls to ps_alloc_plasma state and ps_store_plasma_state, but a single grid needs to be done completely before calling ps_store_plasma_state.

2) There are several different types (7 at last count) of plasma species in the state. [thermal, fast, minority, , all] and each exists as a separate list. Some of these lists can have zero species in them, but species "all" has to be set or you will get plasma state errors. "all" is a separate list containing a union of all the other lists. You set "all" by just calling the function ps_merge_speices_lists(ierr). It is required that you call ps_merge_speices_lists even if you only have thermal ion species. If you don't do this you will get Plasma State internal error messages that look like:

?ps_name_check: blank names detected in: all_name
   list is dimensioned (           0 :           1 ).
   ...name #           0 is blank.
   ...name #           1 is blank.

(Default Message) error "state_file_write" in routine PS_STORE_PLASMA_STATE

failed to write state to file: TEST_monitor_ps.cdf

3) Every plasma species has: name (character), charge, mass, type, atomic charge when fully stripped. I'm not sure whether the state requires that all of these be set, but certainly the name must be. There exists a handy function in the plasma state called ps_label_spectype that will set all these things for you. The calling sequence is

SUBROUTINE ps_label_spectype(iZatom,iZcharge,iAMU,itype, &
       label, qatom, qcharge, mass)

    !--------------------------------
    ! D. McCune Oct 10 2006
    ! module subroutine for plasma species labeling
    !src
    ! ...to be used during initialization of plasma state system at the
    ! start of a simulation...
    !
    ! Mod dmc Mar 2007-- DELETED: tokamakium support.
    !--------------------------------

    integer, intent(in) :: iZatom    ! atomic number: 1 for H, 2 for He, ...
    ! special: -1 for electrons

    integer, intent(in) :: iZcharge ! ionic charge btw 1 and iZatom
    ! ignored for electrons

    integer, intent(in) :: iAMU      ! atomic weight in AMU, nearest integer
    ! ignored unless 1 <= iZatom <= 2

    integer, intent(inout) :: iType     ! species "type", defined as follows:
    ! For thermal species, itype can be input or output:
    !   if itype.eq.-1: electrons (thermal; blank suffix)
    !   if 1.le.itype.le.3: thermal ion species (blank suffix)
    !
    !   if itype.eq.0: *modify on output* according to value of iZatom:
    !          -1 for electrons (iZatom=-1)
    !           1 for thermal ions (1 <= iZatom <= 2)
    !           2 for "impurity" thermal ions (iZatom > 2)
    !
    ! To specify a non-thermal species, a type code must be provided as input:
    !
    !   if itype.eq.4: beam specie (suffix="beam")
    !   if itype.eq.5: RF Minority specie (suffix="rfmi")
    !   if itype.eq.6: fusion product species (suffix="fusn")

! if itype.eq.7: fast electrons

!-------------

    character*(*), intent(out) :: label   ! generated species label

    REAL(KIND=rspec), intent(out) :: qatom    ! charge of fully stripped ion
    REAL(KIND=rspec), intent(out) :: qcharge ! actual charge

REAL(KIND=rspec), intent(out) :: mass ! mass in kg

This can be found in plasma_state_mod.f90. Examples of its use are in the swim_state_test.f90 program in the trunk/components/state/src/test directory, and in my simple fake EPA component trunk/components/epa/change_n_T/change_n_T_EPA.f90. Here's what it looks like in action:

ps%s_type = 0

CALL ps_label_spectype(-1, -1, 0, ps%s_type(0), &
ps%s_name(0), ps%qatom_s(0), ps%q_s(0), ps%m_s(0)) ! electron

CALL ps_label_spectype(1, 1, 2, ps%s_type(1), &
ps%s_name(1), ps%qatom_s(1), ps%q_s(1), ps%m_s(1)) ! Deuterium

You put in (charge, atomic number ->charge-fully-striped, mass rounded to nearest au, and specied type 0 -> thermal) and you get out all the ps% stuff you need (including the character name). Using this is a good way to be sure we all use EXACTLY the same ion masses every time.

4) Going through this exercise has convinced me that every component needs an init function that allocates and sets all of the grids, and that allocates and initializes all of the time evolving profiles and other data that that component provides. That way we will know where everything is initialized. We will be in much better shape to sort out the requirements of components that depend on initial data from other components in order to intialize themselves.

Happy SWIMming.

DBB

--

Donald B. Batchelor
Plasma Theory Group

Phone: (865) 574-1288
Fax: (865) 576-7926
E-mail: batchelordb@ornl.gov

Oak Ridge National Laboratory
Fusion Energy Division
P. O. Box 2008
Oak Ridge, TN 37831-6169

Created by bernhold
Last modified 2008-02-27 09:44

CSWIM

Sections

Personal tools

Navigation

News

Plasma State Pitfalls

Document Actions