SPARTA WWW Site - SPARTA Documentation - SPARTA Commands

fix ave/grid command

Syntax:

fix ID ave/grid group-ID Nevery Nrepeat Nfreq value1 value2 ... keyword args ... 

Examples:

fix 1 ave/grid all 10 20 1000 c_mine
fix 1 ave/grid all 1 100 100 c_2[1] ave running
fix 1 ave/grid all 1 100 100 c_2[*] ave running
fix 1 ave/grid section1 5 20 100 v_myEng 

These commands will dump averages for each species and each grid cell to a file every 1000 steps:

compute 1 grid species n u v w usq vsq wsq
fix 1 ave/grid 10 100 1000 c_1[*]
dump 1 grid all 1000 tmp.grid id f_1[*] 

Description:

Use one or more per-grid vectors as inputs every few timesteps, and average by grid cell over longer timescales, applying appropriate normalization factors. The resulting per grid cell averages can be used by other output commands such as the dump grid command. Only grid cells in the grid group specified by group-ID are included in the averaging. See the group grid command for info on how grid cells can be assigned to grid groups.

Each input value can be the result of a compute or fix or grid-style variable. The compute or fix must produce a per-grid vector or array, not a global or per-particle or per-surf quantity. If you wish to time-average global quantities from a compute, fix, or variable, then see the fix ave/time command. To time-average per-surf quantities, see the fix ave/surf command.

Each per-grid value of each input vector is averaged independently.

Computes that produce per-grid vectors or arrays are those which have the word grid in their style name. See the doc pages for individual fixes to determine which ones produce per-grid vectors or arrays.

Note that for values from a compute or fix, the bracketed index I can be specified using a wildcard asterisk with the index to effectively specify multiple values. This takes the form "*" or "*n" or "n*" or "m*n". If N = the size of the vector (for mode = scalar) or the number of columns in the array (for mode = vector), then an asterisk with no numeric values means all indices from 1 to N. A leading asterisk means all indices from 1 to n (inclusive). A trailing asterisk means all indices from n to N (inclusive). A middle asterisk means all indices from m to n (inclusive).

Using a wildcard is the same as if the individual columns of the array had been listed one by one. E.g. these 2 fix ave/grid commands are equivalent, since the compute grid command creates a per-grid array with 3 columns:

compute myGrid all all u v w
fix 1 ave/grid all 10 20 1000 c_myGrid[*]
fix 1 ave/grid all 10 20 1000 c_myGrid[1] c_myGrid[2] c_myGrid[3] 

The Nevery, Nrepeat, and Nfreq arguments specify on what timesteps the input values will be used in order to contribute to the average. The final averaged quantities are generated on timesteps that are a multiple of Nfreq. The average is over Nrepeat quantities, computed in the preceding portion of the simulation every Nevery timesteps. Nfreq must be a multiple of Nevery and Nevery must be non-zero even if Nrepeat is 1. Also, the timesteps contributing to the average value cannot overlap, i.e. Nfreq > (Nrepeat-1)*Nevery is required.

For example, if Nevery=2, Nrepeat=6, and Nfreq=100, then values on timesteps 90,92,94,96,98,100 will be used to compute the final average on timestep 100. Similarly for timesteps 190,192,194,196,198,200 on timestep 200, etc.


If a value begins with "c_", a compute ID must follow which has been previously defined in the input script. If no bracketed term is appended, and the compute calculates a per-grid vector, then the per-grid vector is used. If c_ID[I] is used, then I must be in the range from 1-M, which will use the Ith column of the M-column per-grid array calculated by the compute. See the discussion above for how I can be specified with a wildcard asterisk to effectively specify multiple values.

Users can also write code for their own compute styles and add them to SPARTA.

If a value begins with "f_", a fix ID must follow which has been previously defined in the input script. If no bracketed term is appended, and the fix calculates a per-grid vector, then the per-grid vector is used. If f_ID[I] is used, then I must be in the range from 1-M, which will use the Ith column of the M-column per-grid array calculated by the fix. See the discussion above for how I can be specified with a wildcard asterisk to effectively specify multiple values.

Note that some fixes only produce their values on certain timesteps, which must be compatible with Nevery, else an error will result. Users can also write code for their own fix styles and add them to SPARTA.

If a value begins with "v_", a variable name must follow which has been previously defined in the input script. Only grid-style variables can be referenced. See the variable command for details. Note that grid-style variables define a formula which can reference stats_style keywords, or they can invoke other computes, fixes, or variables when they are evaluated, so this is a very general means of specifying quantities to time average.


For averaging of a value that comes from a compute or fix, normalization is performed as follows. Note that no normalization is performed on a value produced by a grid-style variable.

If the compute or fix is summing over particles in a grid cell to calculate a per-grid quantity (e.g. energy or temperature), this takes the form of a numerator divided by a denominator. For example, see the formulas discussed on the compute grid doc page, where the denominator is 1 (for keyword n), or the number of particles (ke, mass, temp), or the sum of particle masses (u, usq, etc). When this command averages over a series of timesteps, the numerator and denominator are summed separately. This means the numerator/denominator division only takes place when this fix produces output, every Nfreq timesteps.

For example, say the Nfreq output is over 2 timesteps, and the value produced by compute grid mass is being averaged. Say a grid cell has 10 particles on the 1st timestep with a numerator value of 10.0, and 100 particles on the 2nd timestep with a numerator value of 50.0. The output of this fix will be (10+50) / (10+100) = 0.54, not ((10/10) + (50/100)) / 2 = 0.75.


Additional optional keywords also affect the operation of this fix.

The ave keyword determines what happens to the accumulation of statistics every Nfreq timesteps.

If the ave setting is one, then the values produced on timesteps that are multiples of Nfreq are independent of each other. Normalization as described above is performed, and all tallies are zeroed before accumulating over the next Nfreq steps.

If the ave setting is running, then tallies are never zeroed. Thus the output at any Nfreq timestep is normalized over all previously accumulated samples since the fix was defined. The tallies can only be zeroed by deleting the fix via the unfix command, or by re-defining the fix, or by re-specifying it.


Restart, output info:

No information about this fix is written to binary restart files.

This fix produces a per-grid vector or array which can be accessed by various output commands. A vector is produced if only a single quantity is averaged by this fix. If two or more quantities are averaged, then an array of values is produced, where the number of columns is the number of quantities averaged. The per-grid values can only be accessed on timesteps that are multiples of Nfreq since that is when averaging is performed.

This fix performs averaging for all child grid cells in the simulation, which includes unsplit, split, and sub cells. Section 4.8 of the manual gives details of how SPARTA defines child, unsplit, split, and sub cells.

Grid cells not in the specified group-ID will output zeroes for all their values.

Restrictions:

If performing on-the-fly grid adaptation every N timesteps, using the fix adapt command, this fix cannot time-average across time windows > N steps, since the grid may change. This means Nfreq cannot be > N, and keyword ave = running is not allowed.

Related commands:

compute, fix ave/time

Default:

The option defaults are ave = one.