MIXFRAC

MIX3 and BLF calculate major- and trace-element differentiation trends for crystallizing magmas. The programs have numerous options, including assimilation of a second material, assimilation of boundary layer material, etc. The distributed code is compiled for Intel (PC) processors. Users of other computers will need to re-compile the code. On mainframe computers, you will want to delete most of the lines that output to the screen (WRITE(*,...) ). As you can see, the program has been renamed since we did the work for the GCA and Reviews in Mineralogy papers. This new version has additional options, which allow one to calculate boundary layer crystallization, and randomize the mixing and sampling parameters in order to create a more realistic synthetic data set. This copy also contains our new experimentally determined REE, Sc and Y augite-melt partition coefficients.

The Included Files Programs Are ...

MIX3.FOR
BLF.FOR
Text files containing source code. Most editors and word processors can read them. MIXFRAC calculates low-pressure differentiation trends assuming homogeneous crystallization. BLF calculates the effect of in situ or boundary-layer crystallization.

MIXNFRAC.DAT
BLF.DAT
Input files, includes the starting composition, assimilant composition and model parameters.

MIX3.EXE
BLF.EXE
Executable files (PC version, recompile source code for other platforms).

MIXPLOT
BLFPLOT
Tab-delimited output file containing percent fractionation, temperature and melt compositions of points along the calculated liquid line of descent. Stored as a text file so that it can be read directly into a spreadsheet.

MIXNFRAC.OUT
BLF.OUT
Full output, including mineral compositions.

NEWDS3.DAT
Input file for the executibles BLF and MIX3. This file contains the partitioning information for the trace elements.

RANDOM.DAT
Table of 10000 random numbers.

Input Data Format

The input formats for these two versions of the differentiation program are fairly similar. The input format for MIXFRAC is described in detail with the variatios for BLF given later. MIX3.EXE reads the starting composition and run conditions from a file called MIXNFRAC.DAT. This file must be edited if you want to change any of the model parameters. The first line of MIXNFRAC.DAT is the title of the run, it is followed by the number of elements to be modeled (maximum of 60). The next section is the major element input data. The name of each major component is followed by its concentration. After MnO, the trace element data are entered. The trace element's name is followed by its concentration in ppm. Its atomic weight, the slope and intercept for the partitioning of that element for each of the seven minerals being modeled in the order olivine, orthopyroxene, clinopyroxene, plagioclase, spinel, ilmenite, pigeonite and apatite are contained in the file NEWDS3.DAT. That file should not need updating except when new partitioning data becomes available.

The next input is the composition of the assimilant. Each element's name is followed by the concentration of that element, weight % for the majors, and ppm for Cr, Ni, and other trace elements in the same order as the starting composition was given.

The next eight entries are the fractionation factors for each mineral. A value of zero represents equilibrium crystallization, 1 is fractional crystallization. The recharge, assimilation and eruption factors are next and are described in detail in Nielsen (1988 Comp & Geosci.), and in comment statements. MIX3.EXE outputs results to disk at intervals of % crystallization. This requires an additional input line for that parameter. An interval of 2 % is appropriate for simple fractionation, 10-50% for open system cases. The next four numbers are, in order, the number of log units above or below QFM the oxygen fugacity is to be set, the periodic mixing percentage (set to 0 for continuous mixing), the output interval (in %), and the crystallization increment (in %). The final group of numbers is a set of flags for the randomization of the sampling interval, periodicity of mixing, and the amount of recharge, assimilation and eruption. A value of 1 will cause these parameters to be randomized, any other value will bypass the randomization. The randomization seed sets your initial position within the table of random numbers loaded from RANDOM.DAT.

BLF.EXE outputs data after each increment of solidification-zone magma is mixed into the magma chamber. In addition, there are two inputs needed to tell the computer the size of the solidification zone relative to the magma chamber (0.05 is appropriate), and the % crystallization in the solidification zone (5-90 %). If these values are set to 0.99 and 99 respectively, the program will calculate a simple homogeneous fractional crystallization liquid line of descent.

Useage Notes

The following points address the most commonly encountered questions about the operation of the models.

1. Input Fe⁺³/Fe⁺²should be close to the appropriate value for the T and f₀₂of the run. This is due to the sensitivity of the spinel calculations. An unrealistic ferric/ferrous ratio will cause the program either to calculate the wrong spinel or to crash.

2. The program will not run without values for Mg, Al, Si, Ca, Ti, P, Fe⁺², and Fe⁺³. All other components may equal zero.

3. Cr values above 500 ppm at f₀₂near QFM may cause problems in the calculation of a spinel because of the extremely high D for Cr in spinel.

4. The range of applicable compositions include alkali basalts, tholeiitic basalts, andesites, and dacites with less than 65% SiO₂. The calculations have not been tested for extremely undersaturated basalts such as olivine mellilite nephelinites.

5. Olivine and pyroxene calculations are the most stable, followed by ilmenite, plagioclase and spinel. Plagioclase calculations give unrealistically high temperatures for liquids with Al₂O₃> 20 wt %.

6. The allowable range of f₀₂is from +2 to -4 log units relative to the QFM buffer.

7. REE may be entered either as ppm or as chondrite-normalized values.

8. The periodic mixing option will be enabled if the value of PER, the % crystallization between periodic mixing events, is greater than zero. For example, if PER is 5, the system will crystallize for 5%, and then recharge and mixing of assimilant and eruption will take place. In continuous mixing, recharge, assimilation, and eruption are modeled to occur as each increment of liquidus mineral is removed.

9. The size of the crystallization increment depends on the type of system being simulated. An increment that is too large will over-fractionate the compatible elements. In addition, because crystallization is modeled by oscillating down the cotectic, a large increment will cause unstable oscillations when equilibrium crystallization is being modeled. There is no fixed method for the selection of a crystallization increment. Some testing should be made using several values. However, for simulations involving fractional crystallization of a closed system, the recommended crystallization increment is 0.2%. For recharge models, a value of 0.5 is appropriate, especially for models without Cr. For equilibrium crystallization models, a value of 0.05-0.01% will produce best results.

10. At the end of each liquidus temperature calculation section, the calculated mineral liquidus temperature is output to the screen. This is useful for the monitoring the approach of minerals to the liquidus as differentiation proceeds. In addition, it can signal the approach (or arrival) of problems with an individual mineral calculation.

11. Do not print out the output files until you have determined their length. Many runs can result in output running into the 100- to 200-page range.

12. The phase compositions, proportions, and temperature are output to file BLF.OUT or MIXNFRAC.OUT. A run title, percent fractionation, temperature and liquid composition are output to BLFPLOT or MIXPLOT. This file is tab-delimited and can be entered directly into a spreadsheet.