Do’s and Don’ts

This section collects various recommendations and cautions for the use of MC_fit.

Do’s

  • Use the latest version of Perple_X and its data files. MC_fit is still in development and evolves rapidly.

  • Do an example before trying your own problem.

  • Use raw analytical data, it need not be normalized. A bulk composition calculated from mineral compositions is not raw analytical data; the raw analytical data are the mineral modes and compositions (Appendix B).

  • Set the molar_composition_input and relative_error options to be consistent with the analytical data. These settings are global, all compositions and uncertainties must be in either molar or mass units and all uncertainties must be relative or absolute.

  • Specify reasonable ranges for inversion parameters in the .imc file (my_project.imc). Excessive ranges reduce the probability of finding the best central model. Restrictive ranges may, paradoxically, slow computations and exclude the best central model. To view the scatter of the central-model results in the parameter space do an initial test with, say, 50 central-model tries (number_of_tries) and no perturbations (number_of_perturbations set to 0), then adjust the parameter ranges accordingly.

  • Include solution models for all phases that might plausibly limit the stability of the observed phases in the problem definition file (my_project.dat).

Don’ts

  • Don’t guestimate unmeasured components such as O2, H2O, CO2, or S. The guestimation scheme may assume site populations that are different from the corresponding solution model used in the inversion and this will create conflicts. MC_fit fits analytical data to the site populations implicit in the specified solution models. If you don’t like those site populations, change solution models.

  • Don’t include components in phase compositions that cannot be predicted by the solution model chosen for that phase. For example, don’t include MnO in an amphibole composition if the amphibole solution model cannot accommodate MnO. The MnO from the observed amphibole composition would need to be accommodated by some other phase that can incorporate MnO and make the composition of that phase excessively Mn-rich. Unpredictable components should be either simply eliminated or replaced by geochemical proxies (e.g., the amount of MnO can be added to FeO).

  • Don’t include a fluid phase as an observed phase unless there is strong evidence for its presence and composition. Instead include a generic fluid solution model ( e.g., COH-fluid) in the problem definition file (my_project.dat). Typically inversion will predict fluid as an extraneous phase if the observed assemblage was truly fluid-saturated.

  • Don’t impose fluid saturated component or saturated phase component constraints. See the previous “Don’t”.

  • Don’t specify components in the problem definition file (my_project.dat) that cannot be accommodated by the observed phases. By definition, these cannot be predicted and will needlessly complicate the inversion.

  • Don’t use bulk compositional data together with modal data unless the two data sets are independently measured (Appendix B).