Sumit asked:

>P.S. I am waiting to hear Jamie's answer to your question.

Since you asked, here it is, compiled from several e-mail exchanges 
with Jamie, more information than a non-perple_x user would ever 
want, but, perhaps, useful to some of you.

When I sent Jamie my question pertaining to the differences in 
results I was seeing between perplex calculation types 5 and 7, he 
asked me to send him the in-files I'd used along with the 
"perplex_option.dat" file I'd used to run them so that he could look 
to see if there was a problem with the program, or with what I'd set 
up.  I did so, along with a summary of what results I achieved with 
each, and how they were different.  Before he looked at them he commented:

>...if you have time you could run meemum to see which starting 
>mineral compositions it gives. there are several reasons for small 
>discrepancies, among these are than in the mode 5 calculation your 
>starting point may not actually correspond to a nodal coordinate in 
>which case the program interpolates the starting compositions from 
>the adjacent nodes. another possibility is that perplex has a 
>learning algorithm which helps it adjust its configuration for a 
>specific problem, it could be that it hasn't "learned" enough on 
>your first mode 7 point to give a good result. does it matter? it's 
>easy to check, by looking at how much the subtracting an aliquot of 
>the wrong composition will effect the bulk.

After looking at my files he first said:

>I'm sorry but it's a bug or at least something the program shouldn't 
>have allowed you to do, specifically the way mode 7 calculations are 
>set up you cannot specify saturated components because the amounts 
>of these components included in the "excess" phases are not taken 
>into account (you can also see this by looking at the back 
>calculated bulk compositions output by werami). If you move SIO2 
>from the saturated components to the normal components you get 
>(within error due to the interpolation used in mode 5) the same 
>results for both calculations. so the bad news is that your 
>pseudosections are correct, but your fractionation calculations are 
>wrong (significantly so). I'll get back to you on your other 
>thoughts and queries after I figure out whether I can modify the 
>program to allow saturated components during fractionation or 
>whether I have to modify build to prevent people from using 
>saturated components in mode 7.

followed very shortly thereafter by:

>My previous answer was premature, it turns out that you can use 
>saturated components, the problem in your case was that you did not 
>constrain the amounts of the saturated component (SIO2) in your 
>fractionation calculations but you did constrain it in your phase 
>diagram section calculations. i.e., your fractionation
>input files can be made to work correctly by replacing the line:
>
>SIO2  0  0.00000      0.00000      0.00000     unconstrained amount
>
>with
>
>SIO2  1  65.0100      0.00000      0.00000     weight amount
>
>if you choose not to constrain the amount of a saturated component 
>then vertex computes the amount of the component necessary to 
>saturate the component (i.e., the condition at which the amount of 
>the "excess" phase is zero; the amounts of the phases are then 
>relative amounts (i.e., they do not include the excess phase).
>In principle leaving the amount of SIO2 unconstrained should not 
>effect the mineral compositions but it does of course mean that the 
>relative amounts of the minerals increase because there is no quartz 
>present. i think the reason it changes your mineral compositions is 
>that you somehow move into a quartz undersaturated bulk, I will have 
>to look some more to see whether this is a bug or simply a logical 
>consequence of fractionating an incompletely constrained bulk composition.
>
>would have no influence on your results though your results would 
>not take into account the amount of quartz actually present in yout 
>rockother than to change the absolute amounts of the phases

and after he played with my files a bit:

>I just checked the fractionation calculation with the unconstrained 
>saturated component and even that yields identical compositions (but 
>different proportions as it should), so it seems like everything is 
>working. that leads me to wonder how you got different Gt 
>compositions, is it possible that you are using different versions? 
>alternatively might it be that you are using an old version that had 
>a bug (there were some problems with fractionation calculations last summer)?

Which is quite possibly the case. In looking closer at my files, It 
is possible that the "first in" was run with an older version--that 
file dates to 1 Nov 2007, while the most recent of the solution 
models I've got in the working folder (solut_07) is dated 9/12/07, 
which could be why these are so different. (When I sent him the 
in-files I also sent him a summary the results I'd obtained from the 
different versions, and for this sample they predict a different 
mineral assemblage). However, I haven't made the time to check is by 
re-running them myself with the latest version yet. I did make time 
to check one of the samples for which I did have "good" results to 
confirm that while the "weight amount" version and the "unconstrained 
amount" of SiO2 do give different amounts of the minerals, they also 
gave the same P/T estimate for the garnet rims.

--Reia


Reia M. Chmielowski
PhD Candidate
Department of Earth Sciences/CODES
University of Tasmania, Australia
(03) 6239 6666
0408 238 590
http://utas.academia.edu/ReiaChmielowski