database choice for AMD waters

[mgsephton]

Hi

I used phreeqc.dat to calculate charge balances for some acid mine drainage samples with high levels of sulfate and iron (from 4000-30000ppm SO4 & 1000-10000ppm Fe) using phreeqc.dat as the database.  The percent error tended to be positive 10-20 for many cases.  I repeated the calculations using the llnl.dat database instead, and the percent errors were much lower, averaging around 2%.

The speciation results using the phreeqc.dat database tended to have higher molarities of FeSO4+,  lower molarities of SO4-2,  Fe+2, Fe+3 and FeSO4 compared to the speciation results using the llnl database.

Comparing the two databases, i noticed the log_k values for Fe3+ + SO4 -2 -> FeSO4 + are quite different between the two databases, 1.92 for llnl database, and 4.0 for phreeqc database.  This seems to explain part of the difference in the results, i haven't discovered yet what explains the rest of the difference.

Is there a listing of sources for the log_k values used in each of the databases somewhere? 
I wonder if anyone knows why the values for log_k for FeSO4+ are so different between the llnl and the phreeqc databases.

Would it be easy to enhance IPhreeqc so that one could write a script that loads two databases, and then automatically find major discrepancies between them?

Thanks

Michael

[dlparkhurst]

I don't think the extent of ion pairing alone should cause a difference in the charge balance between the two data bases. If you have Na+, Cl-, and a hypothesized NaCl(aq). The charge sum will be the same regardless of the amount of NaCl(aq) that forms in the speciation calculation.

First check is whether the total concentrations of each element (listed directly under "Solution composition") is the same for each database, and second, the distribution among the redox valences in "Distribution of species" (Fe(2) and Fe(3) for example). My guess is that the two databases are distributing the redox states differently.

[mgsephton]

The total concentrations of each element are the same.  The distribution of species is different: eg.
with llnl.dat database I get
4.527e-003 Fe(2) and 1.067e-002 Fe(3) and a percent error of -1.73
whereas with phreeqc.dat i get
1.028e-003 Fe(2) and 1.416e-002 Fe(3) and a percent error of 12.25


Why do the two databases distribute the redox states differently?

[dlparkhurst]

Sorry, I have to revise my earlier statements. If you have measured Fe(II) and Fe(III), and enter them in SOLUTION as Fe(2) and Fe(3), then the log Ks will not make a difference in the charge balance.

However, log Ks for ion pairing will make a difference in charge balance when it comes to distributing a redox element among its redox states. For iron, it will make a difference if more Fe ends up as Fe(II) vs Fe(III) with one database versus the other because you will have more Fe+2 species and fewer Fe+3 species, for a net of less positively charged ions.

Both databases have the same log K for Fe+2 = Fe+3 + e-, when converted to a common equation. However, Fe+3 bare ion is always a small fraction of the total Fe(III) species. So, the difference is that phreeqc.dat must have stronger Ks for Fe(III) pairs and complexes. You were correct in the first place when you noted FeSO4+ is stronger in phreeqc.dat, which distributes the iron more in the Fe(III) species, and greater positive charge.

You could also go to the literature and try to figure out the sources for the two log Ks and try to make an assessment.

Which is right? Perhaps the charge balance indicates that llnl.dat is better. However, the charge balance will also be very dependent on how you specified pe or -redox in the SOLUTION definition. Usually, there is a lot of uncertainty in the definition of pe. If a redox couple is used to estimate pe, then again there are large uncertainties either in the measurements of the redox states (often one state is small and derived by difference) or in the assumption that the redox conditions for one couple apply to another element (Fe in this case). So another option is to adjust the pe until you get charge balance.

Unfortunately, there will always be a lot of uncertainty. Ideally, you get good measurements of Fe(II) and Fe(III). Lacking that you must make some choices on how to determine the distribution of Fe(II) and Fe(III). At neutral pH, Fe(II) predominates, but in acid conditions both Fe(II) and Fe(III) are stable.