alkalinity of seawater is too high for phreeqc engines

[jjstickel]

I think there is a bug with the alkalinity calculation in pyEQL.Solution. Example:

In [1]: import pyEQL as pyeql

In [2]: sol = pyeql.Solution.from_preset("seawater", engine="phreeqc2026", balance_charge="pH")
/Users/jonathan/Library/Python/3.12/lib/python/site-packages/pyEQL/solution.py:339: UserWarning: After initialization, the calculated solution pH of 8.220 does not match the specified pH of 8.099. This might be a result of erroneous input (e.g., specifying both pH and H+), or it can happen during from_dict / from_preset if you use a different engine than the one which generated the original dict.
  warnings.warn(

In [3]: sol.alkalinity # this looks OK
Out[3]: <Quantity(117.4375715148465, 'milligram / liter')>

In [4]: sol.equilibrate(atmosphere=True)

In [5]: sol.alkalinity # too high!
Out[5]: <Quantity(753.3507804556849, 'milligram / liter')>

Otherwise, the solution and its properties seem reasonable, e.g., pH = 8.2.

As a workaround, I'd like to retrieve the alkalinity calculated by the phreeqc2026 (or phreeqc) engine, but I could not find a way to do that. I do see some discussion of retrieving alkalinity and other properties in issue #310, but it is not clear what are the low-level calls to do this.

[rkingsbury]

Hmmm, thanks for flagging this @jjstickel . @SuixiongTay , can you investigate when you have a few minutes? I wonder if this is related to #398 .

@jjstickel , the underlying PHREEQC object is query-able (although it's generally intended for use only by developers), but I believe the correct code should be something like

sol.engine.ppsol.alkalinity

The engine.ppsol refers to the in-memory PHREEQC solution, and it's the entity that the pyEQL.Solution pulls properties from.

[jjstickel]

sol.engine.ppsol.alkalinity
Unfortunately, this attribute is not available in the current version of pyEQL. I submitted a PR #459 to bring in alkalinity from the phreeqc2026 engine.

[rkingsbury]

Digging further here, it's interesting that the discrepancy is not nearly as large with the native engine (default), which uses llnl.dat rather than phreeqc.dat. This results in a little bit of MgCl[+1] and CaCl[+1] that you don't get from the phreeqc2026 engine.

>>> s=Solution.from_preset('seawater', engine='native', balance_charge='pH')
>>> s.alkalinity
<Quantity(117.4375715148465, 'milligram / liter')>
>>> s.equilibrate()
>>> s.alkalinity
<Quantity(138.58643312800692, 'milligram / liter')>
>>> from pprint import pprint
>>> pprint(s.cations, sort_dicts=False)
{'Na[+1]': 0.4433714025177017,
 'Mg[+2]': 0.03989283543128503,
 'K[+1]': 0.00993963969389008,
 'Ca[+2]': 0.00912890816370233,
 'MgCl[+1]': 0.0048322059815775215,
 'CaCl[+1]': 0.0002349720175303199,
 'MgHCO3[+1]': 0.00017589217605201544,
 'Sr[+2]': 7.81912802211947e-05,
 'MgB(OH)4[+1]': 5.3366260877350515e-05,
 'CaHCO3[+1]': 3.2143754439079596e-05,
 'MgF[+1]': 1.5231616415800963e-05,
 'CaB(OH)4[+1]': 8.033054971734781e-06,
 'SrCl[+1]': 4.78669172386515e-06,
 'CaF[+1]': 5.74606853281073e-07,
 'CaOH[+1]': 1.0658588072772545e-07,
 'H[+1]': 5.442064470280126e-09,
 'SrF[+1]': 1.2010306904232672e-09,
 'SrOH[+1]': 2.8133147113171034e-10,
 'MgOH[+1]': 4.963854324274498e-12}
>>> pprint(s.anions, sort_dicts=False)
{'Cl[-1]': 0.5217477571135759,
 'SO4[-2]': 0.013070267924265954,
 'NaSO4[-1]': 0.006580128353233588,
 'HCO3[-1]': 0.0011945573924713067,
 'Br[-1]': 0.0008321477568366732,
 'KSO4[-1]': 0.0001686344960494305,
 'BO2[-1]': 5.204325165701761e-05,
 'F[-1]': 5.1346905890040505e-05,
 'CO3[-2]': 4.25613922170462e-05,
 'NaCO3[-1]': 1.265396061973774e-05,
 'OH[-1]': 3.302315125178726e-06,
 'HSO4[-1]': 1.4741764891422578e-09,
 'BH2(OF)2[-1]': 8.697358994517376e-15,
 'HF2[-1]': 2.7223126226384054e-15,
 'B2H5O6[-1]': 5.536726401768357e-18,
 'BOHF3[-1]': 5.557698941026352e-21,
 'BrO[-1]': 4.758462532332134e-24,
 'ClO[-1]': 1.9451356506408476e-25,
 'BF4[-1]': 5.518561133906346e-29,
 'Br3[-1]': 2.73868612789012e-29}

Comparing phreeqc2026 engine (so, identical speciation engine, but different activity coefficient methods):

>>> s=Solution.from_preset('seawater', engine='phreeqc2026', balance_charge='pH')
>>> s.alkalinity
<Quantity(117.4375715148465, 'milligram / liter')>
>>> s.equilibrate()
>>> s.alkalinity
<Quantity(750.1329716570397, 'milligram / liter')>
>>> pprint(s.cations, sort_dicts=False)
{'Na[+1]': 0.454076694579413,
 'Mg[+2]': 0.047568706901239065,
 'K[+1]': 0.009994141997974333,
 'Ca[+2]': 0.00956909884097742,
 'MgHCO3[+1]': 0.0002241702871149978,
 'Sr[+2]': 8.39764935415523e-05,
 'MgF[+1]': 2.990307255052956e-05,
 'MgOH[+1]': 2.002642507765606e-05,
 'CaHCO3[+1]': 1.1767367923649085e-05,
 'SrHCO3[+1]': 4.1161804789948005e-07,
 'CaOH[+1]': 1.519796167500615e-07,
 'H[+1]': 4.562072948936468e-09,
 'SrOH[+1]': 4.5220049834386825e-10,
 'CaHSO4[+1]': 2.231148160876166e-11}
>>> pprint(s.anions, sort_dicts=False)
{'Cl[-1]': 0.5442578559819381,
 'NaSO4[-1]': 0.013317073194106448,
 'SO4[-2]': 0.00921392866557388,
 'HCO3[-1]': 0.0012967454181522632,
 'Br[-1]': 0.0008395244921424564,
 'KSO4[-1]': 0.00018089820687298627,
 'Mg(SO4)2[-2]': 0.00012150409668995436,
 'BH2O3[-1]': 8.52909405909086e-05,
 'CO3[-2]': 5.651972067316964e-05,
 'F[-1]': 3.4861809834616796e-05,
 'OH[-1]': 4.67668745098709e-06,
 'BH3O3F[-1]': 4.368815670373759e-09,
 'HSO4[-1]': 7.702228593072517e-10,
 'BH2(OF)2[-1]': 3.532024945674748e-14,
 'HF2[-1]': 1.221783332859699e-14,
 'BOHF3[-1]': 2.922024164532674e-21,
 'BF4[-1]': 8.981395973376415e-28}

Note that I get simliar (but not identical) results if I disable the Stumm-Morgan "conservative ions" definition and rely solely on the weak acid/base definition.

TL;DR I believe this is a weakness in our implementation of get_alkalinity (and, arguably, the S-M definition) related to how to we handle complexes. Things like MgCl[+1] and CaCl[+1] don't get counted, and since the S-M definition sums over all the free cations and anions, the resulting alkalinity for speciated solutions is a bit of a wildcard - it depends on what fraction of the total cations vs. the total anions are complexed. It might be possible to work around this using total concentrations but I'd need to think about it more.