[OLD] Thermodynamic database discussion

[webmaster]

From: Paul Barton

Subject: thermodynamic data bases

There is a substantial level of discontent with the "data base", and (at last) apparent support to do something to improve it. Great! But a word of caution lest we damage the credibility that we all share. The recent flurry identifying specific defects is useful and, of course, needed; but it may mask a larger problem: band aids can improve appearances and functions, but what we really need is a thorough overhaul. In a recent e-mail Dave Wesolowski appropriately described the data base as hanging by a few well defined anchors. That is a good place from which to start, but let us be thoughtful about the way we proceed; that is one reason why Alex's idea of a workshop is appealing. I am reminded of John Haas' examination of the literature on the thermodynamics of the iron oxides maybe 20 years ago. He found that the largest discrepancies among investigations lay not in the measurements themselves, but in the fundamental data used for the calculations, such as those for the H2O decomposition, and the CO-CO2 reaction. We all can find errors, but none of us really knows what the truth is; hopefully, we approach it (probably asymptotically). Every new experimental study may have ripples that infermodification of prior "facts". Therefore, we need a broadly-supported mechanism that continually re-evaluates where we stand and permits steady upgrading of the whole edifice, and the product needs to be transparent so that the chain of assumptions and input data is traceable without major research effort. Today we can only trace back to the next link of what might turn out to be a long and extensively branched chain. Now it is someone else's turn on the soapbox.

 

From: Gregory Miller

Subject: RE: thermodynamic data bases

Paul Barton's comment that "Today we can only trace back to the next link of what might turn out to be a long and extensively branched chain." May well identify the core issue, and possibly a logical structure to a solution and development of the "FAQ". I envision a database that contains the citations for thermo property sources, and (this is the more difficult part) the sources for the 'lower level' reaction constants used by the citation author. In this manner it may be possible to identify the source of database issues; in some cases problems may have a common root! The database could be annotated by users in an informal manner, and users can access the comments. We could start with the Fe.OH and Ce comments we have. This is an expansion of Andy Wilde's and Mark Logsdon's ideas in that not only can a data base be searched for constants for a compound, but queried for data restricted to certain authors or programs. Practitioners can

describe applications and successes and failures of thermo data, other can point to their published work. The need is recognized, there are some approaches in IT that are suited to this, and it seems that the folks at CRC are ready to get started. How should we start?

 

From: Mark J. Logsdon

Subject: Re: thermodynamic data bases

Seems to me that there are two levels of response needed:

1. GWB-related. The FAQ and archive that may be appropriate for the user-group is not an undertaking at the same level as the much more deep-seated issues of thermodynamics that are being raised. We have some needs in the short term for information among GWB users that can be managed without Craig having to re-answer matters that are already on record somewhere. I had imagined the database questions to be some kind of subset of that, to which users could turn for quick information on database (for this code) issues and leads on how to proceed to the extent that those are known to the user group.

2. "Hard core" Thermo issues. The issues raised by Paul, Greg, Andy, Dave and others are much more fundamental to both "pure" and "applied" geochemistry and extend far outside the GWB users, or even geochemical modeling per se. The resolution of the questions will require major resources of time and effort (i.e., $), and these calls on resources cannot and should not be seen as GWB's responsibility. [Note that the databases in question in our user-group discussions are public-domain products that have been developed by DOE and USGS. Please see discussions in the user's manual about these databases and how and why they are included.]

It would be grand - and a goal in the (Robert) Brownian sense of "reach should exceed one's grasp"- to have a "final" database that has some or all of the characteristics described by Greg. Paul's message of today refers to some incipient efforts to jump-start efforts for the basic science and documentation, efforts that, if they are successful, will require support from DOE, NSF and other funding agencies. But those efforts imply a much wider and more time-consuming exercise than the GWB users can mount on their own.

 

From: Dave Decker

Subject: Re: thermodynamic data bases

The recent discussion about thermodynamic database values has inspired me to write my first post to the GWB group. I have found it interesting (and heartening) that somebody is wondering about the origins of the data posted in thermodynamic databases, and perturbing that the response to this query has not been particularly vigorous. As somebody who has yet to climb very far up on the GWB learning curve, but who has occasionally used the 'other' geochemical models out there, and who has written a little bit of code to deal with arsenic, I would like to throw in my two cents (though I suspect that some of you will question such a high valuation on these comments! ).

1. When assembling a model, one should strive to use thermodynamic values that are consistent with one or two sources, i.e. use values from one research lab - almost guarantees consistent laboratory methodology (if that methodology is appropriate for the modeling problem at hand) from one mineral to the next. Unfortunately, one can't always do this, which leads to comment (2).

2. Clearly, the results of geochemical model simulations (using GWB or other codes) are affected by the range of thermo values in the published literature (i.e. database). More to the point, the outcome of the modeling exercise is dependent upon the modeler's selection of thermo values. Other than comment (1) above, how else are people selecting thermodynamic values for a suite of minerals from a variety of sources that are defensible and consistent?

 

From: Geoffrey Thyne

Subject: Re: GWB thermo database questions

Dave inspired me to all add my first comment to this group. After a few years of playing with various models, I have found that adjusting the thermo constants to better reflect observations of complex natural systems seems to facilitate using models to represent such systems. And while it is true that changes in thermo data can significantly affect the results for simple systems, usually models of more complex (natural) systems are much less sensitive than you might expect. Finally, as a lab person I can attest that whatever experiment you perform and constants you measure, they do not accurately reflect the natural system.

So the question for the group that I would like to hear about is: why shouldn't we modify the lab-derived thermo values (within reason) to make our models more accurately reflect the system we are trying to model?

 

From: D J Wesolowski

Subject: Re: GWB thermo database questions

I am also inspired for the first time. If you substituted "preconceived notion" for "natural system" in the paragraphs below, then Dr. Thynes suggestion might have some merit. If the model doesn't fit the field observations, it may not be the input thermodynamic data that are at fault, but rather the method attempted to model the more complex natural system. Another way of saying this is that "chaos" is just a mathematician's way of describing a lack of sufficient constraints.

 

From: Ron Schmiermund

Subject: Re: GWB thermo database questions

I am yet another closet GWB user inspired by the ongoing exchange to make a first posting to the user's group. Each of the recent comments in the thermodynamic data exchange make valid points. There is clearly a lot a of frustration/concern about the very bases of our calculations and a recognition that we can and must do the very best job possible. To that end, I believe that the suggestions of Andy Wilde and Robert Lee should be given much consideration. Goef Thyne raises a good point arguing for the application of appropriate flexibility/judgement in thermodynamic "constants" while Dave Decker argues for intelligent inspection and honoring the best available and most credible and consistent data. Altering values carries with it the burden of justifying those changes each time the adjusted "constants" are used, and being suspected of fudging to obtain preferred results, not to mention potentially causing chaos in the literature. On the other hand, if one were to simply stick by the published values without question there is potential for not recognizing a legitimate data issue and becoming overly confident in a model result because only the best data was used. A transparent data forum might help resolve these issues. The idea of a "universal" database is very appealing and I would like to suggest that it have one additional attribute. Specifically, a linkage from a given mineral or aqueous-species record set to a bulletin board or text file where observations, experience, suggestions or recommended modifications could be posted.

 

From: Mark J. Logsdon

Subject: Re: GWB thermo database questions

In order to understand Dr. Thyne's suggestion, it seems that one would need to know what the purpose of the modeling under consideration might be. And, then, as with any question, one also would need to understand what counts as an answer. If the ability to measure the "natural system" is better than the ability to measure the thermodynamic behavior of individual minerals, then what's left to model? Or, why would one not use an entirely empirical model rather than invoke any theory at all (for example if one were exercising an initial speciation model)? It seems that the solution proposed by Dr.Thyne would arise only in the instance of an inverse model. But - even putting aside the issue of the thermodynamic data - pretty much everyone agrees that inverse models will be non-unique. Therefore, what would be the advantage of adding subjective degrees of freedom in the thermodynamics. I take this to be related to Dr. Wesolowski's response. For forward modeling, I can see no way in which this could be implemented and lots of pitfalls (of the sort described by Ron Schmiermund) if one were to attempt to do. A forward-modeling exercise (imagine: the HLW program) done with subjectively determined thermodynamic data would never survive a "Daubert" review on at least two grounds (lack of peer review and inability to describe error rates). To be fair, to the extent that the thermodynamic data are not well posed, it is possible that these same objections could be raised with respect to any modeling that used such data. Finally, in the natural sciences it must be the case (except for nearly trivial examples) that a model is a simplification compared to the natural world, whether the model is an experiment or a computation. Surely, that alone (i.e., simplicity) cannot count against using models to explore well-posed questions.

 

From: D J Wesolowski

Subject: Re: GWB thermo database questions

OK, Here's another go at the issue. I think experimental studies, thermodynamic/kinetic theory and correlation algorithms should be thought of as climber's pitons. Some you can and MUST stake your life on, others can only be considered as tentatively anchored, and clearly some that we think are secure are in fact just about to pull out. The problem in constructing data bases for field applications is that it is difficult to know which ones are reliable. This is a problem that is continuously addressed (examples CODATA, NIST, the MULTEQ program used in the power industry, SUPCRT, etc., etc.) Those close to the experiment or the theory may have a better feel for the reliability of individual anchor points, but they may be so "close to the wall" that they put in too many pitons and never make it to the top. On the other hand, there are many "free climbers" in the game of modeling natural systems who may not even know they've already come off the face and are falling. These folks should think twice about unclipping from their pitons to get to the top faster. This is a problem that is never going to be adequately resolved. In my opinion, at the current time, the semitheoretical, but largely empirical correlation approach being employed by Helgeson, Shock, Sverjensky and co-workers provides the most comprehensive and useful interpolative and extrapolative model for the relevant thermodynamic properties of water/rock interaction systems, and these fellows are very good at evaluating and incorporating the best experimental data and new theoretical developments. However, there remain problems. I offer as an example our recent studies of the first hydrolysis constant of Zn2+ (Benezeth et al. GCA, 1999, v63, pp1571-1586). The constants extracted from the SUPCRT database were found to be orders of magnitude off. In fact, our new results actually improve the positioning of zinc within the correlation framework of this model. So, there is still plenty of room to hammer in more solid anchors, and plenty for the synthesizers to do to keep up with the experiment and theory. Sooner or later the computational folks may put us out of business, but probably not right away. I envy the climbers, and wish them well. Hope they hook on to the right pins! In the meantime, it is very important for all of us to have respect for what each type of geochemist adds to the pot, and we should present a united, enthusiastic and aggressive face to our funders. Confusion and mistrust lead to debacles like Yucca Mountain, which has been under evaluation as a nuclear waste repository for more years than I can even remember. This is a critical societal issue and one that we as geochemists are uniquely equipped to address. I think we need to take a longer view on what is important for individuals to achieve in a given year or even over a lifetime, within the context of why we are doing what we are doing. We should have greater praise for small steps that are solidly placed.