Help with REACT and kinetic minerals input

[ASchu2010]

I am working on a carbon sequestration project and have both formation water and reservoir mineralogy data to use as inputs. I have calculated the surface area and looked up rate constants to input the minerals as kinetic minerals, but when I try to get my model to run I am getting an error message that says:

 

--Can't converge, abandoning path

--Xi step is too small

 

I tried looking some on the forum to see if anyone else mentioned this Xi step but couldn't find anything. I have attached my script so you don't have to enter the data manually.

 

Any help on why this error might be occuring would be helpful!

 

Thanks!

 

-Anne

Reynolds8_Avg_KineticMins2.rea

[Tom Meuzelaar]

I am working on a carbon sequestration project and have both formation water and reservoir mineralogy data to use as inputs. I have calculated the surface area and looked up rate constants to input the minerals as kinetic minerals, but when I try to get my model to run I am getting an error message that says:

 

--Can't converge, abandoning path

--Xi step is too small

 

I tried looking some on the forum to see if anyone else mentioned this Xi step but couldn't find anything. I have attached my script so you don't have to enter the data manually.

 

Any help on why this error might be occuring would be helpful!

 

Thanks!

 

-Anne

 

Hi Anne:

 

A couple of suggestions that I think will help-

 

1. Have a look at the notes on kinetics and step size on page 33 of the GWB Reaction Modeling Guide.

2. Rather than titrating all of your minerals in at once, start with your initial speciation model, and make sure it converges. Then titrate minerals in one at a time, making sure your model is stable (start with Quartz for instance).

3. You have negative rate constants- while this is allowable in React, be sure that you've accounted for the (1-Q/K) term in the built-in rate law, which yields a negative rate when a mineral is supersaturated (precipitation) and a positive rate when a mineral is undersaturated

 

Regards,

 

Tom Meuzelaar

RockWare, Inc.

[ASchu2010]

Hi Anne:

 

A couple of suggestions that I think will help-

 

1. Have a look at the notes on kinetics and step size on page 33 of the GWB Reaction Modeling Guide.

2. Rather than titrating all of your minerals in at once, start with your initial speciation model, and make sure it converges. Then titrate minerals in one at a time, making sure your model is stable (start with Quartz for instance).

3. You have negative rate constants- while this is allowable in React, be sure that you've accounted for the (1-Q/K) term in the built-in rate law, which yields a negative rate when a mineral is supersaturated (precipitation) and a positive rate when a mineral is undersaturated

 

Regards,

 

Tom Meuzelaar

RockWare, Inc.

 

 

Hi Tom,

 

Thanks for such a prompt response!

 

1. I spent some time reading over the Kinetic Reactions section in the modeling guide today. I did notice the mention of the step size which corresponds to the "delxi" variable. I tried increasing this value by small incriments as well as by orders of magnitude to see if I could get the "Xi step size too small" error to disappear but my efforts were unsuccessful. Am I missing something?

 

2. I'm assuming by "initial speciation model" you mean the original values I used in SpecE8? If so, those values are the formation water data and converge fine in SpecE8. When I do a batch run with the water and minerals in REACT the system converges successfully as long as there are no clays present (if there are clays the system error reads: Didn't wake up, abandoning path). I tried your suggestion of titrating the minerals in one at a time but found none of them to allow the system to run.

 

3. I fixed my rate constants because, as you mentioned, they were negative and I realized I had log k rate constants. I converted them to the appropriate values which all turned out to be around +0.99. Despite this change, the system still showed the "Xi step size too small" error. With this change to non log values am I accounting for the (1-Q/K) portion of the rate constant equation that you refered to?

 

4. Additionally, I have heard from another GWB user that starting with a water/rock ratio of 1 can sometimes help, so I recalculated my rock masses to reflect this W/R=1. Even if I can get these kinetic reactions to converge and be successfull will me lowering the W/R ratio back down to more realistic reservoir ratios affect the reactions again?

 

I have attached my edited script with the new rate constants.

Reynolds8_Avg_KineticMins3.rea

[Tom Meuzelaar]

Hi Ann,

 

1. I spent some time reading over the Kinetic Reactions section in the modeling guide today. I did notice the mention of the step size which corresponds to the "delxi" variable. I tried increasing this value by small incriments as well as by orders of magnitude to see if I could get the "Xi step size too small" error to disappear but my efforts were unsuccessful. Am I missing something?

 

Changing the delxi variable won't have much effect in a kinetic run, as GWB chooses it's own timestep. The delxi variable is more suitable to equilibrium models.

 

Part of the problem is likely the fact that you've set exceedingly fast reaction rates for each of the minerals- in the Reaction Modeling Guide, p. 35, rate constants for Quartz and Albite are 2e-16 and 1e-15, respectively. Most rate constants for silicates are around this order of magnitude. In your simulation, all rate constants are close to a value of 1, reflecting very rapid dissolution/precipitation rates. This will force React to choose exceedingly small timesteps.

 

2. I'm assuming by "initial speciation model" you mean the original values I used in SpecE8? If so, those values are the formation water data and converge fine in SpecE8. When I do a batch run with the water and minerals in REACT the system converges successfully as long as there are no clays present (if there are clays the system error reads: Didn't wake up, abandoning path). I tried your suggestion of titrating the minerals in one at a time but found none of them to allow the system to run.

 

Yes, that's what I'm referring to. Clays (and the Al component) are notoriously difficult when it comes to creating equilibrium models of water-rock systems - in your model, you have 3 clays and 2 feldspars. You might try starting with one, and adding complexity as you go along.

 

I tried your suggestion of titrating the minerals in one at a time but found none of them to allow the system to run.

 

This is likely indicative that your kinetic models aren't yet properly configured to begin with. It's almost always better to start with fewer components, and fewer minerals, making sure that you have a stable model first, before adding additional complexity.

 

 

3. I fixed my rate constants because, as you mentioned, they were negative and I realized I had log k rate constants. I converted them to the appropriate values which all turned out to be around +0.99. Despite this change, the system still showed the "Xi step size too small" error. With this change to non log values am I accounting for the (1-Q/K) portion of the rate constant equation that you refered to?

 

The (1-Q/K) portion of the rate law is built in, and is simply calculated by GWB based on fluid saturation conditions- in other words, it is implicitly accounted for regardless of what values you use for rate constants and surface areas.

 

4. Additionally, I have heard from another GWB user that starting with a water/rock ratio of 1 can sometimes help, so I recalculated my rock masses to reflect this W/R=1. Even if I can get these kinetic reactions to converge and be successfull will me lowering the W/R ratio back down to more realistic reservoir ratios affect the reactions again?

 

Sometimes recreating the actual W/R is important- it really depends on what your modeling objective is. For most models I run, it's more important that I simply include enough mineral mass to reach saturation with the fluid.

 

I think if you revisit your rate constants, you'll get quite a bit further. Have a look at some of the examples in the user's guide and textbook.

 

Hope that helps,

 

Tom

[ASchu2010]

Hi Ann,

 

 

 

Changing the delxi variable won't have much effect in a kinetic run, as GWB chooses it's own timestep. The delxi variable is more suitable to equilibrium models.

 

Part of the problem is likely the fact that you've set exceedingly fast reaction rates for each of the minerals- in the Reaction Modeling Guide, p. 35, rate constants for Quartz and Albite are 2e-16 and 1e-15, respectively. Most rate constants for silicates are around this order of magnitude. In your simulation, all rate constants are close to a value of 1, reflecting very rapid dissolution/precipitation rates. This will force React to choose exceedingly small timesteps.

 

 

 

Yes, that's what I'm referring to. Clays (and the Al component) are notoriously difficult when it comes to creating equilibrium models of water-rock systems - in your model, you have 3 clays and 2 feldspars. You might try starting with one, and adding complexity as you go along.

 

 

 

This is likely indicative that your kinetic models aren't yet properly configured to begin with. It's almost always better to start with fewer components, and fewer minerals, making sure that you have a stable model first, before adding additional complexity.

 

 

 

 

The (1-Q/K) portion of the rate law is built in, and is simply calculated by GWB based on fluid saturation conditions- in other words, it is implicitly accounted for regardless of what values you use for rate constants and surface areas.

 

 

 

Sometimes recreating the actual W/R is important- it really depends on what your modeling objective is. For most models I run, it's more important that I simply include enough mineral mass to reach saturation with the fluid.

 

I think if you revisit your rate constants, you'll get quite a bit further. Have a look at some of the examples in the user's guide and textbook.

 

Hope that helps,

 

Tom

 

 

Tom-

 

Your comments have been immensely helpful! After revisiting the rate constants for slower rates and entering them, the system ran fine with all of the minerals. One of the critical questions I am trying to answer with respect to this project is how long the system takes to buffer the CO2 after injection has stopped. After running the system successfully several times and looking at plots in GtPlot I noticed that despite changing the start/end times the pH, for example, remained in the same pattern. So, running a system starting at 0 and ending at 2 days gave a pH of 5.2 at 0.5 days. Changing the start/end times to 0-30 days and looking at the pH vs. time chart again you would assume that it would still show 5.2 at 0.5 days, but instead the chart showed the pH decrease stretched out over the whole 30 days.

 

I looked in the reaction modeling guide and input the kinetic dissolution of albite example on page 36/37. I then plotted the time vs. Q/K value just like the example showed on page 38. The plot was for 1500 years and so I wondered if I shortened the time frame as I did with my own system if GWB would shrink the profile to fit or if it would show a cut off version of what was shown for the 1500 year simulation. GWB didn't shrink the profile, it showed a cut off version which is what I would like to happen with my own system. Am I typing something in wrong or maybe missing something?

 

Also, I was looking in the Short Course Notebook from Denver last year and on page 96 there is a table with rate constants. Do you know what the reference is for the values in the table?

 

I have attached my updated script which now runs successfully using slower rate constants . Thank you so much for your help.

 

-Anne

Reynolds8_Avg_KineticMins4.rea

[Tom Meuzelaar]

Hi Ann:

 

Am I typing something in wrong or maybe missing something?

 

When I run the script you give me, I get the same result for both 2 and 30 days- however, I'm not privy to the thermo database you are using, so my pH is higher than what you are calculating. If you want to email me your database (gwb@rockware.com) I can take another look.

 

 

Also, I was looking in the Short Course Notebook from Denver last year and on page 96 there is a table with rate constants. Do you know what the reference is for the values in the table?

 

Let me get back to you on this....

 

Tom

[Tom Meuzelaar]

Let me get back to you on this....

 

Ann- the rate constant data in the workbook comes from a variety of sources. However, two current and excellent compilations which contain significant rate constant data are:

 

1. Palandri, J.L. and Y.K. Kharaka, 2004, A compilation of rate parameters of water-mineral interaction kinetics for application to geochemical modeling, USGS Open File Report 2004-1068, pp. 1-70
   
2. Brantley, S.L., Kubicki, J.D. and A.T. White, 2008, Kinetics of water-rock interation, Springer, pp. 737-824
   

 

I received your database, and will try to respond within the next day or two,

 

Tom

[ASchu2010]

Ann- the rate constant data in the workbook comes from a variety of sources. However, two current and excellent compilations which contain significant rate constant data are:

 

1. Palandri, J.L. and Y.K. Kharaka, 2004, A compilation of rate parameters of water-mineral interaction kinetics for application to geochemical modeling, USGS Open File Report 2004-1068, pp. 1-70
   
2. Brantley, S.L., Kubicki, J.D. and A.T. White, 2008, Kinetics of water-rock interation, Springer, pp. 737-824
   

 

I received your database, and will try to respond within the next day or two,

 

Tom

 

Tom,

 

Great, yes I am familiar with those two sources and they have indeed proved quite helpful! Also, thank you for looking at my script again. The system is running great now even with all of the minerals and water data included but I won't be able to make any quantitative results statements without solving this time issue and appreciate very much your expertise and suggestions.

 

Thanks!!

 

Anne

[Tom Meuzelaar]

Tom,

 

Great, yes I am familiar with those two sources and they have indeed proved quite helpful! Also, thank you for looking at my script again. The system is running great now even with all of the minerals and water data included but I won't be able to make any quantitative results statements without solving this time issue and appreciate very much your expertise and suggestions.

 

Thanks!!

 

Anne

 

Anne:

 

Just a quick note here to let our users know that we continued our discussions via personal correspondence, and that your questions did not go unanswered.

 

Regards,

 

Tom