Aloooooone

This is my second day without my graduate student. So far, this week is going pretty poorly.

I spent all of Friday and some of Monday analyzing my CaTiO3 results. When I moved on to BaTiO3 and SrTiO3 on Monday, several structures failed much in the same way my CaTiO3 structures were failing before. But I fixed that problem in calcium titanate. Shouldn't the solution be transferable to other compounds? There was no reason I could see that it shouldn't. Titanium is still the dominant cation in each compound. The Ti-O potential parameters should work. (I did try the Ba-O and Sr-O parameters just for kicks. They predictably failed.) It took me an entire day to figure out what the problem might be with my BaTiO3 calculations. Hours of poring over papers I don't understand (and still don't).

Once I had a theory as to why they might be failing, I tried to figure out how I would fix it. The material became highly technical very quickly. I do not have the knowledge or expertise to understand how to derive these potentials--that's the stuff that PhD theses are made of. So I decided to email my professor this morning to ask to meet with him. His response: "Oh, crap I never mentioned to you .... I left for 3 weeks. I won't be back until July 19."

He asked me about the problem, in case there was a grad student or postdoc he could point me to, and I summed it up for him. He said no one in town would know enough about interatomic potentials to help, but gave me the email for one of his former postdocs who is now an assistant professor at University of Texas, Austin.

Here is that email, with the subject, "Hello! I'm Prof. Fennie's REU student." I hope she responds.

Hi! So I think Craig cc'd you on my last email to him, but I started running into some problems the day after he left for a trip, and while my graduate student (Eva Smith) is at a conference for the week. 

My project is that I'm using GULP to optimize structures using buckingham potentials (and spring potentials, for the core-shell model) and comparing the optimized structures from this method to those calculated using DFT methods.

I've been using potentials from Lewis and Catlow (http://iopscience.iop.org/0022-3719/18/6/010) and started with CaTiO3. I originally had some problems with an error message reading "Largest core-shell distance exceeds cutoff of cuts" on the structures P4mm, Amm2, R3m, and Pnma. Once I found out that the spring potential and shell charge of the oxygen ion depend on the cation (from the paper above), I changed the O parameters to those corresponding with Ti instead of Ca and I no longer got that error message. 

Now, I'm trying to apply the same method to BaTiO3 and SrTiO3. I'm still using the O parameters corresponding to Ti, but now P4mm is failing (with the same original "Largest core-shell distance" error message) in STO, and P4mm and R3m are both failing in BTO. (They also fail when I use the O parameters corresponding to Ba.) 

I thought the Ti-O potential parameters would be transferable between compounds, but this is apparently not the case. My hypothesis is that I'm using these potentials incorrectly. They were originally calculated for binary oxides, and I'm using them for ternary oxides. But I don't have enough technical knowledge of the material (I just finished my second year of undergrad) to know if my hypothesis is correct, and if it is, how to derive the ternary potentials from the binary potentials. The literature on this is extremely vague and unclear, especially for an amateur like me. 

If you have the time, and if you have any insight into this situation, I would be incredibly grateful. With Eva, Brian Abett (my other guiding graduate student), AND Craig all gone, and none of the other students in the group familiar with interatomic potentials, I'm feeling a little lost and adrift.

I did a lot of research into the error message itself, but I truly think that it's a symptom of something else being in error. With CaTiO3, it was the potential parameters, which is what makes me think I've been using them incorrectly, possibly by not accounting for the fact that they're ternary oxides. Does this seem likely to you? Do you know anything about potentials in ternary oxides?
Thank you so much. I've heard your name many times; they miss you back here. :)

Best wishes,
Meredith Staub

I want to solve this problem. I want Eva and Craig to come back, and I can show them how I fit all the pieces back together, how hard I worked to use concepts I barely understand to fix a problem I could barely diagnose. I just don't have the technical knowledge to do it by myself.