Maggie
- Jul 16, 2019
- 3 min read

My project has changed its scope

A lot has happened since the last project update. We went to Beijing and some of us went to Xi'an. The time between then and now as allowed my advisors and I to reconsider the scope of the project. Initially, I thought I would be testing one functional (MN15-L) and given its outstanding results for simple molecules and atoms, we thought it would yield the same for ferroelectric materials. However, as I mentioned in my last project update, the MN15-L functional appears to better estimate only the band gap and is about the same at estimating polarization, energy change, and the cubic lattice constant. Since these types of results aren't indicative of an impactful paper, we decided to begin testing other new functionals on ferroelectrics and for which I haven't found previous publications with respect to these particular ferroelectric materials.

So far, I've finished running MN15L on PTO (see previous research update post for details pertaining to that). This functional provides the closest estimate of the band gap amongst other tested functionals to the experimental results. It also provides a good polarization estimation. Again, it estimates about the same energy change and cubic lattice constant as PBE and LDA.

I've also run revM06-L, GAM, HLE17, and MN12-L on PTO. For BTO, I've run GAM, HLE17, and MN12-L. I ran into some trouble with the revM06-L for BTO and I've decided to come back to it once Dr. He responds to my email pertaining to it. For PTO, revM06-L is roughly the same in terms of polarization and change in energy as MN15L, but it worse at estimating the band gap.

GAM yields a positive Fermi energy for both BTO and PTO which indicates instability of the structure. This can't be correct since all other functionals predict negative fermi energy. I plan to look into this functional more to find the issue here. Interestingly, the change in energy is about the same as PBE and LDA. In terms of polarization, it's about the same as MN15-L and revM06-L for both materials. Finally, it estimates the band gap poorly (about the same as PBE's predictions).

For PTO, HLE17 predicts a midrange accuracy of polarization and band gap, but underestimates the energy change (with respect to other tested functionals) by 2 eV. For BTO, it predicts roughly the same band gap as PBE. Like before, it underestimates the energy change by 2 eV and predicts an a midrange accuracy of the polarization (again, when these results are compared to the other functionals). Basically, HLE17 is bad at predicting the properties of these ferroelectrics across the board.

Finally, MN12-L does very well at predicting change in energy. For BTO, it is only 3 eV less from experiment which is impressive consider all other tested functionals so far predict around 10 eV less than experiment. It is the second best at predicting the band gap (MN15-L is better) and predicts the lowest polarization amongst the other functionals which is good since every other functional greatly overestimates polarization with respect to experiment. For PTO, MN12-L estimates a change in energy only 23 eV less than experiment where as other functionals predict 56 eV less so that's a massive improvement. Predicted polarization is within experimental range and it's the best prediction of band gap so far. This functional is promising due to its good results for these two materials and that its fairly computationally cheap.

Moving forward for the rest of my time here, I plan to run N12, N12-SX, SOGGA, SOGGA11, and SOGGA11-X. If time permits, I'll run DFT+U, DFT+D, BEEF, mBEEF, MS0, MS1, and MS2. Again, this will be for both BTO and PTO. I've currently submitted N12 calculations for both PTO and BTO, but they're waiting in the queue.

In the mean time (and now that I've bored you with a highly technical summary of my work over the past week and a half), enjoy these few pictures from Xi'an!