Dear colleagues,
This Spring semester, I had
the opportunity to teach an elective course called "Calculation of Material
and Molecular Properties" to a few undergraduate and graduate students.
Most of the students coming into the course had little or no background in
computational chemistry. The course was focused on using quantum chemistry (DFT
mainly) to compute optimized geometries, infrared vibrational spectra,
thermodynamic properties (enthalpies, free energies, entropies, zero-point
energies, and heat capacities), heats of reaction, transition states, and
various atomistic descriptors (bond orders, net atomic charges, atomic spin
moments, etc.).
For the course, we primarily used
two software packages: Gaussian 09/16 (for the quantum chemistry calculations)
and Chargemol (to compute the atomistic descriptors, using the
Gaussian-generated wfx file as input).
The beginning of
the course included an introduction to the different types of computational
chemistry calculations. I explained why computational chemistry is so widely
used across various scientific disciplines including materials science,
chemistry, engineering, biochemistry, physics, etc. As examples of the
widespread importance, I used the Nature 2014 article on the top 100 cited
science articles (in which DFT articles were one of the categories mentioned)
and the 1998 Nobel prize to Kohn and Pople for their work in computational
chemistry. The difference between classical atomistic (e.g., molecular dynamics
and Monte Carlo) calculations versus quantum chemistry calculations (e.g., DFT)
was explained. A brief background into Density Functional Theory was given, with
students reading and writing brief (1-2 page) highlights of a few introductory
articles on DFT (e.g., the DFT "Jacob's ladder" article by Perdew
et al.). After a few weeks of introductory material, including a discussion of
different classes of DFT functionals, the students were taught how to carry out
basic calculation types (e.g., ground state optimization, transition state
optimization, frequency calculations, constrained geometry optimizations) in
Gaussian software. The students were also taught the mechanics of computing the
atomistic descriptors using Chargemol software. The students completed a couple
of homework assignments related to these topics, in which they performed
calculations on different materials.
After this, the
students selected a course project. The students could choose a topic of their
choice, although I offered some guidance on which kinds of topics might be more
suitable for the length of time available. The topics spanned a diverse range
including hypercoordinate molecules, dinuclear transition organometallic
compounds, solid state crystalline materials, nuclear chemistry, boron and
carbon containing molecules that exhibit unusual bonding, force field
parameterization, thermodynamics of explosive materials, etc. All of the class
projects were focused on computing things that were new, that no one had
computed the answers to before. Some of the class projects used Gaussian
software, but others involved other software (RASPA, VASP, Matlab, etc.) Most of
them used Chargemol to compute the atomistic
descriptors.
We had a lot of fun exploring some unusual
bonding properties of materials. In one of the class periods I posed an unsolved
problem concerning the structure of diazomethane. The goal was to calculate the
sum of bond orders for each atom, and to use this information to determine
whether the central nitrogen atom is hypercoordinate or not. What made this so
fun was that I had never calculated it before, so I learned the answer together
with the students. Many of these classes were held a computer lab where each
student had individual desktop computers, and the instructor had a computer that
could be projected onto the main screen in front of the
class.
The students turned in three brief reports
of their project: an early one describing the project to be studied which also
included a brief overview of related literature, a middle one describing initial
calculation results and any difficulties they encountered (some students had to
change project at this point if their initial idea was not working), and a final
brief written report. The students also gave a 10 minute oral powerpoint
presentation at the end of their project.
I believe the
students learned a lot. Some of the students commented in the course evaluations
that it was one of the most useful elective courses they took. Things didn't
go perfectly during the course, but overall the students learned some valuable
new skills. We didn't go into great depth of the theory, but the students
learned enough to be able to carry out meaningful calculations. On
the course homepage, I posted journal articles describing the theoretical
methods in more depth in case some of the students wanted to explore the
details, but I didn't require the students to read these. They were required
to read some introductory journal articles.
I believe
the ability to compute meaningful and reliable bond orders really enriched the
course and made for some fun projects. Although bond order is a widely used
concept, only recently has a reliable way to compute it been developed that
works across an extremely wide range of material types (
T.
A. Manz, “Introducing DDEC6 atomic population analysis: part
3. Comprehensive method to compute bond orders
,” RSC
Advances,
7 (2017) 45552-45581 (open access
)
http://doi.org/10.1039/c7ra07400j .
). It's great for a student who is just learning computational chemistry to
be able to take a molecule or system that no one knows the bond orders for and
be able within a few days to calculate a meaningful answer for the first time.
It really empowers them and makes them feel that they can do research that is
meaningful and cutting edge. There are millions of interesting and novel
materials for which the bond orders are unknown, and students can really get in
on the ground floor. It's quite an opportunity.
If
any of you are teaching a computational chemistry course during the next year,
you may want to incorporate some of these ideas into your curriculum. If any of
you have recently taught a computational chemistry course, feel free to mention
your experiences regarding what worked well and what
didn't.
Sincerely,