Research Interests

Solids under high pressure

Under pressure, our chemical intuition breaks down. This is not surprising, since this intuition is based upon what we experience at 1 atmosphere! For example, recently it has been shown that if sufficiently compressed two of the so-called alkali metals (lithium and sodium), actually become semiconductors or even insulators. Moreover, lithium and beryllium (two elements which do not mix under normal conditions) form stable compounds when squeezed.

We have recently shown that under pressure various LiHn (n > 1) phases become stable or metastable (at 1 atmosphere only Li, H and LiH are known) [23]. What makes our work so exciting is that all of these materials become metals at about 1/4th the pressure necessary to metalize pure hydrogen itself. Our calculations indicate that these hydrogen-rich systems may be high-temperature superconductors at experimentally accessible presures.

We are currently using genetic algorithms (see below) in order to predict if other hydrogen-rich systems, and exotic chemical combinations which are not found on earth may be made in a diamond anvil cell.

This work is funded by the NSF Division of Materials Research and the Office of Cyberinfrastructure.

Evolutionary algorithms

Determining the geometries of molecules and solids is one of our main interests. One way to predict stable structures is by using "chemical intuition". But what happens when we enter a new situation, especially one where we do not have enough experience to build up an intuitive way of thinking? Take for example solids under high pressure. The periodic table, and all of the typical bonding rules we employ are based upon our experience at 1 atm. When things are squeezed, it is necessary to determine new rules which can be useful for finding out what the most stable structure will be.

In order to help us gain some chemical intuition in these types of circumstances we need experience. And we also need a good way to predict the most likely geometries. Evolutionary algorithms (EAs) are one method which can be employed. These use the ideas of Lamarckian evolution. The initial structures are randomly generated or seeded by the user. After optimization, the most stable systems are used to generate a set of "children" by combination of two "parents", or mutation by genetic operators.

XtalOpt, our free and open source EA designed to predict crystal structures, has just been released! Get the source code and documentation. XtalOpt has been implemented as an extension to the Avogadro molecular editor.

This work is funded by the NSF Division of Materials Research and the Office of Cyberinfrastructure.


MAO, methylaluminoxane, is one of the most commonly used activators in olefin polymerization. Despite extensive wor k MAO is often considered a ''black-box''. It would be desirable to know the structure and function of MAO, since a high Al:catalyst ratio (~10,000:1) is necessary for good polymer activities. Previously, we have used the results of DFT calcul ations to propose a model for MAO, study its interaction with trimethylaluminum, predict which species are dormant and act ive, as well as investigate the mechanism of olefin polymerization [8].

Yet various puzzles remain unsolved, especially when supported MAO is considered. We would eventually like to answe r the question: "Why does employing a support lower the Al:catalyst ratio required for polymerization by a dramatic t wo orders of magnitude?" Modelling this complicated system is a challenge we would love to pursue!

This work is funded by a Doctoral New Investigator Grant from the American Chemical Society Petroleum Research Fund.

Organic/Metal Interfaces

The rational engineering of self-assembled monolayers (SAMs) is one of the goals of the bottom-up approach of nanotechnology. SAMs show promise in various applications ranging from protective coatings, to organic light emitting diodes (OLEDs), and even molecular electronics and molecular traps.

We are carrying out computations whose goal is to unveil a chemical understanding of the bonding in these systems. The chemistry of the adsorbate-surface junction can be explained using the results of DFT-D calculations coupled with a hybrid of MO and band theory. Our ultimate goal is to explain and predict the supramolecular self-assembly (or lack theroef). We hope our results will aid the rational design of architectures with desired properties!

Solvated electrons and electrides

The amazing blue and gold colors of metal-ammonia solutions have fascinated chemists ever since they were first obs erved by Sir Humphry Davy just over 200 years ago. We have recently carried out a detailed molecular orbital analysis of t he species which may be present in lithium-ammonia solutions [21]. Briefly, we found that the excess electron enters a diffuse orbital built up largel y from the LUMOs of the ammonia molecules. Overlap of these orbitals gives rise to ubiquitous H---H bonding interactions. TD-DFT calculations of the optical absorption spectra shed light on the "fine blue colour" of the dilute mixture s.

Yet, there is still lots of work to be done in this exciting field! Upon cooling, metal-ammonia solutions form solids such as Li(NH3)4 and Ca(NH3)6. These systems can be thought of as electrides: compounds in which the electrons , trapped in cavities and channels within the crystal, are the anions. Other related materials include organic electrides: materials in which the alkali cation is typically complexed b y crown ethers, or inside a cryptand molecule. We are interested in carrying out band structure calculations to get a bett er understanding of the electronic structure and properties of these systems.

Moreover, solvated electrons can be produced in other solvents such as ethylenediamine. In contrast to metal-ammoni a solutions, the optical absorption spectra depends upon the metal. The metal dependent bands have been attributed to alka li metal anions, M-. Indeed, the group I metals are known to have three oxidation states: 0, +1 and also -1! We would like to explore this computationally.


A tremendous amount of research has been directed towards studies of clusters since they may be used as the building blocks for tailored nanomaterials ("bottom-up" approach of nanotechnology). It is particularly interesting to determine the factors which make a cluster magic, that is give it enhanced stability.

So far, our work in this field has focused on metal-fullerene clusters [12, 18], M(C60)2 where M=K, Ba. Here, the most stable systems determined experimentally could not be rationalized by electronic or geometrical shell filling arguments. Our analysis showed an interplay between ionic (K, Ba) and covalent (Ba) bonding. The entropic term to the Gibbs free energy was also found to be crucial in determining which clusters were the most stable.

Our interests have now turned towards gold clusters. We would like to determine their geometrical and electronic structures, and to find out how their properties change with increasing dimensionality. What will the electronic structure and properties of a 1D wire, a 2D sheet or a 3D solid made up of these units be? And how will it depend upon the nature of the molecular building block itself? These are some of the fascinating questions we would like to address.

Carbon nanotubes

One of the reasons why so much research has been directed towards both single (SWNTs) and multi-walled carbon nanotubes is that their physical and chemical properties are intimitely dependent upon their geometries, structural defects, interactions with other tubes, as well as chemical modifications. This versatility has resulted in a wide range of proposed applications spanning from molecular electronics to biotechnology. In order to achieve rational materials design it is important in order to be able to fully characterize samples synthesized using a given method.

NMR (nuclear magnetic resonance) is one of the most versatile tools to study the geometry and electronic structure of solids and molecules. We have calculated the NMR chemical shifts of pristine (finite and infinite) SWNTs, as well as infinite tubes possessing Stone-Wales defects, and those which have been covalently functionalized [22]. Our results indicate that a wealth of knowledge may be obtained from the NMR spectra of nanotube samples.

Another area we are interested in is how finite size, and capping affect the electronic structure of SWNTs. Can the properties of the tubes be engineered by controlling their length, and how they are terminated?

Wannier functions

Wannier functions are a powerful tool to visualize bonding in the solid state. The NMTO method (muffin-tin orbitals of order N) can be used to construct minimal basis sets that accurately describe a group of bands, a band, or even just the occupied art of a band! Symmetrical orthonormalization of the NMTOs yields a set of Wannier (or Wannier-like) functions which are atom-centered and localized by construction.

We have employed the NMTO method in order to visualize the pressure induced electronic s to d transition in cesium [10], and also to gain insight into the mechanism responsible for the superconductivity in the alpha and beta allotropes of ThSi2 [24]. The Wannier function for the inerlayer band showed a striking resemblance to those found previously for superconducting intercalated graphites.