M. S. Senn, C. A. Murray, X. Luo, L. Wang, F.-T. Huang, S.-W. Cheng, A. Bombardi, C. Ablitt, A. A. Mostofi and N. C. Bristowe, Symmetry switching of negative thermal expansion by chemical control,
J. Am. Chem. Soc. 138, 5479 (2016)

Most materials expand when they are heated, but a few do the opposite as a result of certain kinds of atomic vibrations. Ideally, we would like to be able to precisely dial-in a material’s thermal response, but in these complex systems fine control has been elusive.

In this work, which appears on the cover of JACS, the layered perovskite Ca_3–xSr_xMn₂O₇ is shown to exhibit a switching from a material exhibiting uniaxial negative to positive thermal expansion as a function of x. The switching is shown to be related to two closely competing phases with different symmetries. The negative thermal expansion (NTE) effect is maximized when the solid solution is tuned closest to this region of phase space but is switched off suddenly on passing though the transition. Our results show for the first time that, by understanding the symmetry of the competing phases alone, one may achieve unprecedented chemical control of this unusual property.

This work was done in collaboration with Mark Senn at the University of Oxford and others and forms part of Chris Ablitt‘s PhD in the Centre for Doctoral Training in Theory and Simulation of Materials. Chris is supervised by Nicholas Bristowe, Mark Senn and Arash Mostofi.

Robert A. Bell, Simon M.-M. Dubois, Mike C. Payne and Arash A. Mostofi, Electronic transport calculations in the ONETEP code: implementation and applications, Comput. Phys. Commun. 193, 78 (2015)

Over the last decade, there has been increasing interest in modelling electronic transport at the atomic scale. In particular, the need to capture the physics of emerging nanostructured electronic devices calls for quantum-mechanical simulations. The Landauer-Buttiker formalism coupled with density functional theory (DFT) is used successfully to model a wide range of transport phenomena. Simulations realistic nanoscale devices containing several thousands, or tens of thousands, of atoms whilst retaining the gold-standard plane-wave accuracy remains a challenge.

In this work, we have developed a DFT-based method to investigate Landauer-Buttiker electronic transport in devices containing several thousands of atoms. The implementation is based within the ONETEP linear-scaling DFT code that relies on a minimal set of localised Wannier-like orbitals. As these generalized Wannier functions are optimised with respect to an underlying systematic basis of functions equivalent to a set of plane-waves, the implementation retains the characteristic of controllable accuracy. The method is demonstrated by applying it to study conductivity in carbon nanotube networks and is a step toward the goal of rational design and optimisation of realistic nanoscale devices from first-principles.

Robert A. Bell, Mike C. Payne and Arash A. Mostofi, Does water dope carbon nanotubes?, J. Chem. Phys. 141, 164703 (2014)

Carbon nanotubes (CNTs) possess remarkable electronic and physical properties, and many applications have been proposed since their discovery. For any application to be practical, however, these properties must be robust to their local environmental conditions. One technologically-important impurity is water, which may be found not only in the surrounding atmosphere, but also as a result of purification processes that are done in solution. Experimental studies show that water does have an effect on CNT conductivity, yet observations have been contradictory, with both increases and decreases in conductivity observed. A disagreement exists over the effect, and indeed the existence, of charge transfer between water and CNTs.

Using large-scale density functional theory calculations, we show that the interaction between water and CNTs is remarkably long ranged, such that it could not be accurately represented in previous calculations in the literature employing smaller system sizes. The dominant effect is due to the classical electrostatic interaction of the water molecule dipole with the electronic charge density of the CNT. This accounts for the vast majority of the redistribution of charge in the CNT, which had previously been misinterpreted as charge transfer. So, with regard to the question posed at the start, water does not dope carbon nanotubes.

This work is part of Rob Bell’s PhD.

Robert A. Bell, Mike C. Payne and Arash A. Mostofi, Improving the conductance of carbon nanotube networks through resonant momentum exchange, Phys. Rev. B 89, 245426 (2014) (PDF, SUPPLEMENTARY)

Individual single-walled carbon nanotubes (CNTs) possess remarkable electron transport properties and many applications have been proposed since their discovery. To have a significant technological impact, however, the properties of single CNTs must be scalable to the level of devices and mass production. One approach is to use the CNTs in so-called ‘bulk form’, such as wires and mats that are fabricated as a complex network of many individual CNTs. These systems have attracted considerable interest recently. Their conductance, however, is orders of magnitude lower than the theoretical maximum achievable on the basis of single CNTs and is believed to be largely limited by the slow tunnelling of electrons between CNTs in the network.

Rob’s theoretical analysis of inter-tube tunnelling between adjacent CNTs identifies a new mechanism to improve the electronic conductance of CNT networks by an order of magnitude. Whilst it has been known theoretically for some time that inter-tube conductance between CNTs of different chirality is strongly suppressed, how this result impacts real CNT networks has not been clear. Previous theoretical and computational approaches have focussed largely on conductance between terminating ends of CNTs and/or considered tunnelling between CNTs of the same chirality. These studies are, therefore, not representative of realistic networks that can consist of relatively well-aligned CNTs in which side-wall tunnelling is significant and in which there are a wide range of chiralities.

Using analytical theory based on scattering theory, validated by numerical tight-binding simulations, Rob has shown unequivocally that significant enhancements of inter-tube conductance can arise from relatively weak interactions with the surrounding environment. The origin of the enhancement is that these interactions enable momentum-conserving tunnelling to occur by accepting any momentum difference between the initial and final states. The improvement in inter-tube conductance is very general and applies to metallic as well as doped semiconductor CNT networks.

A key point is that the interaction with the environment need only be weak, therefore, it can be effected without chemical bonding. A number of possibilities for the experimental realisation and control of such conductance-enhancing interactions are suggested and our results highlight the crucial importance of considering the effect of the environment when studying nano-scale devices.

This work is part of Rob Bell’s PhD.

Fabiano Corsetti and Arash Mostofi, Negative-U properties for substitutional Au in Si, Europhysics Letters 105, 57006 (2014) [ONLINE JOURNAL|PDF]

It’s well-known that atomic defects are crucial for the operation of semiconductor devices. Not all defects, however, are desirable. Gold dopants in silicon, for example, introduce electronic levels that sit deep within the band gap of silicon, acting as electron traps, leading to a large decrease in conductivity. Understanding the properties of gold in silicon has also received recent interest due to the use of gold as a catalyst in the growth of semiconducting nanowires in a process that can result in the presence of gold impurities in significant excess of the solubility limit that would be expected in bulk silicon.

Fabiano has addressed a long-standing puzzle in the properties of the gold substitutional defect in silicon, namely that there is no electron-paramagnetic resonance (EPR) signal associated with it. This is particularly puzzling since the isoelectronic Pt^1- impurity in silicon does show a strong EPR signal. Why, then, does gold, which is adjacent to platinum in the periodic table, behave so differently? This question was first posed by F.G. Anderson [J. Phys.: Condens. Matter 3, 4421 (1991)] and has since remained without a definitive answer.

His state-of-the-art first-principles calculations suggest that the reason for the lack of an EPR signal for Au in Si is that this defect is what is known as a “negative-U” center, i.e., the system is thermodynamically stable only when the electronic states associated with the defect are doubly occupied. It is this suppression of unpaired spins associated with the defect that results in the lack of an EPR signal in experiments. The robustness of his results is supported by the quantitative agreement, within chemical accuracy, between the experimentally measured defect level transitions and his calculated predictions.

This work was part of Fabiano Corsetti’s PhD.

Fabiano Corsetti and Arash Mostofi, A first-principles study of As doping at a disordered Si-SiO₂ interface, J. Phys.: Condens. Matter 26, 055002 (2014) [ONLINE JOURNAL|PDF]

Understanding the interaction between dopants and semiconductor-oxide interfaces becomes increasingly important as transistors are made smaller and smaller. As part of this work we developed atomistic models of realistic silicon-silica interfaces using a combination of first-principles density-functional theory and a continuous random network Monte Carlo method. We then used these models to study the properties and energetics of segregation of electrically active arsenic donors.

We found that a small percentage (around 10%) of the atomic sites in the first few monolayers on the silicon side of the interface are energetically favourable for segregation, and that this is controlled by the local bonding and local strain of the defect centre. We also found that there is a long-range quantum confinement effect due to the presence of the interface, which results in an energy barrier for dopant segregation, but that this barrier is small in comparison to the effect of the local chemical environment. Finally, we showed that segregation can be controlled by the application of strain to the interface and that this could provide a mechanism for experimental control of dopant segregation in real devices.

This work was part of Fabiano Corsetti’s PhD.

N. Marzari, A. A. Mostofi, J. Yates, I. Souza and D. Vanderbilt, Maximally-localized Wannier functions: theory and applications, Rev. Mod. Phys. 84, 1419–1475 (2012) [ONLINE JOURNAL|PDF]

Arash, together with collaborators from EPFL, Rutgers, Oxford and Universidad del Pais Vasco, has published a paper in the Reviews of Modern Physics entitled “Maximally localized Wannier functions: Theory and applications”. The dichotomy between extended, or Bloch, states, and localized, or Wannier, states in solid-state physics has been known for 75 years. However, the notion of a unique, or maximally localized, Wannier function is of much more recent origin. This review explains this concept, shows how to explicitly construct maximally localized Wannier functions, and explains why they are useful for describing systems from solids to optical lattices and cold atoms.