July 2012 summaries
In recent years, the study of nanomaterials under pressure has attracted multidisciplinary scientific and technological interest. In part due to their increased ratio of surface to volume atoms, nano-objects display a host of properties that differ from those of their bulk counterparts. Using pressure to control semiconducting materials such as silicon (Si) and cadmium sulfide (CdS) nanoclusters, which display physical properties depending sensitively on their structure, has generated particular interest. Attaining such control at the nanoscale holds the promise of a host of novel technological applications such as biomarkers, tunable photovoltaic devices, shock-absorbers, biomarkers, nanoscale stress sensors and quantum transistors; all of which could have major societal and economic impact.
Accurate quantum-mechanical simulations are essential for understanding the structural transformations and resulting changes in optical and electronic properties of nanomaterials under pressure, which are difficult to observe clearly experimentally. However, the computational cost limits the length- and time-scales that can accurately be simulated. In this project,we have successfully overcome some of these bottlenecks by combining a quantum-mechanical code for which the computational cost increases linearly with the number of atoms (in contrast to the cubic scaling of traditional simulations) with a novel method for applying pressure. Our focus in this project has been on Si and CdS nanocrystals that have been widely studied experimentally and are favoured in technological applications. Some of the work was recently published and we hope it will help theorists modelling nanomaterials under pressure as well as experimentalists devising improved devices made of these materials.
Our previous work had indicated that calculations of the formation energy and final atomic geometry of isolated vacancy (missing atom) defects in carbon nanotubes were strongly influenced by the type of boundaries placed upon the structure: if the structure is allowed to have ‘free ends’ this leads to a lower vacancy formation energy and different structure to the case when the nanotube is periodically repeated (which is the model most commonly used in the literature). The reason for this difference is that there are modes of relaxation (structural changes to minimise energy) which involve twisting and bending, which can’t be captured in a periodic model. The objectives of this project were to confirm our findings and demonstrate that this was a general phenomenon not limited to single vacancies in carbon nanotubes. In particular, we wished to extend our studies to narrow ribbons of graphene, which have recently attracted significant attention for their potential in nano-electronic devices.
As a result of detailed, accurate calculations, we determined that our hypothesis was correct. For example, divacancies (two missing atoms) in carbon nanotubes caused them to bend and twist, a phenomenon which could only be captured with ‘free ends’. Likewise, vacancies and Stone-Wales defects (the inclusion of non six-membered rings into the structure without the addition or removal of atoms) in graphene nanoribbons showed the same behaviour. As a result, to find the formation energy, geometry and physical properties of isolated vacancy and Stone-Wales defects calculations should be performed with open boundaries (‘free ends’).
The dwindling amounts of naturally found hydrocarbon-based fuels and stringent emission regulations demand a search for alternative fuels for power generation using industrial gas turbines. Fuels such as synthetic gas (Syngas) and Blast Furnace Gas (BFG), are already being considered for this purpose. These fuels are mainly composed of fuel gases such as carbon monoxide (CO), hydrogen (H2), methane (CH4), and so on, having widely differing thermo-chemical characteristics along with other gases such as carbon dioxide (CO2) and water (H2O). The relative proportions of these gases in the fuel gas mixture depend on the production process and the calorific value of these fuel mixtures are usually low. Thus, the turbulent combustion characteristics of these fuels under gas turbine conditions vary widely and predicting them is a challenging task using the currently available combustion models. This is mainly because of non-equal molecular diffusivities of and heat release rates from the various fuel gases present in the multi-component fuel gas mixture, while the current modelling for turbulent combustion is predominantly for single component fuels. These models are typically developed using data from direct numerical simulations (DNS) of turbulent combustion. These simulations are often conduced using a reduced chemistry because of the computational cost incurred for simulations with detailed chemistry and molecular transport. This typical approach for DNS of multi-species fuels combustion is inadequate and it is unclear whether the currently available turbulent combustion models are applicable to multi-species fuels with further chemical and physical complexities.
The allocated computational resources were used to conduct DNS of turbulent premixed combustion of a multi-species fuel mixture containing CO, H2, H2O, CH4 and CO2 and air with composition akin to BFG. The combustion chemistry is represented using skeletal mechanism involving 49 reactions and 15 species developed specifically for such fuels . These simulations are first of their kind and three simulations involving skeletal and two involving reduced chemistry are conducted. The main objectives were (1) to analyse turbulent flame structure of such fuels, (2) to examine the performance of currently used flame observables employed for laser imaging of heat release rate (HRR), (3) to evaluate the performance of some commonly used combustion sub-models, and (4) to evaluate the performance of a reduced mechanism, also developed in , under turbulent conditions. These objectives are the same as those proposed in the case for this resource allocation panel (RAP) award.
It was found that heat release occurs over a wider temperature range for the multi-species fuel-air combustion on contrary to methane-air flames . This is because of the presence of many distinct and non-overlapping reaction zones consuming the fuel species present in the multi-species fuel mixture. Furthermore, it was shown in  that the currently used flame observables are inadequate for the visualization of the HRR, and alternative flame markers are proposed in . The DNS data are used in  to validate the use of these markers under turbulent combustion, which showed significantly improved HRR correlations. Five different commonly used mean reaction rate closures are tested using these DNS data. Although the overall agreement is found to be good for all these models at low turbulence levels, significant differences were observed between the models at higher turbulence levels.
These results are reported in . It was observed that some of the models' constants showed increased sensitivities to the turbulence level often requiring adjustment. In addition, for low turbulence levels the value of these constants was different from the proposed constants thus revealing the sensitivity of these models to the mixture composition also. Further parametric studies are thus required to correctly calibrate these constants. A plausible modelling for large eddy simulation (LES) of multi-species fuel combustion is tested in . Finally, the two sets of DNS data generated using the reduced chemical kinetics are compared to the ones obtained using skeletal mechanism. It was found that the reduced mechanism requires significantly reduced computational time and memory, often by a factor of two, while retaining similar level of accuracy. Furthermore, the reduced mechanism developed in  is shown to produce the same turbulent flame statistics as the skeletal mechanism implying that the reduced mechanism is a good substitute and this sort of evaluation has not been attempted before. The results of these analysis are reported in .
 Z. M. Nikolaou, J. Y. Chen, N. Swaminathan. `A 5-step reduced mechanism for combustion of CO, H2, CH4, CO2 mixtures with low hydrogen/methane and high H2O content', Combust. Flame 160 (2013) 56-75.
 Z. M. Nikolaou, N. Swaminathan. `Direct Numerical Simulation of multi-component fuel combustion with detailed chemistry,' Combust. Flame, Submitted October 2013.
 Z. M. Nikolaou, N. Swaminathan. 'Flame markers for heat release rate visualization,' Combust. Flame, Submitted in November 2013.
 Z. M. Nikolaou, N. Swaminathan. `An evaluation of combustion sub-models for RANS simulation LES of multi-component fuels,’ to be submitted, January 2014.
 Z. M. Nikolaou, N. Swaminathan. `Scale-similarity modelling in the context of LES using detailed chemistry DNS data’, to be submitted, January 2014.
 Z. M. Nikolaou, N. Swaminathan, J. Y. Chen. `Evaluation of a reduced mechanism for turbulent premixed combustion,' Combust. Flame, Submitted November, being revised December 2013.
Self-sustained oscillations occur in many processes involving fluid dynamics. These can be at the micro-scale (eg micro-emulsions) or at the large-scale (eg combustion). Analyses of self-sustained oscillations with advanced instability methods (see the EPSRC-funded Advanced Institute of Management (AIM) Network have so far been performed on simple geometries because the techniques could be most easily proven on simple systems. The aim of this project was to use these methods on a complex geometry. Specifically, we chose to examine the influence of the geometry of a fuel injector nozzle on the self-sustained hydrodynamic oscillations in a combustion chamber. This is a particularly interesting problem because the most practical way to control the flow is by changing the nozzle geometry and these techniques show the optimal way to do this.
The fuel injector we studied was from a helicopter engine (designed by Turbomeca). Its behaviour has already been studied extensively with Computational Fluid Dynamics and experiments. We studied the 3D global stability and sensitivity of the swirling flow set up in a combustion chamber by this injector. We investigated how changes at the nozzle walls affect the stability of this flow. With adjoint-based sensitivity analysis, we identified the locations of the nozzle that are most influential at changing the stability of the flow, and the directions in which they should be changed in order to stabilize or destabilize the flow most effectively. As expected, we found that changes in the nozzle and inlet profiles are very influential because they change the flow in the most sensitive regions of the upstream part of the combustion chamber. This shows that important insights on control of global instabilities can be gained by including a nozzle region.
The research of this 3D flow problem was computationally expensive, and was made possible by the High End Computing Terascale Resources (HECToR) 1b award. The award resulted in a large amount of data on linear and nonlinear behaviour of the flow, which is currently being summarized in form of publications to be submitted to high-impact fluid dynamics journals.
Understanding the relation between the morphology and the electronic properties of semiconducting polymers through large-scale simulations
Semiconducting polymers are potentially useful for disparate electronics applications requiring inexpensive and flexible materials including light emitting diodes (LED), thin-film transistors and photovoltaic cells. Unfortunately, new materials are essentially discovered and introduced by trial and error because the relation between the chemical composition of the material and the charge mobility (the most important characteristic for an organic semiconductor) is not known. In the past years we have shown that it is possible to describe the electronic structure of relatively simple organic semiconductors (molecular crystal or semiconducting polymers) from the bottom up, in other words starting from an atomistic classical study of their morphology followed by large scale electronic structure calculations. With this project, we have extended these computational capabilities to the more challenging case of amorphous polymers. The objective of this project was to consider two prototypical typical amorphous semiconductors, in other words two derivatives of polyphenylene vinylene (PPV) (MEH-PPV and CN-PPV) and to derive the most accurate description of its atomistic bulk structure and the corresponding electronic structure. Such benchmark calculations have been used to answer a number of key questions: (i) What is the electronic density of states of PPV in the energy range relevant for transport? (ii) What it the typical localization of the charge carrier states in PPV? (iii) How strong is the coupling between localized states in PPV? The results of the calculation provide the first reliable description of these key properties of the polymer and can be used to build a model of charge transport based on realistic atomistic model, in other words one can begin to explore computationally the relation between structure and properties of semiconducting polymers. Interesting, the answers to the questions above contained few surprises. It was found that the one-electron states that describe the charge carrier in semiconducting polymer cannot be identified on the basis on the geometric conjugation breaks of the polymer and the full calculation of the electronic structure is necessary. Moreover, the delocalization of such states is energy dependent, in other words it is not possible to define a characteristic localization length for these states.
Ultra long time window simulations of optical communication systems on General-Purpose computation on Graphics Processing Units (GPGPUs)
The Digital Economy has been one of the strongest growth markets over the past decade and, with high definition video on demand, online gaming, e-health and other bandwidth hungry applications on the horizon, will continue to do so for the foreseeable future. Access networks are currently being upgraded to “very high speed digital subscriber lines” (BT Infinity) and “fibre to the home”, which will bring more of the benefits of high speed networking to ever more people living in the UK. The driving force behind this success story is the availability of networks with vast amounts of high quality affordable bandwidth. The most prominent example of such a network is of course the Internet itself. However, for this new economy to continue to thrive an ever increasing amount of bandwidth at decreasing costs is needed. With customer access speeds currently increasing significantly, the bottleneck is once again shifting to the backbone networks and creates demand for bandwidth upgrades. This project developed and tested the simulation tools required to deliver on this need.
The overall scientific aim of our project was to develop and apply a massively parallel simulation code for next generation fibre optic based communication systems. This proposal builds on our current HECToR class 2a pump-priming grant, where we developed an Open Multi-Processing (OpenMP)/Message Passing Interface (MPI)-based scalable code, exploiting the multi-core fat node structure of HECToR. Our strategy was to extend this implementation into an even more parallel version using thousands of threads on a GPGPU. Halo data are exchanged between compute nodes via standard MPI communication routines. Our code automatically exploits the computational capabilities of the Compute Unified Device Architecture (CUDA) GPGPUs optimally.