Search

BACKGROUND: Dengue virus outbreaks are increasing in number and severity worldwide. Viral transmission is assumed to require a minimum time period of viral replication within the mosquito midgut. It is unknown if alternative transmission periods not requiring replication are possible. METHODS: We used a mouse model of dengue virus transmission to investigate the potential of mechanical transmission of dengue virus. We investigated minimal viral titres necessary for development of symptoms in bitten mice and used resulting parameters to inform a new model of dengue virus transmission within a susceptible population. FINDINGS: Naïve mice bitten by mosquitoes immediately after they took partial blood meals from dengue infected mice showed symptoms of dengue virus, followed by mortality. Incorporation of mechanical transmission into mathematical models of dengue virus transmission suggest that this supplemental transmission route could result in larger outbreaks which peak sooner. INTERPRETATION: The potential of dengue transmission routes independent of midgut viral replication has implications for vector control strategies that target mosquito lifespan and suggest the possibility of similar mechanical transmission routes in other disease-carrying mosquitoes. FUNDING: This study was funded by grants from the National Health Research Institutes, Taiwan (04D2-MMMOST02), the Human Frontier Science Program (RGP0033/2021), the National Institutes of Health (1R01AI143698-01A1, R01AI151004 and DP2AI152071) and the Ministry of Science and Technology, Taiwan (MOST104-2321-B-400-016).

The current state of development in battery technology is not adequate for modern needs in renewable energy. Li-ion batteries provide the greatest capacity, but Li metal anodes can lead to dendrite growth and catastrophic failure. We can reduce this by pulse charging, but this slows the charging rate [#1056, 1117, 1133]. To reduce the danger of dendrite growth, we can use polymer [#1016] or ionic liquid electrolytes, but this decreases the Li mobility. We can replace Li anode with Na, but this reduces capacity (see [#1168]). We can use other electrodes but this generally reduces capacity [1154, 1098, 1046, 1068]. These are all areas that we have modeled at the atomistic level with some success in understanding some of the issues. The most important area for the theory is to understand may be the nature of the solid–electrolyte interface (SEI). The Li-ions and Li metal electrode react with the electrolyte to produce the SEI, which must play an essential role in charging and discharging since the ions must be transported through this layer and then must be desolvated as they are deposited on the Li metal surface. We have used QM to characterize the SEI for the Li electrode-ionic liquid interface.

We carried out the first simulations of Ned Seeman’s DNA crossover structures that laid out how to do self-assembly with DNA [#594, 651, 707, 1182]. Tod Pascal studied the branched three-way junction [#995] where he found that the stable structure is strongly affected by salt concentration and by the ratio of MgCl₂ to NaCl [995]. He also applied 2PT method to calculate the entropy and free energy changes during the rearrangement showing that the barrier is entropic. While he was a materials science graduate student, Si-Ping Han used DNA origami to self-organize carbon nanotubes (CNT) into field-effect transistors. In addition he used DNA to organize CNTs into parallel bundles. But, the major breakthrough in nucleic acid devices was Si-Ping Han’s idea to develop a dual double helix of RNA we refer to a conditional siRNA (cond-siRNA). This incorporates a sensor element targeted to say HIV or AML that in turn serves as a logic device that only in the cells expressing proteins for this virus disassembles to release a 23 RNA fragment that incorporates into the dicer element of the ribosome to kill the cell by preventing expression of a protein vital to the cell. This promises to enable very selectively killing of diseased cells while not affecting adjacent cells that do not have the virus.

Our studies on thermoelectrics have been primarily motivated by the innovative developments by Jeff Snyder mostly in collaboration with Guodong Li of Wuhan University in China, but we did one paper with Jim Heath.

In the early 2000s, there was a great deal of interest in finding materials that could store H₂ at high density and small volume for use in H₂ fuel cells, particularly for transportation. The concern was that pressured H₂ gas would be too dangerous. DOE set standards but no experimental system came close. It seems like the industry has decided that pressured H₂ is OK. In 2004 we designed a series of systems that met DOE standards for H₂ and CH₄ storage, but generally did not find experimentalists to test our designs. Then in 2006 I learned about the MOF’s being developed by my friend Omar Yaghi at UCLA that I thought would be ideal for storing H₂ and CH₄. We did a series of papers, sometimes in collaboration with Omar, showing designs of MOFs and later COFs that would satisfy DOE requirements.

In recent years, the Goddard group has turned to electrocatalysts as ideal systems for catalyst design where one can combine anode and cathode electrodes, with electrolytes, and applied fields to provide the flexibility to control complex reactions. In addition to the oxygen reduction reaction (ORR) at the cathode of the hydrogen fuel cell discussed in Chap. 64 and the water-splitting reactions, HER and OER discussed in Chap. 65, the other the main applications relate to CO₂ reduction (CO₂R) to form fuels or hydrocarbon products. These studies validate the level of DFT theory and the level of explicit solvent that can be used for the design of improved electrocatalysts.

In 1998 Vaidehi Nagarajan and I initiated a project to predict the structures of G protein-coupled receptors (GPCR) from first principles (there were no crystal structures at the time). Our early methods were successful in predicting fairly accurate structures for several GPCRs, and we published the first GPCR crystal structures. First for olfactory receptors and then in 2004, the dopamine D₂ and adrenergic b2 structures. The first experimental X-ray crystallography structure was for bovine rhodopsin and then for b2 AR in 2005, which showed that our 1st principle structure was rather accurate. Later we used the templates of transmembrane tilt from experiment to develop methods for predicting the low energy packing of GPCRs sufficiently stable to bind to ligands, with a number of successes. More recently, we have focused on the mechanism by which the GPCR attached to a Gprotein activates the G protein after binding an agonis.

A major breakthrough for multiscale reactive simulations is the ReaxFF reactive force field [#471] developed by van Duin and me in 2001. ReaxFF has enabled the simulations of wide-ranging reactive systems, including shock impact on an energetic materials polymer composites (model for a plastic bonded explosive, PBX) with 3.7 million atoms for 60 ns (see [#951]), the simulation of the effect of hydration on the strength of concrete [#978], elucidation of the mechanism of complex ammoxidation catalysts [#934], and the product distributions from hypervelocity impact with molecular clusters [1004]. Recently, we have developed a new generation, RexPoN that may provide even higher accuracy.

It is straightforward to predict electronic energy from QM and the potential energy from FF-based MD and more recently from QM-based MD. However, extracting entropy and free energy has been problematic. Generally, the accepted methodology, going back to Jack Kirkwood and Richard Tolman, is thermodynamic integration theory or free energy perturbation theory. These methods are rigorous for obtaining free energy differences if the perturbations are sufficiently slow that the system remains in equilibrium as system A morphs into B. This generally requires repeated equilibrium calculations during the MD, which makes it very expensive for large-scale (100,000 atom) systems. A major advance here is the validation of FEP technology by Bill Jorgensen and its implementation into an automatic module by Schodinger. To make entropy and free energy calculations practical for nanosecond reactive simulations of large systems with up to millions of atoms, Lin, Blanco, and I (LBG) developed the two-phase thermodynamics (2PT) method that is generally 1000’s of time faster than thermodynamic integration (TI) but equally accurate.

A valuable aspect of reactive MD is to provide rate constants of complex reactions. We give some examples for cases of detonation and pyrolysis. For the general case we need a good machine learning model to collect together the reactions at any one temperature so that we can extract rate constants. Since a reactive force field such as ReaxFF, describes all possible reactions of a system, it could be the basis for general approaches to extract the fundamental kinetic parameters to describe combustion (say of diesel full, gasoline, or JP-10 Hydrocarbon Jet Fuel) or shock-induced decompositions. Although great progress has been achieved in first principles modeling of low-pressure gas-phase reactions, it has been difficult to provide first principles predictions of Combustion and Shock processes for condensed phases. We will discuss several activities we have been pursuing to accomplish this.

Artificial photosynthesis (AP) using solar energy to convert H₂O into H₂ (a fuel) and O₂ is a most promising approach to a carbon-neutral cycle and scalable energy storage. Electrocatalysis provides an attractive candidate route to AP, which could extend to all intermittent renewable energy resources. A major challenge in renewable energy technology is water splitting, which uses solar radiation to photoelectrochemically convert water molecules into H₂ (a fuel) and O₂. Here both the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER) present challenges for the catalysts. The detailed reaction mechanisms had not yet been established for either one. We did the first mechanisms under electrochemical conditions and including free energy reaction barriers for the transition states for both systems. Here we separately consider the two electrochemical half-reactions, HER and OER, which require drastically different catalysts for optimal performance.

As a graduate student (GS), I had worked out that the GVB wavefunction for CH₄ could lead to localized pairs of VB-like orbitals along each bond but the full wavefunction would still have Td symmetry. Also, I had shown that the GVB wavefunctions could dissociate properly as a bond is broken, going to atomic limits. However, I had not yet learned to program a computer, and indeed, I avoided reading any papers trying to do QM on materials because I had learned that they were completely useless. After joining the chemistry department in Nov. 1964 I learned how to program and over the next few years learned about bonding for main group and for bonds to metals. We also learned about mechanisms for reactions based on Valence Bond concepts. Ideas about resonance were worked out. For H₂ and He₂ we were able to get a VB view of all excited states. These ideas were extended later to compounds involving other main group elements and transition metal. However, for metallic materials such as Fe and brass, our methods are still deficient.

When I joined Pol Duwez’s group in Sept. 1960, I was interested in understanding how the interaction of electrons between atoms controlled the crystal structures of metal alloys. I wanted to derive this understanding from first principles. This led eventually to my development of the GVB method and then joining the chemistry faculty. This chapter summarizes the insights we have obtained about metal and their alloys, but we have still not succeeded at my original goals. Hopefully we can succeed over the next 20 years.

Although not a prime interest in my group (and never funded), we got heavily involved in high Tc materials. After a false start where we assumed incorrectly that the doping leads to hole in the CuO₂ plane, we eventually showed that the hole is out of the plane, next to the dopant. It leads to a degenerate state with 3 electrons between two degenerate e levels, leading one hole that become conducting above 5% doping and couple to antiferromagnetic Cu at the boundary. This explains most strange properties of Cuprates. We have not been able to predict Tc, but we showed how to order the dopants to double Tc.

Fraser Stoddart and I have interacted closely since 2004 (monthly and sometimes weekly when he was still at UCLA) and we have published no fewer than 46 papers together. This has been a great collaboration because Fraser relies on advice from computations for nearly all of the revolutionary materials utilizing the mechanical bond that he has been developing. We have used a variety of theory and methods: quantum mechanics (QM), molecular dynamics (MD), Monte Carlo, or QM tunneling, which we have extended to problems necessary to deal with the complex mechanically constrained systems Stoddart has been designing, synthesizing, and characterizing.

The Goddard group has been a leading force in developing new generations of organometallic catalysts, working with experimentalists like Brent Gunnoe, Jay Groves, Roy Periana, and Alan Goldman to design the ligands in advance of experiments, greatly accelerating the development of improved catalysts. In these studies, the Goddard group showed tremendous insight and intuition about reaction mechanisms combined with their deep understanding of how to include high levels of correlation and solvation in quantum mechanics [QM] calculations of complex reaction mechanisms in various solvents has been critical. Their predictions nearly always preceded the experiments, often eliminating many potential catalysts that would have to be much less active or selective, and were nearly always correct. Our only real mistake was waiting for the experimental confirmation before publishing. This made the theory easier to publish in good journals, but the result is that the experimental community does not have a way to judge the accuracy in advance of experimental validation.

With the development of ReaxFF, an important application is toward impact induced reactive decompositions of large-scale polymer-energetic material (EM) composites. Here we were able to simulate the steps leading to decomposition and reactions for systems with 3.7 million atoms. This allowed us to explain the origin of hot spots in EM and to suggest ways to eliminate them. We also used ReaxFF to characterize the CJ state for steady detonation waves in these systems (discussed in Chap. 45).

This chapter brings together various studies on materials that happen to be ceramics. This includes • Boron Carbide, (C₁₁B) CBC or B₄C, where we seek alloys or other changes to make it ductile, • The BaTiO₃ ferroelectric where we examined 90 and 180 degree domain wall boundaries, and • Silica glass • Applications to BaZrO₃ electrolytes are in the fuel cell chapter.

One of the major goals set by our humanity is to make breakthroughs in key technologies pertaining to energy, environment, and water. The main purpose of this chapter is to give an overview over sustainable CO₂ conversion technologies that could be demonstrated from high-performance materials both design and also the realization, so that they could prevent simultaneously environmental problems such as global warming.

Initially, my goal as a GS was to extract from QM unique new concepts into how chemical bonding determines the remarkable varieties of structures, properties, and reactivity. However the new GVB wave functions required more sophisticated methods to optimize the wave functions. We describe here some of these advances.

In 1962, I struck up a collaboration with Bill Goddard and several solid-state physics labs to do a systematic study of the properties of metal-semiconductor junctions. These structures are ubiquitous in semiconductor devices and are central to their operation. Understanding them is at the boundary of Physics and Chemistry.

Our group has been a leader in building realistic models of polymers using force fields based upon quantum chemical calculations. We have carried out large-scale molecular dynamics simulations of these systems to elucidate fundamental materials properties, using the molecular simulation technology described above. Highlights are the first simulations of the Tomalia dendrimers, the Percec dendritic system and the Frechet systems. We also carried out the first simulations on polyethylene and nylon. Later we carried out the first realistic simulations on the Nafion electrolyte for fuel cells. More recent work is on PEO polymer electrolytes for Li batteries. Along the way, we developed efficient methodologies to predict the structures of amorphous polymers using thermal and volume quenching and finally scaled effective solvent (SES) methodology. We extended these ideas to network polymers and nonlinear optical materials.

MD requires charges, but the only way to assign charges was to do QM, which is totally impractical for most materials problems. For proteins and DNA, one can limit the calculation to the monomers (or better trimers). Tony Rappe and I solved this problem in 1991. Saber Naserifar and I extended to polarization in 2017.

Ion-enhanced dry etch methods inflict “etch process damage” through surface ion bombardment. These inherent limitations in conventional dry etch methods create potential roadblocks to achieving device properties necessary for scaling below 10 nm. We describe an alternative dry etch method in which electrons with energies below about 100 eV stimulate precision etching of features as small as 5 nm without damage. This Low Energy Electron Enhanced Etching (LE4) method also gives atomically smooth etched surfaces, very high selectivity between materials, and maintains stoichiometry of compound materials. LE4 etches low K dielectric materials with no loss of carbon and gives Line Width Roughness (LWR) values dramatically smaller than achieved by ion-enhanced etching. In addition, LE4 is used to modify substrates in a variety of applications, like low-temperature surface cleaning and the modification of the surface architecture on multiple length scales for biotechnology applications. We have developed the electron force field (eFF) method to describe electron dynamics in highly excited electronic states and use it to show preferential bond breaking and product desorption after electronic excitation of the sample surface.

We report a first principles Quantum Mechanics (QM) study to screen chemical functionalities in polymer-based chemical sensing materials to detect sulfur dioxide (SO₂) and elemental mercury (Hg) vapors. The screening methodology involves evaluating the performance of various chemical functionalities in polymers based on their binding energy scores of the target molecules (SO₂ and Hg). The QM results were validated by comparing the actual sensor response trends with the calculated binding energy values, by performing experiments using polymer-carbon composite sensors made from the polymers with the recommended chemical functionalities. A good correlation is found between the experimental sensor responses (strong or weak) to SO₂ and Hg and the calculated binding energy values (strong or weak). The sensors were successfully used in the Third Generation JPL Electronic Nose (ENose) Technology Demonstration Experiment on the International Space Station (ISS).

Quantifying how assemblies of molecules coordinate their motions is key to understanding the operation of biological machines and other modular nanosystems. We explore the question of how tightly coupled the motions of adjacent units should be to optimize the transmission of signals, in an environment beset by thermal fluctuations. In this paper, we consider an allosteric dimer consisting of two ligand-binding domains and study how variations of ligand binding at one site are translated to similar variations in ligand affinity at the adjacent site. We compute the dynamics using a simple energy landscape model, and using mathematical methods used to study phenomenon of stochastic resonance, quantify the fraction of signal that propagates through the allosteric system as a transmission coefficient. We find that the transmission coefficient is optimized for an intermediate coupling between domains. The intermediate coupling is a compromise—when it is too low, the motion of the domains become uncorrelated, and fidelity suffers, but when it is too high, the system must surmount multiple transition barriers to change conformation, and communication slows down. There are two consequences. First, for a wide range of energy landscape parameters, the intermediate coupling corresponds to stepwise switching of domains via high energy (1.4 kcal/mol) lowly populated (<10% occupancy) intermediates. Second, the compromise between fidelity and rate caps the transmission coefficient at 40%, so that the majority of signal is degraded, even at the most perfectly formed allosteric interface. We propose that the stepwise switching observed in many allosteric proteins does not arise from an inherent limit in the mechanical couplings within the protein, but is a consequence of optimizing signal transmission; and that in general, if signals are to be communicated between domains via conformational change, a certain looseness of coupling is preferred, to provide a useful freedom of motion.

To provide some insight on odorant recognition, we partitioned the 398 human olfactory receptors (OR) into six subclasses based on the phylogenetic tree and predicted the 3D structure and binding site for the family head of each subclass. We used the GPCR Ensemble of Structures in Membrane BiLayer Environment (GEnSeMBLE) method that samples 10 trillion combinations of helix tilts and rotations to select an ensemble of the 25 most stable 7-helix packings. We found that the ensembles for the OR in all 6 subclasses exhibit the TM1-2-7 coupling characteristic of Class A G protein-coupled receptors (GPCRs). In many cases of tetrapod subclasses, the conserved R3.50:E6.30 TM3-TM6 coupling in the inactive GPCRs is replaced with D(E)3.39:H6.40 and D3.49:R(K)6.30 interactions. We found a different pattern for the fish-like class I, where D3.49 couples with residues in TM4 (R4.44 for O52D1-I and Q4.41/K4.38 for O52B6-I) instead of TM6. A variety of residues lining the binding site are involved depending on the ligand: Q103^(3.28) and N282^(7.39( for O52D1-I, Y91^(2.53) for O11H6-II, Q97^(3.28) for OR4Q2-III, S272^(7.35) for OR7D4-IV, S206^(5.43) for OR2J3-V, and T249^(6.45) for OR5P3-VI. Here, we provide the structures for the top25 for all 11 ORs, including the structures with ligands.

Including solvation is very important for systems with polar solvents. This motivated development of a number of methods for describing solvent implicitly. Our more recent methods do this at constant potential.

My group has been involved with several projects involving solar cells: The CIGS (Cu-In-Ga-Se Solar Cells), Dye-Sensitized Solar Cells, and organic PbI3 solar cells. Also we have been involved with surface plasmon generation of hot carriers near interfaces.

Materials at extreme conditions in which high temperature and pressure are fluctuating rapidly with molecule, atoms, ions and electrons (e.g., plasma) all interacting dynamically has been a no-man’s land for atomistic simulations. It is clearly foolhardy to attack this problem with our current tools and methods. Even so Su and Goddard (SG) jumped headlong into the problem, developing the eFF method to describe the non-adiabatic evolution of large-scale quantum systems.

In the 1960s, no serious calculations were done on transition metals or even the Si row of the periodic table because the core electrons had to be present, adding substantially to the cost (state-of-the-art computers were about 0.1 mips and memory was 36 K words, including operating system). To get around this problem, the concept of a pseudopotential that cancels the deep potential in the core region was introduced by Phillips, Kleinman, Cohen, Heine, and Ziman as a way to think about bonding in metals such as Li. As formulated, their pseudopotential had serious problems. 1. The pseudopotential was not unique; there are an infinite number of different pseudopotentials each of which leads to a different pseudo-orbitals, but with the same E. 2. Second, the new Hamiltonian is not Hermitian, leading to complications when considering scattering. 3. Third, the pseudopotential was an integral operator; not a local potential, causing problems when considering scattering. The Goddard group (Carl Melius and Luis Kahn) solved this problem by introducing the Effective Core Potential (ECP). Adding in core orbital character to make the valence orbital nodeless while going smoothly to zero as the distance from the nucleus goes to 0 led a pseudo-orbital that leads to a unique ECP (or pseudopotential). Most importantly, the ECP depended dramatically on symmetry. It was very different for s states than p states then d states, etc. This led to the concept of writing the ECP in terms of angular momentum projection operators. We showed that the excited states of each symmetry are well described with the same ECP. This was completed by 1974. The final refinement (1977) was restricting the form of the smooth pseudo-orbital so that the long range size was not changed (norm conserving). This concept of ECP with angular momentum projection operators extracted from ab initio atomic wavefunctions has been the basis of the enormous progress in the last 50 years for treating materials for the f

In the early 1970s, advances in experiments were giving the first clues about surface reconstruction. Our GVB calculations on clusters representing semiconductor and metal surfaces were very useful in predicting and interpreting the experiment.

Our studies on hydrogen PEM fuel cells have focused on the Oxygen Reduction Reaction (ORR), which is currently the biggest obstacle to wide spread application, and the role of the Nafion electrolyte in proton conduction and degradation. The big breakthrough here was full solvent (5 layers) solvation, which required ReaxFF to pre-equilibrate the system.

The practical size and time scale for QM based MD (QM-MD or AIMD) is 200 atoms for 50 ps. But we want to do MD on systems with 100,000 to 3 million atoms for 100’s of ns. Thus the forces from QM must be replaced for analytic functions that depend directly on distances and angles. The parameters could be trained with QM but better is to develop rules that lead to an acceptable accuracy for wide classes of materials. Thus we developed DREIDING for main group elements (columns 14–17) and later UFF for all elements (up to Lr, Z=103) aimed at inorganic materials. These are aimed at structures but not bond breaking. For bond breaking we developed the ReaxFF and RexPoN FF described in chapter 47.

My group got interested in designing nanoscale systems that could self-organize in the 1980s. This led to papers on molecular machines, where we carried out MD for systems designed by others. Usually they did not work so well because the atom sizes are discrete. One of my graduate students (Ching-Hwa Kiang) discovered the first single wall nanotube, which led to papers characterizing them and the mechanism by which they formed. We also had some collaborations with the Atwater group.

I first met Bill in the spring of 1965. My good friend George Hammond had told me that Bill was terrific and I should hang out with him.

The Goddard group was the first to apply ab initio QM to determining the mechanisms of complex heterogeneous catalysts. Particular emphasis has been on activation of propane, propene, and butane also ammoxiation to acrylonitral and N2 reducton to NH₃. Since then we have used QM to discover the reaction mechanisms for many practical heterogeneous catalysts for butane to maleic anhydride, ammoxidation, and N2 reduction (Haber Bosch).

Simulations of nanoelectronic devices with nonequilibrium Green's function are computationally very intensive, in particular, when combined with multiband approaches, such as the k . p methods. To reduce the cost and make the simulation of realistic devices tractable, we have developed a model order reduction method for the simulation of hole transport in silicon nanowires using three- and six-band k . p models. It is shown in this paper that, with a spurious band elimination process, the method can be readily extended to the eight-band case that enables us to simulate band-to-band tunneling devices. The method is demonstrated via constructing reduced models for indium arsenide (InAs) nanowires and simulation of I-V characteristics of InAs tunneling field-effect transistors (TFETs). The results indicate that significant model reduction can be achieved with good accuracy retained. The method is then applied to study InAs TFETs with different channel orientations and source-pocket TFETs with n-p-i-p doping profiles.

Simulations of nanoelectronic devices with nonequilibrium Green's function are computationally very intensive, in particular, when combined with multiband approaches, such as the k . p methods. To reduce the cost and make the simulation of realistic devices tractable, we have developed a model order reduction method for the simulation of hole transport in silicon nanowires using three- and six-band k . p models. It is shown in this paper that, with a spurious band elimination process, the method can be readily extended to the eight-band case that enables us to simulate band-to-band tunneling devices. The method is demonstrated via constructing reduced models for indium arsenide (InAs) nanowires and simulation of I-V characteristics of InAs tunneling field-effect transistors (TFETs). The results indicate that significant model reduction can be achieved with good accuracy retained. The method is then applied to study InAs TFETs with different channel orientations and source-pocket TFETs with n-p-i-p doping profiles.

Simulations of nanoelectronic devices with nonequilibrium Green's function are computationally very intensive, in particular, when combined with multiband approaches, such as the k . p methods. To reduce the cost and make the simulation of realistic devices tractable, we have developed a model order reduction method for the simulation of hole transport in silicon nanowires using three- and six-band k . p models. It is shown in this paper that, with a spurious band elimination process, the method can be readily extended to the eight-band case that enables us to simulate band-to-band tunneling devices. The method is demonstrated via constructing reduced models for indium arsenide (InAs) nanowires and simulation of I-V characteristics of InAs tunneling field-effect transistors (TFETs). The results indicate that significant model reduction can be achieved with good accuracy retained. The method is then applied to study InAs TFETs with different channel orientations and source-pocket TFETs with n-p-i-p doping profiles.

Simulations of nanoelectronic devices with nonequilibrium Green's function are computationally very intensive, in particular, when combined with multiband approaches, such as the k . p methods. To reduce the cost and make the simulation of realistic devices tractable, we have developed a model order reduction method for the simulation of hole transport in silicon nanowires using three- and six-band k . p models. It is shown in this paper that, with a spurious band elimination process, the method can be readily extended to the eight-band case that enables us to simulate band-to-band tunneling devices. The method is demonstrated via constructing reduced models for indium arsenide (InAs) nanowires and simulation of I-V characteristics of InAs tunneling field-effect transistors (TFETs). The results indicate that significant model reduction can be achieved with good accuracy retained. The method is then applied to study InAs TFETs with different channel orientations and source-pocket TFETs with n-p-i-p doping profiles.

Simulations of nanoelectronic devices with nonequilibrium Green's function are computationally very intensive, in particular, when combined with multiband approaches, such as the k . p methods. To reduce the cost and make the simulation of realistic devices tractable, we have developed a model order reduction method for the simulation of hole transport in silicon nanowires using three- and six-band k . p models. It is shown in this paper that, with a spurious band elimination process, the method can be readily extended to the eight-band case that enables us to simulate band-to-band tunneling devices. The method is demonstrated via constructing reduced models for indium arsenide (InAs) nanowires and simulation of I-V characteristics of InAs tunneling field-effect transistors (TFETs). The results indicate that significant model reduction can be achieved with good accuracy retained. The method is then applied to study InAs TFETs with different channel orientations and source-pocket TFETs with n-p-i-p doping profiles.

Simulations of nanoelectronic devices with nonequilibrium Green's function are computationally very intensive, in particular, when combined with multiband approaches, such as the k . p methods. To reduce the cost and make the simulation of realistic devices tractable, we have developed a model order reduction method for the simulation of hole transport in silicon nanowires using three- and six-band k . p models. It is shown in this paper that, with a spurious band elimination process, the method can be readily extended to the eight-band case that enables us to simulate band-to-band tunneling devices. The method is demonstrated via constructing reduced models for indium arsenide (InAs) nanowires and simulation of I-V characteristics of InAs tunneling field-effect transistors (TFETs). The results indicate that significant model reduction can be achieved with good accuracy retained. The method is then applied to study InAs TFETs with different channel orientations and source-pocket TFETs with n-p-i-p doping profiles.