Point defects directly impact solar cell device performance by limiting the carrier lifetime. In this work, density functional theory calculations are first used to determine the formation energy and diffusion energy barriers of dominant defects in Cu(In,Ga)Se-2. Next, continuum reaction-diffusion models are developed to analyze the redistribution of defects during manufacturing processes. We estimate defect capture cross sections using a first-principles-based approach. These cross sections are combined with our calculated defect profiles and trap energy levels to parameterize a Shockley-Read-Hall recombination model, which we implement into a device simulator to predict carrier lifetimes and device performance. In that way, a predictive technology computer aided design model is built to predict and optimize the performance of Cu(In,Ga)Se-2 solar cells.

To improve the performance of Cu(In,Ga)Se2 thin-film photovoltaic devices, a robust understanding of the dominant diffusion pathways of the alloy species In and Ga is needed. Here, the most probable defect complexes and mechanisms for In and Ga diffusion are identified with the aid of density functional theory. The binding energies and migration barriers for these complexes are calculated in bulk CuInSe2 and CuGaSe2. Analytic models and kinetic lattice Monte Carlo simulations are employed to predict the diffusivity of In and Ga under variations in composition and temperature. We find that a model based on coulombic interactions between group III antisites and vacancies on the Cu-sublattice produces results that match well with experiment.

Ytterbium-doped all-inorganic lead-halide perovskites (Yb3+:CsPb(Cl1-xBrx)(3)) generate near-infrared photoluminescence (PL) quantum yields exceeding 100% by quantum cutting. Experimental and computational studies have suggested complex dopant speciation in these materials arising from the formation of lattice defects needed to compensate for the excess charge of Yb3+ relative to Pb2+, but the relationship between quantum cutting and such speciation is still poorly understood. Here, we use cryogenic photoluminescence spectroscopy and density functional theory-assisted kinetic Monte Carlo simulations to investigate changes in Yb3+ speciation induced by anion (Cl- vs Br-) and trivalent-dopant (Yb3+ vs Gd3+) alloying in CsPb(Cl1-xBrx)(3) (0.00 <= x <= 1.00) perovskite nanocrystals (NCs). The experimental results reveal nonstatistical distributions of Yb-Cl and Yb-Br bonds in Yb3+:CsPb(Cl1-xBrx)(3) NCs. Monte Carlo simulations reproduce the experimental trends well and predict thermodynamic favorability for Yb3+ dopants to retain Cl- coordination, even in the presence of high lattice Br- concentrations. For a given lattice composition (e.g., CsPbCl3), low-temperature PL spectra reveal that the relative populations of three dominant Yb3+ species change substantially with Gd3+ codoping. These results further show that quantum cutting is largely insensitive to these differences in Yb3+ speciation, ruling out any "magic" configuration of Yb3+ ions and defects. These results are discussed in relation to the microscopic prerequisites for the concerted sensitization of two Yb3+ ions during quantum cutting. Overall, these findings highlight the mechanistic robustness of Yb3+:CsPb(Cl1-xBrx)(3) quantum cutting in lead-halide perovskites and provide deeper insight into the inner workings of this unique phenomenon.

CsPb(Cl1-xBrx)3(0 <= x <= 1) nanocrystals and thinfilms doped with a series oftrivalent rare-earth ions (RE3+=Y3+,La3+,Ce3+,Gd3+,Er3+,Lu3+) have been prepared andstudied using variable-temperature and time-resolved photoluminescence spectroscopies. Wedemonstrate that aliovalent (trivalent) doping of this type universally generates a new and often-emissive defect state ca. 50 meV inside the perovskite band gap, independent of the specificRE3+dopant identity or of the perovskite form (nanocrystals vs thinfilms). Chloride-to-bromideanion exchange is used to demonstrate that this near-band-edge photoluminescence shifts withchanging band-gap energy to remain just below the excitonic luminescence for all compositionsof CsPb(Cl1-xBrx)3(0 <= x <= 1). Computations show that this shift stems from the effect of the changing lattice dielectric constantson a shallow defect-bound exciton. Microscopic descriptions of this dopant-induced near-band-edge state and its relation toquantum cutting in Yb3+-doped CsPb(Cl1-xBrx)3are discussed.

Progress in the application of machine learning techniques to the prediction of solid-state and molecular materials properties has been greatly facilitated by the development state-of-the-art feature representations and novel deep learning architectures. A large class of atomic structure representations based on expansions of smoothed atomic densities have been shown to correspond to specific choices of basis sets in an abstract many-body Hilbert space. Concurrently, tensor network structures, conventionally the purview of quantum many-body physics and quantum information, have been successfully applied in supervised and unsupervised learning tasks in computer vision and natural language processing. In this work, we argue that architectures based on tensor networks are well-suited to machine learning on Hilbert-space representations of atomic structures. This is demonstrated on supervised learning tasks involving widely available datasets of density functional theory calculations of metal and semiconductor alloys. In particular, we show that certain standard tensor network topologies exhibit strong generalizability even on small training datasets while being parametrically efficient. We further relate this generalizability to the presence of complex entanglement in the trained tensor networks. We also discuss connections to learning with generalized structural kernels and related strategies for compressing large input feature spaces.

Ytterbium doping in all-inorganic lead-halide perovskites [CsPb(Cl1-xBrx)(3)] generates interesting properties including quantum cutting and narrow line emission, making these materials attractive spectral down converters for solar photovoltaics. The relationship between this optical efficiency and the defect structure(s) associated with Yb3+ dopants within perovskites is not well understood. Various charge-neutral doping motifs have previously been proposed and studied computationally, including clusters involving two substitutional Yb3+ ions charge compensated by a single local Pb2+ vacancy. Near-band-edge defect states associated with such motifs are believed to play an important mechanistic role in quantum cutting itself. Here, we report the results of x-ray absorption and x-ray total-scattering measurements on ytterbium-doped CsPbCl3. XANES shows that the dopant oxidation state is exclusively Yb3+, and a combination of Yb L-3 and Pb L-3 extended x-ray absorption fine structure (EXAFS) shows that this Yb3+ substitutes exclusively at Pb2+ sites, where it adopts a pseudo-octahedral [YbCl6](3-) coordination environment. Shell-by-shell fits to the data show a short Yb-Cl bond distance of 2.58 angstrom compared to the Pb-Cl bond distance of 2.83 angstrom. We confirm this finding by x-ray pair distribution function analysis, which also shows evidence of additional Pb2+ vacancy formation induced by Yb3+ doping. We evaluate whether this is the primary mechanism of charge compensation using simulated EXAFS and pair distribution function data for several computed defect structures. Together, these results resolve the local dopant structures and charge-compensation mechanisms in lanthanide-doped all-inorganic lead-halide perovskites, and, as such, significantly advance the understanding of structure-function relationships in this important class of materials.

Exceptionally high experimental photoluminescence quantum yields attributed to highly efficient quantum cutting have recently been observed in ytterbium-doped inorganic metal-halide perovskites such as Yb:CsPb(X = Cl, Br)(3). Combined with strong, tunable, broadband absorption in the visible spectrum, these materials show great promise for applications in solar down-converter technologies. Much subsequent work has been dedicated to uncovering the fundamental mechanisms behind Yb-mediated quantum cutting, and an accumulation of experimental evidence has shown that a particular speciation of Yb, believed to be a fully compensated (2Yb(Pb) + V-Pb)(0) defect complex, dominates this process. In this work, we investigate Yb defect formation in single-crystal CsPbCl3, and in particular the feasibility of forming (2Yb(Pb) + V-Pb)(0) defect complexes, using first-principles electronic structure calculations. A simple thermodynamic model based on defect formation energies, binding energies, and charge transition levels provides some insight into the distribution of YbPb substitutionals and Pb vacancies, underscoring the range of material compositions and synthesis conditions over which locally bound configurations of (2Yb(Pb) + V-Pb)(0) defect complexes become significant. We complement this analysis with additional calculations of structural and electronic properties and discuss the consistency of these results with the observed onset of quantum cutting.

We investigate the diffusion of copper in CuInSe2 using thermodynamic and kinetic models based on density functional theory calculations, attempting to reconcile large differences in reported experimental diffusivities. We find that observations of rapid chemical diffusion can be explained by large thermodynamic factors, which we calculate using a compositionally constrained model of intrinsic point defect formation. We further characterize how copper diffusion coefficients depend on material synthesis conditions and exhibit their variation across the CuInSe2 secondary phase diagram. In doing so, we identify stable off-stoichiometries that are dominated by either vacancy- or interstitial-mediated diffusion mechanisms. These results are employed in the development of a continuum reaction–diffusion model, which we use to simulate experimental depth profiles.

Point defects directly impact solar cell device performance by limiting the carrier lifetime. In this work, density functional theory calculations are first conducted to determine the formation energy and diffusion energy barriers of dominant defects in Cu(In,Ga)Se-2. Next, continuum models are developed to model the redistribution of defects during cooling and annealing processes. The calculated defect profiles as well as the corresponding capture cross sections and trap energy levels are implemented into device simulation via Shockley-Read-Hall (SRH) recombination models to calculate carrier lifetime and device performance. In that way, a predictive TCAD model is built to optimize the composition and performance of Cu(In,Ga)Se-2 solar cells.

Liquefied natural gas (LNG) is becoming increasingly popular as a marine fuel as emission regulations become more stringent. However, very little data are available on the particulate matter (PM) emissions of modern marine natural gas engines. In this study, we present a first detailed characterization of the composition of the PM emitted by a modern, in-use, natural-gas-powered vessel. The vessel engines use compression-ignition and only a small amount of diesel fuel as pilot. These engines drive electrical generators, providing propulsion as well as auxiliary power for the vessel. Our emissions characterization includes six different techniques to measure black carbon (BC), including all methods determined as appropriate for measuring BC emissions from ships by the International Maritime Organization, as well as particle size distributions, metal concentrations, and organic particulate emissions. PM emissions differed significantly between idle and at-sea operating conditions. At idle, PM emission factors were primarily organic (approximately 1500mg/kWh), with BC emission factors over two orders of magnitude lower (5.6±0.4mg/kWh). At engine loads above 25%, all emissions were independent of load and substantially lower than at idle, at 4.4±1.7mg/kWh for organics and 0.8±0.2mg/kWh for black carbon. When operated only on diesel fuel, this engine emitted 8-fold more organic PM (38±15mg/kWh) and 37-fold more BC (30±11mg/kWh) at loads above 25%. At idle loads, the diesel-fuel emissions were comparable to the natural-gas emissions. In addition to organics and BC, a third category of non-volatile sub-10-nm particles was identified. A detailed consideration of our measurements indicated that the sources of the organic, BC, and sub-10-nm particles were lubrication oil, diesel pilot fuel, and lubrication-oil metals, respectively. Future studies should seek to quantify the emissions of other dual-fuel engines that will be entering the market.

In-use exhaust stream CH4 emissions from two dual fuel marine engines were characterized and strategies for CH4 reduction were identified and evaluated. For this, a low-cost, portable, wavelength modulation spectroscopy (WMS) system was developed. The performance of the developed WMS sensor was assessed using gas standards and demonstrated on a heavy-duty, diesel pilot ignited, direct-injection natural gas research engine through comparison to a flame ionization detector. The WMS sensor was subsequently used to measure the exhaust-stream CH4 concentration from two diesel pilot-ignited, port-injected natural gas engines on a coastal vessel while under normal operation. Using cylinder deactivation to reduce the excess air ratio, λ, and vessel operation changes to minimize operation at lower loads, the total CH4 emission were reduced by up to 33%. The measured, load specific CH4 emissions were subsequently used to identify an improved vessel operation strategy, with an estimated 56-60% reduction in CH4 emissions. These results demonstrate the importance of considering the real-world engine operation profile for accurate estimates of the global warming potential, as well as the utility of a WMS sensor for characterizing and mitigating in-use CH4 emissions.

Manganese(II)-doped cesium-lead-chloride (Mn2+:CsPbCl3) perovskite nanocrystals have recently been developed as promising luminescent materials and attractive candidates for white-light generation. One approach to tuning the luminescence of these materials has involved anion exchange to incorporate Br-, but the effects of anion exchange on Mn2+ speciation in doped metal-halide perovskites is not well understood at a microscopic level. Here, we use a combination of X-band electron paramagnetic resonance (EPR) and photoluminescence spectroscopies to monitor the Mn2+ dopants in Mn2+:CsPbCl3 nanocrystals during Cl- -> Br(- )anion exchange. Analytical measurements show that the nanocrystals retain their Mn(2+ )over the course of Cl- -> Br- anion exchange and they continue to show strong Mn2+ d-d luminescence but, surprisingly, the Mn2+ EPR intensities all but vanish. Further results suggest that Mn2+ ions migrate during anion exchange to form clusters that are still luminescent but show no EPR signal due to antiferromagnetic superexchange coupling. Monte Carlo simulation and analysis of the Mn2+:CsPb(Cl1-xBrx)(3) lattice at various halide compositions (x) bolsters this interpretation by indicating a propensity for Mn2+-Cl(- )units to cluster as the Br(- )content increases, increasing the probability of the nearest-neighbor Mn2+-Mn2+ interactions. The driving force for this clustering is retention of the stronger Mn-Cl bonds compared to Mn-Br bonds. In addition, modeling predicts spinodal decomposition to form Mn2+-enriched domains even at the end point compositions of x = 0 and 1, with Mn(2+ )ordering in next-nearest-neighbor positions driven by Coulomb interactions and lattice-strain minimization. These results haVe important implications for both fundamental studies and applications of doped and alloyed metal-halide perovskites.

Mixed ionic-electronic conducting (MIEC) membrane reactors are attractive for partial oxidation and oxidative coupling of methane (OCM) because of their ability to separate oxygen from air with a low energetic penalty, and introduce the oxygen into a reaction zone with spatial and temporal control. To facilitate design and optimization of such reactors, two numerical models for MIEC membrane reactors for methane reforming are introduced: a computationally inexpensive, coupled, two-chamber CSTR model, and a CFD model for reacting flow. The CFD model considers diffusive transport with a mixture averaged model, and was shown to agree well with previous experimental and modelling works. Both models utilize a detailed gas-phase chemical kinetic mechanism and a membrane oxygen permeation model that considers the local oxygen concentration. It is demonstrated that the CSTR model can be used to evaluate a large number of reactor parameters efficiently to identify optimal conditions for OCM. The most promising reactor operating points identified using the CSTR for a particular reactor size are further considered with spatial detail using the CFD model. It is shown that the coupling between oxygen permeation, gas phase chemical kinetics, and flow field can have a significant impact on the reactor performance. Here the flow direction in a typical button reactor was reversed, which resulted in a 40% increase in predicted C-2 yield. The increase is attributed to the shorter post-reaction residence time of the C-2 products in the reactor, thereby inhibiting deeper oxidation. This significant impact on the reactor performance can be utilized to improve the yield of partially oxidized products from membrane reactors. (C) 2018 Elsevier Ltd. All rights reserved.

We construct an analytic model for the composition dependence of the vacancy-mediated Cu diffusion coefficient in undoped CuInSe2 using parameters from density functional theory. The applicability of this model is supported numerically with kinetic lattice Monte Carlo and Onsager transport tensors. We discuss how this model relates to experimental measurements of Cu diffusion, arguing that our results can account for significant contributions to the bulk diffusion of Cu tracers in non-stoichiometric CuInSe2.

We extend a canonical-thermodynamic method for computing intrinsic point defect concentrations as a function of chemical stoichiometry and temperature to include extrinsic defect contributions, while applying this method to study defects in CuInSe 2 . This method relies on a large set of defect formation energies calculated from first principles, which require corrections for known errors arising from spurious interactions. Guided by recent experimental work exhibiting the complex interplay between Na, K, and Cd incorporation, we examine the behavior of dominant defects as the material composition varies under experimentally relevant conditions. In addition to identifying the regions of composition-parameter space relevant to the incorporation of impurities Na, K, and Cd, and stable against the formation of secondary compounds, we also study defect kinetics susceptible to the presence of these impurities. From this analysis, we propose a simple model for the enhanced diffusion of Cu-vacancies mediated by K, which could lead to the enhanced incorporation of Cd observed experimentally.

Recent experimental work has revealed the distinct and beneficial role of K incorporation on the fabrication of increasingly efficient thin-film photovoltaic devices with Cu(In,Ga)Se2 (CIGS) absorber layers. This has been attributed, in part, to improved CdS/CIGS heterojunction quality due to the enhanced diffusion of Cd into the near-interface region of CIGS. In this work, we try to distinguish the role of K compared to Na in enhancing Cd incorporation in CuInSe2 (CIS) based on first-principles calculations. Using a canonical method for calculating defect concentrations as a function of temperature and material stoichiometry, we identify experimentally relevant conditions under which a simple model for Na and K kinetics can lead to such an effect. We argue that a sufficiently low migration barrier for K diffusion mediated by Cu vacancies can lead to Cu-depletion near the CdS interface, allowing Cd to occupy greater numbers of vacant Cu sites.

Energetically favorable configurations of defect structures in CIGS are studied with a Metropolis-Hastings lattice Monte Carlo (MC) method. Empirical models or the binding energy of arbitrary CIGS configurations are tested and fit to ab-initio data. The MC simulations yield energetically favorable crystal structures for the optimized binding energy model. These structures reveal temperature-dependent phases of indium-gallium segregation and copper-vacancy segregation. These results are then used in kinetic lattice Monte Carlo (KLMC) simulations to study cation diffusion.