DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Design and fabrication of the teepee photonic crystal for high-efficiency thin film solar cell architectures
No full textAugust 2022School of ScienceThe goal of this thesis is to provide the fabrication foundation and analysis of a photonic crystal-based tunnel oxide passivated contact (TOPCon) solar cell. A theoretical framework will be provided that studies the light interaction with photonic crystals and discusses funda- mental effects associated with photonic crystals such as parallel-to-interface refraction, slow light modes and enhanced light trapping. Additionally, a review of solar cell theory will be given, reviewing semiconductor properties, the generation and recombination of carriers, p-n junction diodes and solar cell efficiency and its limits.The fabrication procedure for the photonic crystal, referred to as the teepee photonic crystal, and the integrated TOPCon solar cell will be discussed, including in-depth discussion of reactive ion etching, the formation of tunnel oxide passivated contacts and each of their challenges. Next, the methods used for the analysis of optical properties of two unique photonic crystals, the teepee and inverted pyramid photonic crystals, and the electronic properties of the teepee photonic crystal TOPCon solar cell will be provided.
In our analysis, we demonstrate that the teepee photonic crystal exceeds the Lamber- tian limit, a fundamental upper limit of solar absorption based on statistical ray trapping. When fabricated on a 10 μm silicon-on-insulator substrate, we show that the teepee photonic crystal exceeds this limit for angles of incidence between 0o and 60o over a majority of the AM1.5 Global Solar Spectrum.
Finally, we present a teepee photonic crystal TOPCon solar cell that demonstrates an efficiency of 11.7%. The low efficiency is largely due to the presence of a blistered passivated contact region that degrades the passivating quality of our device resulting in very low open- circuit voltages. Alternate methods of fabrication to prevent blistering are presented, laying the ground work for future fabrication.Ph
Synchrophasor-based monitoring, control, and protection for distributed energy resources
No full textDecember 2021School of EngineeringThe proliferation of distributed energy resources (DERs) are transforming the landscape of today's electrical grid. Non-conventional DERs such as photovoltaic systems, Battery Energy Storage Systems (BESS), and wind turbines exhibit strikingly different dynamics when compared to traditional energy sources driven by synchronous generators, due to their power electronic interfaces with the grid. Thus, a modern electrical grid that utilizes these new energy resources, would require a reconfigurable, fast, accurate, and time critical monitoring, protection and control framework in order to operate harmoniously and resiliently. Synchrophasor technology and Phasor Measurement Units (PMUs) provide a unique, time-critical, accurate, reliable and standardized framework for power system measurements under these new energy resources.
This thesis proposes a synchrophasor-based monitoring, management, protection, and control architecture for power systems with DERs. The usage of PMUs and synchrophasor technology is appropriate for such applications, because they provide accurately time-stamped measurements for power network management functions such as monitoring and control. This approach proposes substantial benefits as compared to conventional supervisory control and data acquisition (SCADA) systems in terms of accuracy, reliability and speed. Because, it is logistically impossible to operate and test the proposed architecture in a real-life power system, real-time hardware-in-the-loop (HIL) simulators are used to mimic the behavior of a power system, containing multiple DERs such as photovoltaic systems, BESSs, and diesel generators. During this process, different types of real-time hardware-in-the-loop (HIL) simulator hardware were explored and their performances were compared. The proposed synchrophasor-based network management and control infrastructure uses deterministic real-time embedded systems with Field Programmable Gate Arrays (FPGA) for data acquisition and timing-management, and real-time processors are utilized for networking and inter-communication purposes. To test the resiliency of the proposed architecture, additional network traffic generator hardware were connected to the communication network that houses the controller and the PMUs. Through the course of experimentation, multi-platform homogeneous real-time simulation models were developed for photovoltaic cells and Li-ion batteries. In the domains of power system protection, controller-hardware-in-the-loop (CHIL) experiments were performed to investigate whether traditional protection methods are suitable for the integration of inverter based DERs. These experiments revealed that unlike traditional plants with synchronous generators, grounding transformers are not required in order to protect inverter based DERs from ground-fault overvoltages (GFOV) under single-line-to-ground (SLG) faults. However, most utilities are still reluctant to abandon the inclusion of grounding transformers. Thus, this observation has the potential to reduce the cost for planning and implementing DER based substations significantly, if adopted by the operators. To perform all these experiments seamlessly within a real-time digital power system simulation ecosystem, a laboratory architecture featuring networking, timing and electrical functionalities was standardized and experiments were performed on it. One such key experiment featured the design and implementation of a synchrophasor synchronization gateway and controller (SSGC) hardware. This hardware has the functionalities to parse multiple synchrophasor streams in real-time, compute and monitor their respective network delays, and provide supplementary control functionalities based on the information retrieved from those synchrophasor streams. This hardware was utilized to monitor and control a microgrid (containing BESS and PV systems) running on a real-time simulator. To test the resilience of this proposed hardware, its communication network was tampered with external hardware, and its performance was analyzed under such conditions.Ph
Data-driven stochastic modeling of guided wave propagation and robust damage diagnosis
No full textMay 2022School of EngineeringModern day civil, mechanical, and aeronautical structures are transitioning towards a continuous, online, and automated maintenance paradigm in order to ensure increased safety and reliability. The field of structural health monitoring (SHM), which is concerned with online damage detection, localization and quantification, is playing a key role in this respect, and a significant amount of research efforts have been directed towards achieving this maintenance paradigm. Active sensing acousto-ultrasound guided-wave based SHM techniques have shown great promise due to their potential sensitivity to small damages. However, the methods' robustness and diagnosis capability become limited in the presence of environmental and operational variability such as a change in surrounding temperature, different load and boundary conditions, variation in material properties, etc. In addition, the currently used techniques rely on deterministic damage diagnosis schemes rather than probabilistic frameworks, which can account for uncertainty arising from different sources. As such, it is critical to model guided wave propagation in the presence of varying external sources, environments, operating conditions, and material property variations that impose uncertainty on guided wave propagation in order to enable the formulation of a robust, reliable, and probabilistic damage diagnostic scheme.
In order to achieve this goal, in this report, a novel stochastic time series based framework was adopted to model guided wave propagation. Different stochastic time-varying time series models, such as Recursive Maximum Likelihood Time-varying Auto-Regressive (RML-TAR) and Functional Series Time-varying Auto-Regressive (FS-TAR) models, and stationary Functionally Pooled (FP) time-series models were put forward to model and capture the uncertainty in guided wave propagation under varying temperature, loads, as well as material property variations based on experimental data. In order to incorporate information from a physics perspective, high-fidelity finite element (FE) models were also established to model the effect of temperature and material property variation on guided wave propagation. The effect of broadband high frequency actuation on guided wave propagation under different temperature was studied with the help of novel Functionally-Pooled Auto-Regressive (FP-AR) models. Finally, surrogate models were formulated through the use of stochastic time-dependent RML-TARX models and compared with the FE models under varying temperatures. The advantages of using surrogate models will be manifested in the future work that has been proposed in this report with the ultimate aim of formulating a probabilistic SHM framework.
Once the modeling part is complete, stochastic time series models are invoked to formulate damage diagnosis algorithm. At first, stochastic stationary time series models such as autoregressive models (AR) are used. In addition to using standard AR-based approach, where all the model parameters are used for formulating a statistical characteristic quantity, two other approaches are also introduced, namely: singular value decomposition (SVD) and principal component analysis (PCA)-based method. The performance of these three methods are analyzed and assessed in detail for damage detection and identification in the aluminum as well as the composite plates. It is shown that the AR-based approach works well for aluminum plates but shows poor performance for composite plates in damage identification.
As guided wave signals are non-stationary in nature, it is more appropriate to use non-stationary modeling techniques for damage detection and identification. An important class of parametric methods for the effective solution of the modeling of non-stationary problems is based on functional series time-dependent autoregressive moving average (FS-TARMA) models. These models have parameters that explicitly depend upon time, with the dependence described by deterministic functions belonging to specific functional subspaces. The advantages of FS-TARMA models involve improved accuracy, improved tracking of the time-varying dynamics, increased predictive ability, and representation parsimony. In this study, FS-TAR models, which is a subclass of FS-TARMA models, are used for damage detection and identification using non-stationary guided wave signals.
When formulating damage detection and identification based algorithms, constant coefficients of projection as well as the time-varying parameters are used. Three types of functional basis functions are investigated, namely: wavelet, Chebyshev and trigonometric basis functions and their performance in terms of damage detection and identification are analyzed and assessed.
Lastly, the idea of using a high-frequency broadband white noise actuation instead of a tone-burst actuation for exciting the guided wave signal is investigated. The hypothesis is that the broadband actuation will excite additional vibratory modes of the structure, which are not realized from a traditional tone-burst actuation. This new type of actuating the structure may help formulate a more robust damage diagnosis scheme.Ph
An information theoretic approach to graph representations, graph embedding and embedding evaluation
No full textMay 2022School of ScienceFor centuries, mineralogists have sought means of classifying the now more than 5500 mineral species based on some physical and chemical attributes. Collectively, these studies have been constrained by tabular representations with a relatively small number of attribute columns. In recent efforts Earth science researchers have been incorporating mineral co-occurrence data and studying patterns of coexisting mineral species in order to understand the evolution of minerals and their interaction with the environment. Paleontologists rely on the fossil record to learn about the development and behaviors of various species of flora and fauna throughout geologic time. Fossil data is used to reconstruct taxonomical maps and hierarchies describing the relationships between species based on evolutionary and environmental criteria. Studying these relationships is key to learning about the evolution of life and its effect on the planet. Underlying each of these fundamental systems in our natural world has been the constraint of low-dimensional tabular representations and thus mostly descriptive analyses. This constraint has been a barrier to harnessing the power of machine learning as a vehicle for predictive analysis of data from these systems. In this thesis, motivated by such problems as well as others in human systems, we utilized graph-based representations of extant data, and developed and applied graph entropy methods to quantify structural information content of graphs to improve upon existing graph embedding methods and introduce an new evaluation method for graph embeddings. Many natural and human systems are modeled well as networks of interacting entities and mathematically represented by graphs which opens the door to the application of quantitative methods to extract useful information from these graphs. This is a challenging task however, given the highly dimensional nature of graphs. Graph embedding methods aim to learn concise vector representations that accurately preserve the graph structure and are often built around a specific representation of the graph. The performance of these methods varies greatly depending on the characteristics of the input graph and the most commonly utilized approach to evaluating this performance is the training and testing of downstream predictive models. Both of these aspects of the embedding process introduce complexity and increase uncertainty, further exacerbating the problem and making it difficult for researchers to make decisions while designing and deploying analysis pipelines. Graph embedding methods operate on high dimensional graph representations such as graph matrices or sets of sampled graph paths to obtain low dimensional vector embeddings by applying various methods including matrix factorization and deep learning. A limitation to current methods is that each is built around one specific graph representation and thus the outcome is sensitive to that one perspective. Graphs are complex structures that can be viewed from several perspectives and are best described through multiple complementary structures. A graph adjacency matrix for example represents distances between nodes hinting at the strength of relationships between entities while a node degree vector captures the connectedness of each node offering potential insight into its role within the graph. Both perspectives provide useful information for downstream analysis tasks such as node classification and link prediction. The main challenge is finding a suitable combination of graph representation, embedding method and machine learning model to maximize predictive performance. In this thesis, we have developed a method to incorporate graph structural information content into the graph embedding process, using graph entropy measures. We begin by computing structural information content of a graph using six methods and then selectively combine the value vectors with the adjacency matrix producing novel graph representations. Every graph information method we have used, captures structural information based on a different graph element and the information functions utilized by these methods can be extended to use new graph measures and metrics. We have also developed a method to obtain vector embeddings from the new structural information matrix using a deep neural network. Deep neural networks are capable of capturing non-linear structure and thus are suitable in modeling complex graph structure. We designed an autoencoder deep neural network to learn low-dimensional vector embeddings from the structural information matrices described above. We utilized labeled graph datasets and an array of machine learning predictive models to show that our embedding method results in improvements upon the predictive accuracy achieved by existing embedding approaches. Finally we have developed a method to evaluate the suitability of graph embeddings by measuring the loss in information between the original graph and a reconstructed graph. We use the previously mentioned graph entropy measures to quantify and compare this loss in information between different embedding methods. The overall goal of this thesis is to introduce an information theoretic approach to graph representation, embedding and embedding evaluation, in order to improve the predictive performance of downstream machine learning models trained using these representations. We have constructed a software workflow to process graph data and execute all the steps needed to compute structural information content, obtain graph embeddings and then train and test machine learning predictive models. After examining the results of the experiments we have identified factors in the input representations that affect the performance of the embedding methods and thus reveal the boundaries of the usefulness of information content based graph representations. We also discuss the implications of the variability in predictive accuracy between combinations of graph embedding methods and machine learning models. Finally we introduce a novel method of evaluating embedding fitness based on graph reconstruction and graph entropy methods.Ph
End-to-end question answering on semi-structured tabular data
No full textMay 2022School of ScienceSemi-structured tables are commonly seen in everyday life as one of the most popular and convenient ways to store and organize data. They widely appear in open-domain digital documents such as PDFs, web pages, knowledge bases (KBs), i.e., Wikipedia, and domain-specific documents such as scientific papers, journals, and enterprise reports. Recently, semi-structured tables have been recognized as a rich knowledge source for the Question Answering (QA) tasks. Unlike relational databases tables, it is more challenging for machines to automatically understand semi-structured tables and use them in downstream tasks. Despite the existing effort, computational approaches still suffer for the following reasons: 1. the abundance of semi-structured tables, 2. lack of an explicit schema, and 3. complex and flexible table structures. Besides, the annotation for Table QA datasets is labor and time-intensive, especially for domain-specific, large-size tables corpus. Due to the complexity of Table QA, existing works tried to crack the problem as sub-tasks, i.e., table retrieval and QA over tables. Given a natural language query or question, table retrieval studies how to locate the table containing the correct answer from the table corpus, while the QA over tables task focuses on finding table cells from a given table to answer the questions. The traditional two-step pipeline has its limitation on performance, mainly due to error propagation. In this thesis, we aim to fill in the vacancy in the research of end-to-end Table QA, leveraging the transformer-based models with the support of semantic-driven approaches. More specifically, with any natural language question, our goal is to design models that can efficiently search through a massive table corpus, retrieve the table containing the correct answer, and finally locate the correct answer to the given questions from the table. This thesis covers a series of supervised Table QA models as well as a brief discussion on unsupervised solutions. We first focus on providing sophisticated solutions to the QA over tables task. While the existing models highly rely on specialized pre-training techniques, we introduce the RCI model, which utilizes an existing language model to build connections between questions and table components. The RCI model locates the correct cells as the intersection of table rows and columns by capturing the row and column semantics with transformer-based architectures. With the RCI model producing the state-of-the-art results in QA over tables, we further extend the RCI architecture to an end-to-end Table QA pipeline called Cell Level Table Retrieval (CLTR). This model consists of two components: 1. a retriever model integrates traditional information retrieval methods with RCI to identify relevant tables; 2. a reader model identifies the correct table cells as answers to the questions using the RCI architecture. To enhance the accuracy and simplify the training of the end-to-end Table QA model, we investigate Dense Passage Retrieval (DPR) and Retrieval-Augmented Generation (RAG). We propose the T-RAG model, which unifies the traditional [retriever + reader] pipeline with a single training step. The T-RAG model holds the current best scores for the end-to-end Table QA and the table retrieval tasks. Finally, we explore unsupervised, neuro-symbolic approaches for Table QA. The lexical- and semantic-driven methods are applied to identify the correct table rows and columns to answer natural language questions without the benefit of supervision and training data.Ph
Application of the phase field model towards selective laser melting of Inconel 718, through an integrated multiscale framework
No full textDecember 2021School of EngineeringThe Selective Laser Melting (SLM) process is a developing additive manufacturing process for metals. This technique allows for very fine details to be produced with minimal post-processing required, and is especially useful for creating parts from traditionally hard-to-manufacture materials, such as Inconel 718. For this technique to mature, a link between the thermal conditions within the SLM process and the final microstructure must be established. In our work, we have used a thermal field for a set of known print parameters from the literature, provided to us by collaborators, to drive the evolution of a multicomponent phase field model approximating that of Inconel 718. Using the frozen gradient approximation, the solidification of γ-FCC dendrites in the liquid melt is observed, and values for primary dendrite arm spacing (PDAS) and composition are computed. In addition, the potential for secondary phase formation is evaluated through the measurement of the free energy of formation of secondary phases at characteristic points in the simulation. Finally, we develop an analytic model, taking certain approximations, demonstrating that as thermal gradient increases, a larger range of dendrite widths can exist at steady state. We also show how grain orientation plays a role in selecting the observed PDAS, as dendrite spacing will tend to be determined by the characteristic dendrite width formed through the tertiary arm branching process. Our work demonstrates that multicomponent phase field simulations of the primary solidification of Inconel 718 in the SLM process are feasible to conduct, and that with future work to improve the quantitative accuracy of the model, increasingly accurate results can be obtained from simulation. As this model is composition-ambivalent, this technique can lead to the the production of new, SLM-optimized alloys, testing their performance in silico before they are ever manufactured.Ph
A modified method for measurement in reverberation rooms
No full textAugust 2021School of ArchitectureUnwanted noise within the built environment present a growing number of concerns for building users. Sound absorbers are frequently utilized to address these concerns. However, the characterization of their absorption coefficients continue to demonstrate challenges for high accuracy and reliability as practice requires. Recently, experimental analysis has demonstrated that the assumptions regarding the diffusivity of sound fields remain unfulfilled. Specifically, chamber-based measurement methods presume sound intensities within reverberation chambers to be isotropic, or diffuse. Diffusion equation models (DEM) have clearly shown the anisotropy of energy flows within these chamber-based measurements, especially when highly absorptive materials under test are present. This phenomenon has attracted the attention of members of the acoustical community who believe this to be the cause for well-documented inconsistencies reported by chamber-based measurement laboratories across the world. DEM offers a viable and efficient method for increasing the reliability of chamber-based measurements through the prediction of sound energy flows with reverberation chambers.M
Physical modeling for LEAP-2017-2020 : assessment of liquefaction hazards for sloping ground and retaining sheet-pile quay walls
No full textAugust 2020School of EngineeringSeismically induced soil liquefaction can occur in saturated cohesionless soils, as a result of an earthquake and leads to significant or total loss of the soils strength. Soil conditions able to trigger liquefaction being rather common globally combined with the catastrophic consequences on buildings and infrastructure render the phenomenon one of the most severe earthquake-related hazards. Soil liquefaction is associated with failures in soil systems, such as slopes and embank- ments, and soil-structure systems, such as retaining walls and foundations, leading to large permanent soil deformations and extensive displacements of the structures, sometimes up to complete collapse. The mechanisms behind these failures have been a subject of research for many a number of decades. Geotechnical centrifuge testing and numerical tools have been employed in order to simulate the phenomenon and decipher the mechanisms leading to potential failure. Notwithstanding the valuable outcomes from these efforts, the challenge remained, since inconsistencies in the experimental results among different centrifuge facilities led in some cases to ambiguous conclusions and variability in the numerical results restricted the confidence in these tools to accurately predict the response. In 2013 the Liquefaction Experiments and Analyses Projects (LEAP) was undertaken by six centrifuge facilities (UC Davis, RPI, Cambridge University, Kyoto University, Na- tional Central University of Taiwan, and Zhejiang University), in an effort to investigate the response of a liquefiable sloping deposit. During that phase, in 2015 two centrifuge tests were performed at RPI, achieving repeatability with high fidelity. Four years later, LEAP-2017 built and expanded on the previous phase, in order to examine the repeatability and reproducibility potential of centrifuge testing in soil liquefaction and to investigate the effect of the variation of specific testing parameters on the system response. To that end the same model tests were repeated in nine centrifuge facilities in China, France, Japan, Korea, Taiwan, UK and USA. As part of the LEAP-2017 experimental campaign, three geotechnical centrifuge experiments were conducted at RPI, investigating the repeatability and reproducibility of centrifuge testing in soil liquefaction and shedding light on the sensitivity of the experimental results when varying the relative density or the input motion. The three models simulated a 5-degree sloping deposit, they were built consistently and employed consistent instrumen- tation. The reference test, RPI01, was built with 65% soil relative density and successfully reproduced the experimental results from 2015. The effect on the response of an additional high frequency component in the input motion was investigated in RPI02, which was also built with 65% soil relative density. The non-mono-frequency input motion led to lower dila- tive soil response and consequently to slightly higher lateral spreading. The effect of lower soil relative density on the response was examined in RPI03, which was built at 45% soil relative density and was subjected to the same input motion as RPI01. The looser deposit was found to be significantly more susceptible to liquefaction leading to higher permanent surficial displacements. In 2018 the LEAP-Asia investigated the effect of the generalized scaling law on the response of the liquefiable sloping deposit examined in the previous exercises. As part of this exercise, one experiment was performed at RPI (Model B) and was compared against RPI01 (Model A). The response of the model designed observing the generalized scaling law was in remarkable agreement with RPI01, prior to liquefaction. After the onset of liquefaction, Model B exhibited higher propensity for liquefaction and accumulated higher lateral surficial soil deformations. Soil-structure interaction (SSI) was introduced in the investigation of soil liquefaction in the LEAP-2020. The examined experimental set-up simulated a 3-m deep excavation of a saturated deposit of 5m total depth. The excavation was retained by a rigid floating sheet- pile quay wall, which was embedded in very dense sand. Six experiments were performed at RPI as part of the experimental campaign for LEAP-2020. RPI05 and RPI06 were built with 65% soil relative density and showed successful repeatability of the experimental results in the presence of SSI. The sensitivity of the experimental results in the initial orientation of the sheet-pile was investigated in RPI07. The installation of the sheet-pile with outward 2-degree rotation introduced a bias in its response leading to rapid accumulation of seaward displacements. The effect of relative density on the systems response was examined in RPI08 (built with Dr = 55%) and RPI09 (built with Dr = 75%), showing dramatic increase in the displacements of the soil and the sheet-pile in the case of the loose deposit. The response of the dense deposit was similar to that of RPI06, exhibiting slightly lower values of soil and the sheet-pile displacements. Last but not least, the effect on response of the system by an additional high frequency component in the input motion was investigated in RPI10. In consistency with the results from the slope, the non-mono-frequency input motion led to lower dilative soil response but to slightly lower lateral spreading, in view of the SSI effects.Ph
ITWS Capstone: Engineering a Semantic Web (Fall 2022)
No full textGuest lecture for ITWS Capstone, Rensselaer Polytechnic Institute, Troy, NY (Fall 2022
Epitaxial growth and properties of transition metal carbide thin films
No full textDecember 2021School of EngineeringTransition metal carbides and nitrides are of broad interest due to their thermal stability, chemical inertness, metallic to semiconducting conductivity, high hardness and wear resistance. In the past decades, transition metal carbides have been widely used in a wide range of technological applications including wear-resistant and decorative coatings, components for cutting and drilling tools, electrical contacts and diffusion barriers in microelectronics, and electrodes and catalysts for electrochemical storage and conversion systems. However, compared to numerous studies of transition metal nitrides on its intrinsic electrical, optical, optoelectrical, piezoelectrical, electrochemical and mechanical properties, much less is known for its transition metal carbide counterparts. To develop more fundamental understandings of transition metal carbide, specifically the early transition metal carbides, I have studied four representative transition metal carbide and carbonitride systems including WC, MoC, TiC and TiCN.Rocksalt WCy(001) layers are deposited onto MgO(001) single crystal substrates using DC magnetron sputtering method. A CH4 fraction of fCH4 = 0.4% - 6%, yields total C-to-W ratios x = 0.57 - 1.25. The C-to-W ratio y in the cubic WCy phase is smaller than x, ranging from y = 0.47 to 0.68, as determined from lattice constant ao measurements in combination with first-principles calculations that predict an increasing ao = (0.4053 + 0.0295y) nm for y = 0.3-1.0. This suggests that the cubic phase is stabilized by carbon vacancies and that the layers contain amorphous C with a volume fraction increasing from 4% - 26% for fCH4 = 0.4% - 6%. Structural analyses confirm the growth of epitaxial rock-salt structure WCy(001) layers with a cube-on-cube epitaxial relationship with the substrate: (001)WC(001)MgO and [100]WC[100]MgO. The measured XRD out-of-plane coherence length of 8 – 14 nm is nearly independent of the film thickness d = 10 or 600 nm, suggesting that growth beyond d = 10 nm leads to an epitaxial breakdown and the nucleation of misoriented hexagonal or orthorhombic W2C grains for fCH4 ≤ 1% and cubic nanocrystalline WCy grains for fCH4 > 1%.
Molybdenum carbide, which are formed by the same group transition metal as the tungsten carbide, are also investigated in this study. Molybdenum carbide layers are grown on Al2O3(0001) substrates with a varying CH4 fraction fCH4 = 0-10%. Structural analyses reveal that fCH4 = 7-8% leads to epitaxial -MoCy(111) grains with ["11" "2" ̅]-MoC["11" "2" ̅"0" ]Al2O3 and biaxial textured -Mo2C(0001) with a preferential ["10" "1" ̅"0" ]-Mo2C["10" "1" ̅"0" ]Al2O3 in-plane orientation. The two phases nucleate epitaxially on the substrate and/or on top of each other, followed by a competitive growth mode which results in a dominant cubic -MoCy(111) or hexagonal -Mo2C(0001) phase at fCH4 = 7 or 8%, respectively, and a reduction in the layer density measured by x-ray reflectivity which suggests the formation of amorphous C clusters above the layer nucleation stage. Deposition at lower fCH4 ≤ 6% leads to polycrystalline -Mo2C and/or bcc Mo phases, while higher fCH4 ≥ 10% yields nanocrystalline -MoCy embedded in an amorphous C matrix. The increase in fCH4 also causes a 3-fold decrease in the Mo deposition rate measured by Rutherford backscattering spectrometry and an 18% increase in the discharge voltage which is attributed to adsorbed CH4¬ and carbide formation on the target surface.
Titanium carbide are deposited onto MgO(001) by reactive DC magnetron sputtering in Ar/CH4 mixtures at 1100 C using a varying CH4 fraction fCH4 = 0.4-8% that yields C-to-Ti ratios x = 0.08-1.8. Structural analyses indicate epitaxial TiCx(001) growth for x = 0.08-1.5, but incorporation of secondary Ti and C impurity phases for x ≤ 0.24 and x 1.5, respectively. First-principles calculations of the formation energy of Ti1-yCy confirms the phase separation of hcp Ti and titanium carbide at low carbon contents. The relaxed lattice constant increases from 0.4304 nm to 0.4325 nm and the measured strain decreases from ε = 0.3% to 0.1 % for phase pure epitaxial TiC0.5 and TiC1.0 layers. Nanomechanical properties include hardness H and elastic modulus E shows a monotonically increasing trend with increasing carbon incorporation from H = 8.7 GPa and E = 143 GPa for TiC0.08 to H =31.2 GPa and E = 462 GPa for TiC1.0, while further increasing x leads to a rapid drop till H = 13.5 GPa and E = 201 GPa for TiC1.8. The measured resistivities at 298 K range from 83-598 cm for TiCx layers with the lowest resistivity of 83 cm for near-stoichiometric epitaxial TiC1.0 layer. The resistivities at 77 K are only 10-26 cm lower than at 298 K, indicating defect scattering dominance.
The electrical and nanomechanical properties of titanium carbide can be tuned by alloying with titanium nitride, which has a 6-fold decrease in the electrical resistivities and an fine regulation of valence electron concentration. For this purpose, titanium carbonitride films are sputter-deposited onto MgO(001) in Ar/CH4/N2 mixtures at 1100 C using a varying N2 partial pressure pN2 = 0.2-0.8 mTorr. The compositional analysis yields Ti0.44C0.39N0.17¬, Ti0.44C0.3N0.26 and Ti0.47C0.21N0.32 for pN2 = 0.3, 0.6 and 0.8 mTorr respectively. X-ray diffraction results indicate epitaxial TiCN(001) growth. The relaxed lattice constant decreases from 0.4296 nm to 0.4266 nm from pN2 = 0.2 to 0.8 mTorr, indicating N replacement of C into the carbonitride lattices. Nanomechanical properties include hardness H and elastic modulus E shows an initial increasing trend with increasing nitrogen incorporation from H = 26 GPa and E = 299 GPa for pN2 = 0.2 mTorr to H =28.3 GPa and E = 495 GPa for pN2 = 0.3 mTorr, while further increasing pN2 drops till H = 25.6 GPa and E = 456 GPa for pN2 = 0.8 mTorr. The measured resistivities at 298 K shows monotonically decreasing trend from 40 to 26 cm as pN2 increasing from 0.2 to 0.8 mTorr.Ph