Search

Backdoor attacks change a small portion of training data by introducing hand-crafted triggers and rewiring the corresponding labels towards a desired target class. Training on such data injects a backdoor which causes malicious inference in selected test samples. Most defenses mitigate such attacks through various modifications of the discriminative learning procedure. In contrast, this paper explores an approach based on generative modelling of per-class distributions in a self-supervised representation space. Interestingly, these representations get either preserved or heavily disturbed under recent backdoor attacks. In both cases, we find that per-class generative models allow to detect poisoned data and cleanse the dataset. Experiments show that training on cleansed dataset greatly reduces the attack success rate and retains the accuracy on benign inputs., Comment: Accepted to BMVC 2024

Spin- and charge-lattice interactions are potential key factors in the microscopic mechanism of high-temperature superconductivity in cuprates. Although both interactions can dramatically shape the low-energy electronic structure, their phenomenological roles in superconductivity are usually investigated independently. Employing angle-resolved photoemission spectroscopy, we reveal the spectroscopic fingerprint of short-range antiferromagnetic order in conjunction with enhanced electron-phonon interaction in the electron-doped cuprate superconductor $\mathrm{Nd_{1.85}Ce_{0.15}CuO_4}$. The observed mode coupling exhibits a strong momentum dependence that is in striking contrast to the node-antinode dichotomy previously observed in the hole-doped cuprates. Our results reveal an intimate relationship between electron-phonon coupling and antiferromagnetic fluctuations, which collectively sets the stage for unconventional superconductivity in the electron-doped cuprates.

Large Language Models (LLMs) have demonstrated remarkable capabilities across a variety of software engineering and coding tasks. However, their application in the domain of code and compiler optimization remains underexplored. Training LLMs is resource-intensive, requiring substantial GPU hours and extensive data collection, which can be prohibitive. To address this gap, we introduce Meta Large Language Model Compiler (LLM Compiler), a suite of robust, openly available, pre-trained models specifically designed for code optimization tasks. Built on the foundation of Code Llama, LLM Compiler enhances the understanding of compiler intermediate representations (IRs), assembly language, and optimization techniques. The model has been trained on a vast corpus of 546 billion tokens of LLVM-IR and assembly code and has undergone instruction fine-tuning to interpret compiler behavior. LLM Compiler is released under a bespoke commercial license to allow wide reuse and is available in two sizes: 7 billion and 13 billion parameters. We also present fine-tuned versions of the model, demonstrating its enhanced capabilities in optimizing code size and disassembling from x86_64 and ARM assembly back into LLVM-IR. These achieve 77% of the optimising potential of an autotuning search, and 45% disassembly round trip (14% exact match). This release aims to provide a scalable, cost-effective foundation for further research and development in compiler optimization by both academic researchers and industry practitioners.

Various iterative eigenvalue solvers have been developed to compute parts of the spectrum for a large sparse matrix, including the power method, Krylov subspace methods, contour integral methods, and preconditioned solvers such as the so called LOBPCG method. All of these solvers rely on random matrices to determine, e.g., starting vectors that have, with high probability, a non-negligible overlap with the eigenvectors of interest. For this purpose, a safe and common choice are unstructured Gaussian random matrices. In this work, we investigate the use of random Khatri-Rao products in eigenvalue solvers. On the one hand, we establish a novel subspace embedding property that provides theoretical justification for the use of such structured random matrices. On the other hand, we highlight the potential algorithmic benefits when solving eigenvalue problems with Kronecker product structure, as they arise frequently from the discretization of eigenvalue problems for differential operators on tensor product domains. In particular, we consider the use of random Khatri-Rao products within a contour integral method and LOBPCG. Numerical experiments indicate that the gains for the contour integral method strongly depend on the ability to efficiently and accurately solve (shifted) matrix equations with low-rank right-hand side. The flexibility of LOBPCG to directly employ preconditioners makes it easier to benefit from Khatri-Rao product structure, at the expense of having less theoretical justification.

We introduce a novel paradigm in compiler optimization powered by Large Language Models with compiler feedback to optimize the code size of LLVM assembly. The model takes unoptimized LLVM IR as input and produces optimized IR, the best optimization passes, and instruction counts of both unoptimized and optimized IRs. Then we compile the input with generated optimization passes and evaluate if the predicted instruction count is correct, generated IR is compilable, and corresponds to compiled code. We provide this feedback back to LLM and give it another chance to optimize code. This approach adds an extra 0.53% improvement over -Oz to the original model. Even though, adding more information with feedback seems intuitive, simple sampling techniques achieve much higher performance given 10 or more samples.

Large language models show great potential in generating and optimizing code. Widely used sampling methods such as Nucleus Sampling increase the diversity of generation but often produce repeated samples for low temperatures and incoherent samples for high temperatures. Furthermore, the temperature coefficient has to be tuned for each task, limiting its usability. We present Priority Sampling, a simple and deterministic sampling technique that produces unique samples ordered by the model's confidence. Each new sample expands the unexpanded token with the highest probability in the augmented search tree. Additionally, Priority Sampling supports generation based on regular expression that provides a controllable and structured exploration process. Priority Sampling outperforms Nucleus Sampling for any number of samples, boosting the performance of the original model from 2.87% to 5% improvement over -Oz. Moreover, it outperforms the autotuner used for the generation of labels for the training of the original model in just 30 samples.

The diverse and intriguing phenomena observed in twisted bilayer systems, such as graphene and transition metal dichalcogenides, prompted new questions about the emergent effects that they may host. However, the practical challenge of realizing these structures on a scale large enough for spectroscopic investigation, remains a significant hurdle, resulting in a scarcity of direct measurements of the electronic band structure of twisted transition metal dichalcogenide bilayers. Here we present a systematic nanoscale angle-resolved photoemission spectroscopy investigation of bulk, single layer, and twisted bilayer WS2 with a small twist angle of 4.4{\deg}. The experimental results are compared with theoretical calculations based on density functional theory along the high-symmetry directions {\Gamma}-K and {\Gamma}-M. Surprisingly, the electronic band structure measurements suggest a structural relaxation occurring at 4.4{\deg} twist angle, and formation of large, untwisted bilayer regions., Comment: 23 pages, 11 figures

We study quasi-freestanding bilayer graphene on silicon carbide intercalated by calcium. The intercalation, and subsequent changes to the system, were investigated by low-energy electron diffraction, angle-resolved photoemission spectroscopy (ARPES) and density-functional theory (DFT). Calcium is found to intercalate only at the graphene-SiC interface, completely displacing the hydrogen terminating SiC. As a consequence, the system becomes highly n-doped. Comparison to DFT calculations shows that the band dispersion, as determined by ARPES, deviates from the band structure expected for Bernal-stacked bilayer graphene. Instead, the electronic structure closely matches AA-stacked bilayer graphene on Ca-terminated SiC, indicating a spontaneous transition from AB- to AA-stacked bilayer graphene following calcium intercalation of the underlying graphene-SiC interface., Comment: 14 pages, 3 figures

We explore the novel application of Large Language Models to code optimization. We present a 7B-parameter transformer model trained from scratch to optimize LLVM assembly for code size. The model takes as input unoptimized assembly and outputs a list of compiler options to best optimize the program. Crucially, during training, we ask the model to predict the instruction counts before and after optimization, and the optimized code itself. These auxiliary learning tasks significantly improve the optimization performance of the model and improve the model's depth of understanding. We evaluate on a large suite of test programs. Our approach achieves a 3.0% improvement in reducing instruction counts over the compiler, outperforming two state-of-the-art baselines that require thousands of compilations. Furthermore, the model shows surprisingly strong code reasoning abilities, generating compilable code 91% of the time and perfectly emulating the output of the compiler 70% of the time.

Advanced compiler technology is crucial for enabling machine learning applications to run on novel hardware, but traditional compilers fail to deliver performance, popular auto-tuners have long search times and expert-optimized libraries introduce unsustainable costs. To address this, we developed LoopTune, a deep reinforcement learning compiler that optimizes tensor computations in deep learning models for the CPU. LoopTune optimizes tensor traversal order while using the ultra-fast lightweight code generator LoopNest to perform hardware-specific optimizations. With a novel graph-based representation and action space, LoopTune speeds up LoopNest by 3.2x, generating an order of magnitude faster code than TVM, 2.8x faster than MetaSchedule, and 1.08x faster than AutoTVM, consistently performing at the level of the hand-tuned library Numpy. Moreover, LoopTune tunes code in order of seconds.

Magnesium intercalated 'quasi-freestanding' bilayer graphene on 6H-SiC(0001) (Mg-QFSBLG) has many favorable properties (e.g., highly n-type doped, relatively stable in ambient conditions). However, intercalation of Mg underneath monolayer graphene is challenging, requiring multiple intercalation steps. Here, we overcome these challenges and subsequently increase the rate of Mg intercalation by laser patterning (ablating) the graphene to form micron-sized discontinuities. We then use low energy electron diffraction to verify Mg-intercalation and conversion to Mg-QFSBLG, and X-ray photoelectron spectroscopy to determine the Mg intercalation rate for patterned and non-patterned samples. By modeling Mg intercalation with the Verhulst equation, we find that the intercalation rate increase for the patterned sample is 4.5$\pm$1.7. Since the edge length of the patterned sample is $\approx$5.2 times that of the non-patterned sample, the model implies that the increased intercalation rate is proportional to the increase in edge length. Moreover, Mg intercalation likely begins at graphene discontinuities in pristine samples (not step edges or flat terraces), where the 2D-like crystal growth of Mg-silicide proceeds. Our laser patterning technique may enable the rapid intercalation of other atomic or molecular species, thereby expanding upon the library of intercalants used to modify the characteristics of graphene, or other 2D materials and heterostructures., Comment: 24 pages, 4 figures

Deep supervised models have an unprecedented capacity to absorb large quantities of training data. Hence, training on multiple datasets becomes a method of choice towards strong generalization in usual scenes and graceful performance degradation in edge cases. Unfortunately, different datasets often have incompatible labels. For instance, the Cityscapes road class subsumes all driving surfaces, while Vistas defines separate classes for road markings, manholes etc. Furthermore, many datasets have overlapping labels. For instance, pickups are labeled as trucks in VIPER, cars in Vistas, and vans in ADE20k. We address this challenge by considering labels as unions of universal visual concepts. This allows seamless and principled learning on multi-domain dataset collections without requiring any relabeling effort. Our method achieves competitive within-dataset and cross-dataset generalization, as well as ability to learn visual concepts which are not separately labeled in any of the training datasets. Experiments reveal competitive or state-of-the-art performance on two multi-domain dataset collections and on the WildDash 2 benchmark., Comment: 27 pages, 16 figures, 10 tables, accepted to International Journal of Computer Vision

Ultrafast light sources have become an indispensable tool to access and understand transient phenomenon in material science. However, a simple and easy-to-implement method for harmonic selection, with high transmission efficiency and pulse duration conservation, is still a challenge. Here we showcase and compare two approaches for selecting the desired harmonic from a high harmonic generation source while achieving the above goals. The first approach is the combination of extreme ultraviolet spherical mirrors with transmission filters and the second approach uses a normal-incidence spherical grating. Both solutions target time- and angle-resolved photoemission spectroscopy with photon energies in the 10-20 eV range but are relevant for other experimental techniques as well. The two approaches for harmonic selection are characterized in terms of focusing quality, efficiency, and temporal broadening. It is demonstrated that a focusing grating is able to provide much higher transmission as compared to the mirror+filter approach (3.3 times higher for 10.8 eV and 12.9 times higher for 18.1 eV), with only a slight temporal broadening (6.8% increase) and a somewhat larger spot size (~30% increase). Overall, our study establishes an experimental perspective on the trade-off between a single grating normal incidence monochromator design and the use of filters. As such, it provides a basis for selecting the most appropriate approach in various fields where an easy-to-implement harmonic selection from high harmonic generation is needed.

The success in studying 2D materials inherently relies on producing samples of large area, and high quality enough for the experimental conditions. Because their 2D nature surface sensitive techniques such as photoemission spectroscopy , tunneling microscopy and electron diffraction, that work in ultra high vacuum (UHV) environment are prime techniques that have been employed with great success in unveiling new properties of 2D materials but it requires samples to be free of adsorbates. The technique that most easily and readily yields 2dmaterials of highest quality is indubitably mechanical exfoliation from bulk grown samples, however as this technique is traditionally done in dedicated environment, the transfer of these samples into UHV setups requires some form of surface cleaning that tempers with the sample quality. In this article, we report on a simple and general method of \textit{in-situ} mechanical exfoliation directly in UHV that yields large-area single-layered films. By employing standard UHV cleaning techniques and by purpusedly exploiting the chemical affinity between the substrate and the sample we could yield large area exfoliation of transition metal dichalcogenides. Multiple transition metal dichalcogenides, both metallic and semiconducting, are exfoliated \textit{in-situ} onto Au and Ag, and Ge. Exfoliated flakes are found to be sub-milimeter size with excellent crystallinity and purity, as evidenced by angle-resolved photoemission spectroscopy, atomic force microscopy and low-energy electron diffraction. In addition, we demonstrate exfoliation of air-sensitive 2D materials and possibility of controlling the substrate-2D material twist angle.

What is the relation between a word and its description, or a word and its embedding? Both descriptions and embeddings are semantic representations of words. But, what information from the original word remains in these representations? Or more importantly, which information about a word do these two representations share? Definition Modeling and Reverse Dictionary are two opposite learning tasks that address these questions. The goal of the Definition Modeling task is to investigate the power of information laying inside a word embedding to express the meaning of the word in a humanly understandable way -- as a dictionary definition. Conversely, the Reverse Dictionary task explores the ability to predict word embeddings directly from its definition. In this paper, by tackling these two tasks, we are exploring the relationship between words and their semantic representations. We present our findings based on the descriptive, exploratory, and predictive data analysis conducted on the CODWOE dataset. We give a detailed overview of the systems that we designed for Definition Modeling and Reverse Dictionary tasks, and that achieved top scores on SemEval-2022 CODWOE challenge in several subtasks. We hope that our experimental results concerning the predictive models and the data analyses we provide will prove useful in future explorations of word representations and their relationships.

Here we present a high repetition rate, narrow band-width, extreme ultraviolet (XUV) photon source for time- and angle-resolved photoemission spectroscopy (tr-ARPES). The narrow band width pulses $\Delta E=9, 14, 18$ meV for photon energies $h\nu=10.8, 18.1, 25.3$ eV are generated through High Harmonic Generation (HHG) using ultra-violet (UV) drive pulses with relatively long pulse lengths (461 fs). The HHG setup employs an annuluar drive beam in a tight focusing geometry at a repetition rate of 250 kHz. Photon energy selection is provided by a series of selectable multilayer bandpass mirrors and thin film filters, thus avoiding any time broadening introduced by single grating monochromators. A two stage optical-parametric amplifier provides $< 100$ fs tunable pump pulses from 0.65 $\mu$m to 9 $\mu$m. The narrow bandwidth performance of the light source is demonstrated through ARPES measurements on a series of quantum materials including the high-temperature superconductor Bi-2212, WSe$_2$ and graphene.

In this paper we consider a constrained parabolic optimal control problem. The cost functional is quadratic and it combines the distance of the trajectory of the system from the desired evolution profile together with the cost of a control. The constraint is given by a term measuring the distance between the final state and the desired state towards which the solution should be steered. The control enters the system through the initial condition. We present a geometric analysis of this problem and provide a closed-form expression for the solution. This approach allows us to present the sensitivity analysis of this problem based on the resolvent estimates for the generator of the system. The numerical implementation is performed by exploring efficient rational Krylov approximation techniques that allow us to approximate a complex function of an operator by a series of linear problems. Our method does not depend on the actual choice of discretization. The main approximation task is to construct an efficient rational approximation of a generalized exponential function. It is well known that this class of functions allows exponentially convergent rational approximations, which, combined with the sensitivity analysis of the closed form solution, allows us to present a robust numerical method. Several case studies are presented to illustrate our results., Comment: 25 pages, 7 figures

To address the challenge of performance analysis on the US DOE's forthcoming exascale supercomputers, Rice University has been extending its HPCToolkit performance tools to support measurement and analysis of GPU-accelerated applications. To help developers understand the performance of accelerated applications as a whole, HPCToolkit's measurement and analysis tools attribute metrics to calling contexts that span both CPUs and GPUs. To measure GPU-accelerated applications efficiently, HPCToolkit employs a novel wait-free data structure to coordinate monitoring and attribution of GPU performance. To help developers understand the performance of complex GPU code generated from high-level programming models, HPCToolkit constructs sophisticated approximations of call path profiles for GPU computations. To support fine-grained analysis and tuning, HPCToolkit uses PC sampling and instrumentation to measure and attribute GPU performance metrics to source lines, loops, and inlined code. To supplement fine-grained measurements, HPCToolkit can measure GPU kernel executions using hardware performance counters. To provide a view of how an execution evolves over time, HPCToolkit can collect, analyze, and visualize call path traces within and across nodes. Finally, on NVIDIA GPUs, HPCToolkit can derive and attribute a collection of useful performance metrics based on measurements using GPU PC samples. We illustrate HPCToolkit's new capabilities for analyzing GPU-accelerated applications with several codes developed as part of the Exascale Computing Project.

Deep supervised models have an unprecedented capacity to absorb large quantities of training data. Hence, training on many datasets becomes a method of choice towards graceful degradation in unusual scenes. Unfortunately, different datasets often use incompatible labels. For instance, the Cityscapes road class subsumes all driving surfaces, while Vistas defines separate classes for road markings, manholes etc. We address this challenge by proposing a principled method for seamless learning on datasets with overlapping classes based on partial labels and probabilistic loss. Our method achieves competitive within-dataset and cross-dataset generalization, as well as ability to learn visual concepts which are not separately labeled in any of the training datasets. Experiments reveal competitive or state-of-the-art performance on two multi-domain dataset collections and on the WildDash 2 benchmark., Comment: 18 pages, 8 figures, 11 tables

Semi-supervised learning an attractive technique in practical deployments of deep models since it relaxes the dependence on labeled data. It is especially important in the scope of dense prediction because pixel-level annotation requires significant effort. This paper considers semi-supervised algorithms that enforce consistent predictions over perturbed unlabeled inputs. We study the advantages of perturbing only one of the two model instances and preventing the backward pass through the unperturbed instance. We also propose a competitive perturbation model as a composition of geometric warp and photometric jittering. We experiment with efficient models due to their importance for real-time and low-power applications. Our experiments show clear advantages of (1) one-way consistency, (2) perturbing only the student branch, and (3) strong photometric and geometric perturbations. Our perturbation model outperforms recent work and most of the contribution comes from photometric component. Experiments with additional data from the large coarsely annotated subset of Cityscapes suggest that semi-supervised training can outperform supervised training with the coarse labels., Comment: The source code is available at https://github.com/Ivan1248/semisup-seg-efficient

Normalizing flows are bijective mappings between inputs and latent representations with a fully factorized distribution. They are very attractive due to exact likelihood valuation and efficient sampling. However, their effective capacity is often insufficient since the bijectivity constraint limits the model width. We address this issue by incrementally padding intermediate representations with noise. We precondition the noise in accordance with previous invertible units, which we describe as cross-unit coupling. Our invertible glow-like modules increase the model expressivity by fusing a densely connected block with Nystrom self-attention. We refer to our architecture as DenseFlow since both cross-unit and intra-module couplings rely on dense connectivity. Experiments show significant improvements due to the proposed contributions and reveal state-of-the-art density estimation under moderate computing budgets., Comment: Accepted at NeurIPS2021

We present a perturbed subspace iteration algorithm to approximate the lowermost eigenvalue cluster of an elliptic eigenvalue problem. As a prototype, we consider the Laplace eigenvalue problem posed in a polygonal domain. The algorithm is motivated by the analysis of inexact (perturbed) inverse iteration algorithms in numerical linear algebra. We couple the perturbed inverse iteration approach with mesh refinement strategy based on residual estimators. We demonstrate our approach on model problems in two and three dimensions., Comment: 30 pages, 15 figures

The industrial realization of graphene has so far been limited by challenges related to the quality, reproducibility, and high process temperatures required to manufacture graphene on suitable substrates. We demonstrate that epitaxial graphene can be grown on transition metal treated 6H-SiC(0001) surfaces, with an onset of graphitization starting around $450-500^\circ\text{C}$. From the chemical reaction between SiC and thin films of Fe or Ru, $\text{sp}^{3}$ carbon is liberated from the SiC crystal and converted to $\text{sp}^{2}$ carbon at the surface. The quality of the graphene is demonstrated using angle-resolved photoemission spectroscopy and low-energy electron diffraction. Furthermore, the orientation and placement of the graphene layers relative to the SiC substrate is verified using angle-resolved absorption spectroscopy and energy-dependent photoelectron spectroscopy, respectively. With subsequent thermal treatments to higher temperatures, a steerable diffusion of the metal layers into the bulk SiC is achieved. The result is graphene supported on magnetic silicide or optionally, directly on semiconductor, at temperatures ideal for further large-scale processing into graphene based device structures., Comment: 10 pages, 4 figures, 51 references

We propose an a posteriori error estimator for high-order $p$- or $hp$-finite element discretizations of selfadjoint linear elliptic eigenvalue problems that is appropriate for estimating the error in the approximation of an eigenvalue cluster and the corresponding invariant subspace. The estimator is based on the computation of approximate error functions in a space that complements the one in which the approximate eigenvectors were computed. These error functions are used to construct estimates of collective measures of error, such as the Hausdorff distance between the true and approximate clusters of eigenvalues, and the subspace gap between the corresponding true and approximate invariant subspaces. Numerical experiments demonstrate the practical effectivity of the approach.

Intrinsic magnetic topological insulators offer low disorder and large magnetic bandgaps for robust magnetic topological phases operating at higher temperatures. By controlling the layer thickness, emergent phenomena such as the Quantum Anomalous Hall (QAH) effect and axion insulator phases have been realised. These observations occur at temperatures significantly lower than the Neel temperature of bulk MnBi2Te4, and measurement of the magnetic energy gap at the Dirac point in ultra-thin MnBi2Te4 has yet to be achieved. Critical to achieving the promise of this system is a direct measurement of the layer-dependent energy gap and verifying whether the gap is magnetic in the QAH phase. Here we utilise temperature dependent angle-resolved photoemission spectroscopy to study epitaxial ultra-thin MnBi2Te4. We directly observe a layer dependent crossover from a 2D ferromagnetic insulator with a bandgap greater than 780 meV in one septuple layer (1 SL) to a QAH insulator with a large energy gap (>100 meV) at 8 K in 3 and 5 SL MnBi2Te4. The QAH gap is confirmed to be magnetic in origin, as it abruptly diminishes with increasing temperature above 8 K. The direct observation of a large magnetic energy gap in the QAH phase of few-SL MnBi2Te4 is promising for further increasing the operating temperature of QAH materials.

We present our submission to the semantic segmentation contest of the Robust Vision Challenge held at ECCV 2020. The contest requires submitting the same model to seven benchmarks from three different domains. Our approach is based on the SwiftNet architecture with pyramidal fusion. We address inconsistent taxonomies with a single-level 193-dimensional softmax output. We strive to train with large batches in order to stabilize optimization of a hard recognition problem, and to favour smooth evolution of batchnorm statistics. We achieve this by implementing a custom backward step through log-sum-prob loss, and by using small crops before freezing the population statistics. Our model ranks first on the RVC semantic segmentation challenge as well as on the WildDash 2 leaderboard. This suggests that pyramidal fusion is competitive not only for efficient inference with lightweight backbones, but also in large-scale setups for multi-domain application., Comment: 2 pages, 2 tables, no figures

We use angle-resolved photoemission spectroscopy to investigate the electronic structure of bilayer graphene at high n-doping and extreme displacement fields, created by intercalating epitaxial monolayer graphene on silicon carbide with magnesium to form quasi-freestanding bilayer graphene on magnesium-terminated silicon carbide. Angle-resolved photoemission spectroscopy reveals that upon magnesium intercalation, the single massless Dirac band of epitaxial monolayer graphene is transformed into the characteristic massive double-band Dirac spectrum of quasi-freestanding bilayer graphene. Analysis of the spectrum using a simple tight binding model indicates that magnesium intercalation results in an n-type doping of 2.1 $\times$ 10$^{14}$ cm$^{-2}$, creates an extremely high displacement field of 2.6 V/nm, opening a considerable gap of 0.36 eV at the Dirac point. This is further confirmed by density-functional theory calculations for quasi-freestanding bilayer graphene on magnesium-terminated silicon carbide, which show a similar doping level, displacement field and bandgap. Finally, magnesium-intercalated samples are surprisingly robust to ambient conditions; no significant changes in the electronic structure are observed after 30 minutes exposure in air., Comment: 46 pages including supporting information, 4 figures in the main paper and 9 figures in the supporting information

The intercalation of epitaxial graphene on SiC(0001) with Ca has been studied extensively, yet precisely where the Ca resides remains elusive. Furthermore, the intercalation of Mg underneath epitaxial graphene on SiC(0001) has not been reported. Here, we use low energy electron diffraction, x-ray photoelectron spectroscopy, secondary electron cut-off photoemission and scanning tunneling microscopy to elucidate the physical and electronic structure of both Ca- and Mg-intercalated epitaxial graphene on 6H-SiC(0001). We find that Ca intercalates underneath the buffer layer and bonds to the Si-terminated SiC surface, breaking the C-Si bonds of the buffer layer i.e. 'freestanding' the buffer layer to form Ca-intercalated quasi-freestanding bilayer graphene (Ca-QFSBLG). The situation is similar for the Mg-intercalation of epitaxial graphene on SiC(0001), where an ordered Mg-terminated reconstruction at the SiC surface and Mg bonds to the Si-terminated SiC surface are formed, resulting in Mg-intercalated quasi-freestanding bilayer graphene (Mg-QFSBLG). Ca-intercalation underneath the buffer layer has not been considered in previous studies of Ca-intercalated epitaxial graphene. Furthermore, we find no evidence that either Ca or Mg intercalates between graphene layers. However, we do find that both Ca-QFSBLG and Mg-QFSBLG exhibit very low workfunctions of 3.68 and 3.78 eV, respectively, indicating high n-type doping. Upon exposure to ambient conditions, we find Ca-QFSBLG degrades rapidly, whereas Mg-QFSBLG remains remarkably stable., Comment: 58 pages, 10 figures, 4 tables. Revised text and figures

Layered indium selenides ($In_{2}Se_{3}$) have recently been discovered to host robust out-of-plane and in-plane ferroelectricity in the $\alpha$ and $\beta$' phases, respectively. In this work, we utilise angle-resolved photoelectron spectroscopy to directly measure the electronic bandstructure of $\beta '-In_{2}Se_{3}$, and compare to hybrid density functional theory (DFT) calculations. In agreement with DFT, we find the band structure is highly two-dimensional, with negligible dispersion along the c-axis. Due to n-type doping we are able to observe the conduction band minima, and directly measure the minimum indirect (0.97 eV) and direct (1.46 eV) bandgaps. We find the Fermi surface in the conduction band is characterized by anisotropic electron pockets with sharp in-plane dispersion about the $\overline{M}$ points, yielding effective masses of 0.21 $m_{0}$ along $\overline{KM}$ and 0.33 $m_{0}$ along $\overline{\Gamma M}$. The measured band structure is well supported by hybrid density functional theory calculations. The highly two-dimensional (2D) bandstructure with moderate bandgap and small effective mass suggest that $\beta'-In_{2}Se_{3}$ is a potentially useful new van der Waals semiconductor. This together with its ferroelectricity makes it a viable material for high-mobility ferroelectric-photovoltaic devices, with applications in non-volatile memory switching and renewable energy technologies., Comment: 19 pages, 4 + 1 figures; typos corrected;added references; updated figures & discussion to reflect changes in model