arXiv subjects

Marc A. Kastner

Publications and source records attributed to Marc A. Kastner.

At least 19 recordsLinked to original sources

arXiv2025

Small Object Detection for Birds with Swin Transformer

Object detection is the task of detecting objects in an image. In this task, the detection of small objects is particularly difficult. Other than the small size, it is also accompanied by difficulties due to blur, occlusion, and so on. Current small object detection methods are tailored to small and dense situations, such as pedestrians in a crowd or far objects in remote sensing scenarios. However, when the target object is small and sparse, there is a lack of objects available for training, making it more difficult to learn effective features. In this paper, we propose a specialized method for detecting a specific category of small objects; birds. Particularly, we improve the features learned by the neck; the sub-network between the backbone and the prediction head, to learn more effective features with a hierarchical design. We employ Swin Transformer to upsample the image features. Moreover, we change the shifted window size for adapting to small objects. Experiments show that the proposed Swin Transformer-based neck combined with CenterNet can lead to good performance by changing the window sizes. We further find that smaller window sizes (default 2) benefit mAPs for small object detection.

cs.CV↗

arXiv2025

Hybrid metal-semiconductor quantum dots in InAs as a platform for quantum simulation

Arrays of hybrid metal-semiconductor islands offer a new approach to quantum simulation, with key advantages over arrays of conventional quantum dots. Because the metallic component of these hybrid islands has a quasi-continuous level spectrum, each site in an array can be effectively electronically identical; in contrast, each conventional semiconductor quantum dot has its own spectral fingerprint. Meanwhile, the semiconductor component retains gate-tunability of intersite coupling. This combination creates a scalable platform for simulating correlated ground states driven by Coulomb interactions. We report the fabrication and characterization of hybrid metal-semiconductor islands, featuring a submicron metallic component transparently contacting a gate-confined region of an InAs quantum well with tunable couplings to macroscopic leads. Tuning to the weak-coupling limit forms a single-electron transistor with highly-uniform Coulomb peaks, with no resolvable excitation spectrum in the Coulomb diamonds. Upon increasing the transmissions toward the ballistic regime we observe an evolution to dynamical Coulomb blockade.

cond-mat.mes-hall↗

arXiv2025

Extended Fractional Chern Insulators Near Half Flux in Twisted Bilayer Graphene Above the Magic Angle

Fractional Chern insulators (FCIs) -- the lattice analog of fractional quantum Hall states -- form as fractionalized quasiparticles emerge in a partially-filled Chern band. This fractionalization is driven by the interplay of electronic interaction and quantum geometry of the underlying wavefunctions. Bilayer graphene with an interlayer twist near the magic angle of 1.1\textdegree\ hosts diverse correlated electronic states at zero magnetic field. When the twist angle exceeds 1.3\textdegree, the electronic bandwidth is sufficient to suppress the zero-field correlated states. Yet applying a magnetic field can restore the importance of electron-electron interactions. Here, we report strongly-correlated phases when a 1.37\textdegree\ twisted bilayer graphene sample is tuned to near half a magnetic flux quantum per moiré cell, deep into the Hofstadter regime. Most notably, well-quantized odd-denominator FCI states appear in multiple Hofstadter subbands over unusually large ranges of density. We also observe a bending and resetting of the Landau minifan reminiscent of behavior commonly seen in magic-angle samples near integer filling at low magnetic field.

cond-mat.mes-hall↗

arXiv2024

Investigating Conceptual Blending of a Diffusion Model for Improving Nonword-to-Image Generation

Text-to-image diffusion models sometimes depict blended concepts in the generated images. One promising use case of this effect would be the nonword-to-image generation task which attempts to generate images intuitively imaginable from a non-existing word (nonword). To realize nonword-to-image generation, an existing study focused on associating nonwords with similar-sounding words. Since each nonword can have multiple similar-sounding words, generating images containing their blended concepts would increase intuitiveness, facilitating creative activities and promoting computational psycholinguistics. Nevertheless, no existing study has quantitatively evaluated this effect in either diffusion models or the nonword-to-image generation paradigm. Therefore, this paper first analyzes the conceptual blending in a pretrained diffusion model, Stable Diffusion. The analysis reveals that a high percentage of generated images depict blended concepts when inputting an embedding interpolating between the text embeddings of two text prompts referring to different concepts. Next, this paper explores the best text embedding space conversion method of an existing nonword-to-image generation framework to ensure both the occurrence of conceptual blending and image generation quality. We compare the conventional direct prediction approach with the proposed method that combines $k$-nearest neighbor search and linear regression. Evaluation reveals that the enhanced accuracy of the embedding space conversion by the proposed method improves the image generation quality, while the emergence of conceptual blending could be attributed mainly to the specific dimensions of the high-dimensional text embedding space.

cs.MM↗

arXiv2024

Quantitative determination of twist angle and strain in Van der Waals moiré superlattices

Scanning probe techniques are popular, non-destructive ways to visualize the real space structure of Van der Waals moirés. The high lateral spatial resolution provided by these techniques enables extracting the moiré lattice vectors from a scanning probe image. We have found that the extracted values, while precise, are not necessarily accurate. Scan-to-scan variations in the behavior of the piezos which drive the scanning probe, and thermally-driven slow relative drift between probe and sample, produce systematic errors in the extraction of lattice vectors. In this Letter, we identify the errors and provide a protocol to correct for them. Applying this protocol to an ensemble of ten successive scans of near-magic-angle twisted bilayer graphene, we are able to reduce our errors in extracting lattice vectors to less than 1%. This translates to extracting twist angles with a statistical uncertainty less than 0.001° and uniaxial heterostrain with uncertainty on the order of 0.002%.

cond-mat.mtrl-sci↗

arXiv2024

Deterministic fabrication of graphene hexagonal boron nitride moiré superlattices

The electronic properties of moiré heterostructures depend sensitively on the relative orientation between layers of the stack. For example, near-magic-angle twisted bilayer graphene (TBG) commonly shows superconductivity, yet a TBG sample with one of the graphene layers rotationally aligned to a hexagonal Boron Nitride (hBN) cladding layer provided the first experimental observation of orbital ferromagnetism. To create samples with aligned graphene/hBN, researchers often align edges of exfoliated flakes that appear straight in optical micrographs. However, graphene or hBN can cleave along either zig-zag or armchair lattice directions, introducing a 30 degree ambiguity in the relative orientation of two flakes. By characterizing the crystal lattice orientation of exfoliated flakes prior to stacking using Raman and second-harmonic generation for graphene and hBN, respectively, we unambiguously align monolayer graphene to hBN at a near-0 degree, not 30 degree, relative twist angle. We confirm this alignment by torsional force microscopy (TFM) of the graphene/hBN moiré on an open-face stack, and then by cryogenic transport measurements, after full encapsulation with a second, non-aligned hBN layer. This work demonstrates a key step toward systematically exploring the effects of the relative twist angle between dissimilar materials within moiré heterostructures.

cond-mat.mes-hall↗

arXiv2023

Torsional Force Microscopy of Van der Waals Moirés and Atomic Lattices

In a stack of atomically-thin Van der Waals layers, introducing interlayer twist creates a moiré superlattice whose period is a function of twist angle. Changes in that twist angle of even hundredths of a degree can dramatically transform the system's electronic properties. Setting a precise and uniform twist angle for a stack remains difficult, hence determining that twist angle and mapping its spatial variation is very important. Techniques have emerged to do this by imaging the moiré, but most of these require sophisticated infrastructure, time-consuming sample preparation beyond stack synthesis, or both. In this work, we show that Torsional Force Microscopy (TFM), a scanning probe technique sensitive to dynamic friction, can reveal surface and shallow subsurface structure of Van der Waals stacks on multiple length scales: the moirés formed between bi-layers of graphene and between graphene and hexagonal boron nitride (hBN), and also the atomic crystal lattices of graphene and hBN. In TFM, torsional motion of an AFM cantilever is monitored as it is actively driven at a torsional resonance while a feedback loop maintains contact at a set force with the sample surface. TFM works at room temperature in air, with no need for an electrical bias between the tip and the sample, making it applicable to a wide array of samples. It should enable determination of precise structural information including twist angles and strain in moiré superlattices and crystallographic orientation of VdW flakes to support predictable moiré heterostructure fabrication.

cond-mat.mtrl-sci↗

arXiv2023

Universal conductance fluctuations in a MnBi$_2$Te$_4$ thin film

Quantum coherence of electrons can produce striking behaviors in mesoscopic conductors, including weak localization and the Aharonov-Bohm effect. Although magnetic order can also strongly affect transport, the combination of coherence and magnetic order has been largely unexplored. Here, we examine quantum coherence-driven universal conductance fluctuations in the antiferromagnetic, canted antiferromagnetic, and ferromagnetic phases of a thin film of the topological material MnBi$_2$Te$_4$. In each magnetic phase we extract a charge carrier phase coherence length of about 100 nm. The conductance magnetofingerprint is repeatable when sweeping applied magnetic field within one magnetic phase, but changes when the applied magnetic field crosses the antiferromagnetic/canted antiferromagnetic magnetic phase boundary. Surprisingly, in the antiferromagnetic and canted antiferromagnetic phase, but not in the ferromagnetic phase, the magnetofingerprint depends on the direction of the field sweep. To explain these observations, we suggest that conductance fluctuation measurements are sensitive to the motion and nucleation of magnetic domain walls in MnBi$_2$Te$_4$.

cond-mat.mes-hall↗

arXiv2023

MVA2023 Small Object Detection Challenge for Spotting Birds: Dataset, Methods, and Results

Small Object Detection (SOD) is an important machine vision topic because (i) a variety of real-world applications require object detection for distant objects and (ii) SOD is a challenging task due to the noisy, blurred, and less-informative image appearances of small objects. This paper proposes a new SOD dataset consisting of 39,070 images including 137,121 bird instances, which is called the Small Object Detection for Spotting Birds (SOD4SB) dataset. The detail of the challenge with the SOD4SB dataset is introduced in this paper. In total, 223 participants joined this challenge. This paper briefly introduces the award-winning methods. The dataset, the baseline code, and the website for evaluation on the public testset are publicly available.

cs.CV↗

arXiv2023

IPA-CLIP: Integrating Phonetic Priors into Vision and Language Pretraining

Recently, large-scale Vision and Language (V\&L) pretraining has become the standard backbone of many multimedia systems. While it has shown remarkable performance even in unseen situations, it often performs in ways not intuitive to humans. Particularly, they usually do not consider the pronunciation of the input, which humans would utilize to understand language, especially when it comes to unknown words. Thus, this paper inserts phonetic prior into Contrastive Language-Image Pretraining (CLIP), one of the V\&L pretrained models, to make it consider the pronunciation similarity among its pronunciation inputs. To achieve this, we first propose a phoneme embedding that utilizes the phoneme relationships provided by the International Phonetic Alphabet (IPA) chart as a phonetic prior. Next, by distilling the frozen CLIP text encoder, we train a pronunciation encoder employing the IPA-based embedding. The proposed model named IPA-CLIP comprises this pronunciation encoder and the original CLIP encoders (image and text). Quantitative evaluation reveals that the phoneme distribution on the embedding space represents phonetic relationships more accurately when using the proposed phoneme embedding. Furthermore, in some multimodal retrieval tasks, we confirm that the proposed pronunciation encoder enhances the performance of the text encoder and that the pronunciation encoder handles nonsense words in a more phonetic manner than the text encoder. Finally, qualitative evaluation verifies the correlation between the pronunciation encoder and human perception regarding pronunciation similarity.

cs.MM↗

arXiv2023

Low-damage electron beam lithography for nanostructures on Bi$_2$Te$_3$-class topological insulator thin films

Nanostructured topological insulators (TIs) have the potential to impact a wide array of condensed matter physics topics, ranging from Majorana physics to spintronics. However, the most common TI materials, the Bi$_2$Se$_3$ family, are easily damaged during nanofabrication of devices. In this paper, we show that electron beam lithography performed with a 30 or 50 kV accelerating voltage -- common for nanopatterning in academic facilities -- damages both nonmagnetic TIs and their magnetically-doped counterparts at unacceptable levels. We additionally demonstrate that electron beam lithography with a 10 kV accelerating voltage produces minimal damage detectable through low-temperature electronic transport. Although reduced accelerating voltages present challenges in creating fine features, we show that with careful choice of processing parameters, particularly the resist, 100 nm features are reliably achievable.

cond-mat.mes-hall↗

arXiv2022

Reproducibility Companion Paper: Describing Subjective Experiment Consistency by $p$-Value P-P Plot

In this paper we reproduce experimental results presented in our earlier work titled "Describing Subjective Experiment Consistency by $p$-Value P-P Plot" that was presented in the course of the 28th ACM International Conference on Multimedia. The paper aims at verifying the soundness of our prior results and helping others understand our software framework. We present artifacts that help reproduce tables, figures and all the data derived from raw subjective responses that were included in our earlier work. Using the artifacts we show that our results are reproducible. We invite everyone to use our software framework for subjective responses analyses going beyond reproducibility efforts.

cs.MM↗

arXiv2022

Clean quantum point contacts in an InAs quantum well grown on a lattice-mismatched InP substrate

Strong spin-orbit coupling, the resulting large $g$ factor, and small effective mass make InAs an attractive material platform for inducing topological superconductivity. The surface Fermi level pinning in the conduction band enables highly transparent ohmic contact without excessive doping. We investigate electrostatically defined quantum point contacts (QPCs) in a deep-well InAs two-dimensional electron gas. Despite the 3.3% lattice mismatch between the InAs quantum well and the InP substrate, we report clean QPCs with up to eight pronounced quantized conductance plateaus at zero magnetic field. Source-drain dc bias spectroscopy reveals a harmonic confinement potential with a nearly $5$ meV subband spacing. We find a many-body exchange interaction enhancement for the out-of-plane $g$ factor $|g_{\perp}^*| = 27 \pm 1$, whereas the in-plane $g$ factor is isotropic $|g^*_{x}| = |g^*_{y}| = 12 \pm 2$, close to the bulk value for InAs.

cond-mat.mes-hall↗

arXiv2022

Quantum simulation of an exotic quantum critical point in a two-site charge Kondo circuit

Tuning a material to the cusp between two distinct ground states can produce exotic physical properties, unlike those in either of the neighboring phases. The prospect of designing a model experimental system to capture such behavior is tantalizing. An array of tunnel-coupled quantum dots, each hosting a local spin, should have an appropriately complex phase diagram, but scaling up from individual dots to uniform clusters or lattices has proven difficult: though each site can be tuned to the same occupancy, each has a different set of localized wavefunctions whose couplings to neighboring sites cannot be made fully uniform. An array of metal nanostructures has complementary strengths and weaknesses: simple electrostatic tuning can make each element behave essentially identically, but intersite coupling is not tunable. In this work, we study a tunable nanoelectronic circuit comprising two coupled hybrid metallic-semiconductor islands, combining the strengths of the two types of materials, and demonstrating the potential for scalability. With two charge states of an island acting as an effective spin-1/2, the new architecture also offers a rich range of coupling interactions, and we exploit this to demonstrate a novel quantum critical point. Experimental results in the vicinity of the critical point match striking theoretical predictions.

cond-mat.mes-hall↗

arXiv2021

Clean ballistic quantum point contact in SrTiO$_3$

Two dimensional electron gases based on SrTiO$_3$ are an intriguing platform for exploring mesoscopic superconductivity combined with spin-orbit coupling, offering electrostatic tunability from insulator to metal to superconductor within a single material. So far, however, quantum effects in SrTiO$_3$ nanostructures have been complicated by disorder. Here we introduce a facile approach to achieving high mobility and patterning gate-tunable structures in SrTiO$_3$, and use it to demonstrate ballistic constrictions with clean normal state conductance quantization. Conductance plateaus show two-fold degeneracy that persists to magnetic fields of at least 5 T - far beyond what one would expect from the $g$-factor extracted at high fields - a potential signature of electron pairing extending outside the superconducting regime.

cond-mat.mes-hall↗

arXiv2021

Quantum Imaging of Single-Atom Spin-Splitting in a Monolayer Semiconductor

Theoretical work has suggested that monolayer MoS2 doped with Mn should behave as a two-dimensional dilute magnetic semiconductor, which would open up possibilities for spintronic applications, device physics, and novel ground states. The magnetic properties on Mn dopants in MoS2 are dependent on the mid-gap impurity states of said dopants as well as the sites of dopant incorporation and dopant concentration. In this work we use STM/STS to characterize multiple impurity types associated with Mn dopants in MoS2, and use ring features that appear in spectral maps due to tip-induced band bending to investigate the nature of the mid-gap impurity states. The doublet nature of the rings and comparison to DFT calculations show that the Mn states exhibit strong spin splitting which can be quantified. We used scanned MOKE experiments to extend these magnetization measurements from atomic scale to mm scales, and detect the spin susceptibility signal which increases with Mn concentration. These experiments show that single Mn atoms in MoS2 function as active unscreened magnetic moments in the TMD monolayer, and can be harnessed for spin physics applications and science.

cond-mat.mes-hall↗

arXiv2021

Bulk dissipation in the quantum anomalous Hall effect

Even at the lowest accessible temperatures, measurements of the quantum anomalous Hall (QAH) effect have indicated the presence of parasitic dissipative conduction channels. There is no consensus whether parasitic conduction is related to processes in the bulk or along the edges. Here, we approach this problem by comparing transport measurements of Hall bar and Corbino geometry devices fabricated from Cr-doped (BiSb)$_2$Te$_3$. We identify bulk conduction as the dominant source of dissipation at all values of temperature and in-plane electric field. Furthermore, we observe identical breakdown phenomenology in both geometries, indicating that breakdown of the QAH phase is a bulk process. The methodology developed in this study could be used to identify dissipative conduction mechanisms in new QAH materials, ultimately guiding material development towards realization of the QAH effect at higher temperatures.

cond-mat.mes-hall↗

arXiv2021

Evidence of orbital ferromagnetism in twisted bilayer graphene aligned to hexagonal boron nitride

We have previously reported ferromagnetism evinced by a large hysteretic anomalous Hall effect in twisted bilayer graphene (tBLG). Subsequent measurements of a quantized Hall resistance and small longitudinal resistance confirmed that this magnetic state is a Chern insulator. Here we report that, when tilting the sample in an external magnetic field, the ferromagnetism is highly anisotropic. Because spin-orbit coupling is negligible in graphene such anisotropy is unlikely to come from spin, but rather favors theories in which the ferromagnetism is orbital. We know of no other case in which ferromagnetism has a purely orbital origin. For an applied in-plane field larger than $5\ \mathrm{T}$, the out-of-plane magnetization is destroyed, suggesting a transition to a new phase.

cond-mat.mes-hall↗