What can we “learn” from atoms?
In machine learning, energy-based models are rooted in concepts common to magnetism, like the Ising model. Within these models, plasticity, learning, and ultimately pattern recognition can be linked to the dynamics of coupled spin ensembles. While this behavior is commercially emulated in software, there are strong pursuits to implement these concepts directly and autonomously in solid-state materials. To date, hybrid approaches, which often use the serendipitous electric, magnetic, or optical response of materials, emulate machine learning functionality with the help of external computers. Yet, there is still no clear understanding of how to create machine learning functionality from fundamental physical concepts in materials, like hysteresis, glassiness, or spin dynamics.
Based on scanning tunneling microscopy, magnetic atoms on surfaces have become a model playground to understand and design magnetic order. However, these model systems historically have been probed in limits that are not usable for machine learning functionality. In this talk, I will illustrate new model platforms to realize machine learning functionality directly in the dynamics of coupled spin ensembles. I will first review the concept of energy-based neural networks and how they are linked to multi-modal landscapes and the physics of spin glasses. I will then highlight a new example based on the recent discovery of orbital memory: an atomic-scale ensemble that mimics the properties of a Boltzmann machine. I will illustrate the creation of atomic-scale neurons and synapses, in addition to new learning concepts based on the separation of time scales and self-adaptive behavior. I will also discuss recent cutting-edge developments that enable magnetic characterization in new extreme limits and how this platform may be applied toward autonomous adaption and quantum machine learning.
Atomic spin structures on surfaces
Scanning tunneling microscopy (STM) has proved to be a mature technique for the study of magnetic impurities on different substrates. Additionally, the STM allows us to manipulate atoms and assemble magnetic structures of atomic dimensions that are going to behave differently depending on their geometrical and chemical environment. We have applied our techniques to the study of magnetic spectra on atomic spin structures not only on metallic system [1] but also on superconducting surfaces [2,3] revealing the topological properties [4]. Such magnetic impurities on different substrates allow us to explore many-body effects and exotic phenomena in different experimental spin systems giving us an understanding on the parameters on each system.
Atomic scale visualization of electron-pair fluids and crystals
Machine Learning of Quantum Images
Probing the electronic structure of monolayer flakes of cuprate superconductors
The role of dimensionality in high Tc superconductivity is an interesting issue: many of the high Tc superconductor have layered atomic structures, and yet the link between the high Tc superconductivity and the two-dimensional nature of the crystal structure remains elusive. We fabricated atomically thin Bi2Sr2CaCu2O8+d (Bi-2212) and Bi2Sr2CuO6+d (Bi-2201) flakes, and used scanning tunneling microscopy/spectroscopy (STM/STS) to investigate their electronic structure. In this talk, I will discuss our recent results on the superconducting gap, pseudogap and charge order in Bi-2212 and Bi-2201 in the ultimate 2D limit.
Probing Atomic Quantum Defects in 2D Semiconductors
2D materials are an exciting host to engineer atomic quantum systems by chemical design rules. In this colloquium, I will give an overview on our efforts to design defect systems in monolayer transition metal dichalcogenides (TMDs) by means of chemical doping, Helium ion beam bombardment and atomic manipulation.
We directly resolve the discrete electronic spectrum of single dopants in a charge neutral or ionized state and map out their associated defect orbitals [1-5]. Different types of defects reveal the interplay between chemical impurity states [1-3], multi-valley hydrogenic bound states [4,6], and electron-phonon coupling [1,7] at reduced dimensions. In particular, we will discuss the atomically controlled generation of magnetic carbon radical ions (CRIs) in synthetic TMDs [7] and electrically driven photon emission from individual defects [8].
Recording available from the organizers
Electrons in Moiré Superlattices: A playground for correlation and topology
Interactions among electrons and the topology of their energy bands can create novel quantum phases of matter. The discovery of electronic bands with flat energy dispersion in magic-angle twisted bilayer graphene (MATBG) has created a unique opportunity to search for new correlated and topological electronic phases. We have developed new scanning tunneling microscopy (STM) and spectroscopy (STS) techniques to probe the nature of electronic correlations and to detect novel topological phases in two-dimensional systems, such as MATBG. Density-tuned STS studies have enabled us to study the properties of MATBG as function of carrier concentration revealing key and new properties of this novel material. These measurements establish that MATBG is a strong correlated system at all partial filling of its flat bands. [1] The strength of the interactions, which can be measured in our experiments, is found to be larger than the flat bandwidth in the non-interacting limit. We demonstrate that these interactions drive a cascade of transitions at each integer filling of these bands, creating likely the insulating states at low temperatures that are spin or valley polarized.[2] Most recently, we developed a new STS technique to detect topological phases and their associated Chern numbers and used it to show that strong interactions drive the formation of unexpected topological insulating phases in MATBG [3]. These phases, which are stabilized by a weak magnetic field, are rare examples of when topology emerges from interaction between electrons. I will describe these experiments, and other ongoing efforts, that illustrate the power of atomic scale experiments in revealing novel physics of electrons in moiré superlattices.
“All that glitters is gold”: First Impressions of Sparse Sampling for fast Quasiparticle Interference measurements with Au(111) Surface State
The band structure conveys the most fundamental information of a material that enables a prediction of important properties, such as its transport behavior, the presence of superconducting phases, optical transitions, magnetism etc. However, theoretical modelling of novel quantum materials, often require an experimental verification in conditions in which established methods, such as ARPES are inhibited by the involvement of magnetic fields, milli Kelvin temperatures, or device geometries. Although the complementary method, scanning tunneling microscopy (STM) based quasiparticle interference (QPI) mapping (1), operates well in such conditions, its band-structure mapping via the scattering space is also excruciatingly slow. Here we investigate the promise of a compressed sensing implementation of QPI measurements with STM that fundamentally speeds up large scale spectroscopic mapping of the local density of states (LDOS)(2). The condition for compressive sensing is fulfilled by the sparsity of the LDOS information in the reciprocal domain, which is also the sought QPI pattern. We utilize the Shockley surface state of Au(111) as a well-studied model system for our QPI implementation(3, 4). Instead of visiting all points on a regular grid, we only visit a random and small fraction of the usual measurement locations for which we use a traveling salesperson optimization to reduce tip-motion related overhead. This random walk separates time and spatial correlation that previously got intermixed in conventional grid mapping schemes and that were particularly problematic for periodic perturbations, such as line-noise or mechanical resonances. Our sparse QPI implementation substantially reduces the overall measurement time without the introducing of artifacts. Looking at the QPI maps of Au(111), we not only find the expected near-free electron dispersion reported before but also a band-gap that can be traced back to the action of the period potential of the herringbone reconstruction (5, 6). We finally discuss limitations and opportunities of the sparse QPI method.