BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Sonia Zhang (Princeton) DTSTART:20200508T170000Z DTEND:20200508T180000Z DTSTAMP:20260422T225717Z UID:PrincetonFPO/1 DESCRIPTION:Title: Topology in quantum magnets and superconductors\nby Sonia Zhang ( Princeton) as part of Princeton FPO\n\n\nAbstract\nA unifying theme in und erstanding the macroscopic properties of quantum materials is the concept of emergence\, where novel effects and unusual phases arise from the inter play of spin-orbit effects\, band topology and strong correlation. In this dissertation\, using scanning tunnelling microscopy and spectroscopy (STM /S)\, we explore a series of unconventional spin-orbit materials including magnets and superconductors. First\, we provide a brief introduction to t he working principles of low temperature atomic resolution STM/S operating in conjunction with a vector magnetic field capability. Utilising this s tate-of-the-art capability we explore the effects of artificial quantum im purity and vortex defects on topological superconductor candidates LiFeAs and PbTaSe2. We find that a controlled deposition of a carefully chosen c lass of atomic scale magnetic impurities on their surfaces generates zero- bias peaks exhibiting signatures of Majorana zero modes\, despite being ab sent in vortices in pristine samples. In a second line of research\, we ex plore wavefunction topology in correlated kagome magnets. In Fe3Sn2\, we d iscover a giant and anisotropic many-body spin-orbit tunability whose orig in remains unclear in current theoretical models. In Co3Sn2S2 we find an unexpected negative magnetic response in the kagome flat band arising fro m the topology. Finally\, we explore topological magnet Mn3Sn and show tha t the unique geometry of the kagome lattice leads to a remarkable manifest ation of an apparent Kondo lattice-type effect\, usually observed in stron gly correlated heavy fermion materials. Our results taken collectively fea ture novel effects and phases arising from rich interplay among spin-orbit effects\, band topology and many-body interactions in quantum magnets and exotic superconductors\, that may potentially lead to new frontiers in co ndensed matter physics.\n LOCATION:https://researchseminars.org/talk/PrincetonFPO/1/ END:VEVENT BEGIN:VEVENT SUMMARY:Yunqin Zheng DTSTART:20200504T170000Z DTEND:20200504T180000Z DTSTAMP:20260422T225717Z UID:PrincetonFPO/2 DESCRIPTION:Title: Tensor network states\, entanglement\, and anomalies of topological p hases of matter\nby Yunqin Zheng as part of Princeton FPO\n\n\nAbstrac t\nThis dissertation investigates two aspects of topological phases of mat ter: 1) the tensor network state (TNS) representations of the ground state s as well as their en- tanglement entropies of gapped Hamiltonians in dive rse dimensions\; 2) the anomalies and dynamics of strongly coupled quantum field theories.\n\nFor the first aspect\, we first show an efficient meth od of analytically deriving the translation invariant TNS and matrix produ ct state (MPS) representation for the ground state of translation invarian t stabilizer code Hamiltonians in both 1d and higher dimensions. These TNS /MPS states have minimal virtual bond dimension. Using the TNS\, we derive the entanglement entropy for a variety of stabilizer codes\, including th e fracton models the Haah code. We further go beyond the stabilizer codes and study the structure of entanglement entropy for generic 3d gapped Hami ltonians. In particular\, an explicit formula for a universal physical obs ervable – topological entanglement entropy (TEE) – has been derived\, which sharpens previous results. Our formula shows that the TEE across an arbitrary entanglement surface is linearly proportional to the TEE across a torus.\n\nFor the second aspect\, we use the global symmetries and their ’t Hooft anomalies of the SU(2) Yang-Mills theory with a theta term to constrain its dynamics. In particular\, we point out that there are four d ifferent such theories\, distinguished by Lorentz symmetry enrichments of the Wilson loops in the SU(2) fundamental representation. We further deriv e a new mixed anomaly between time reversal and one form symmetry which ca n only be seen on an unorientable manifold. We further use the anomalies t o explore various possible dynamics\, such as nontrivial degrees of freedo m localized on the domain wall due to spontaneously broken time reversal s ymmetry\, as well as a potentially possible but exotic quantum phase trans ition — Gauge Enhanced Quantum Critical Point.\n LOCATION:https://researchseminars.org/talk/PrincetonFPO/2/ END:VEVENT BEGIN:VEVENT SUMMARY:Jaan Altosaar DTSTART:20200515T170000Z DTEND:20200515T180000Z DTSTAMP:20260422T225717Z UID:PrincetonFPO/4 DESCRIPTION:Title: Probabilistic Modeling of Structure in Science: Statistical Physics t o Recommender Systems\nby Jaan Altosaar as part of Princeton FPO\n\n\n Abstract\nApplied machine learning relies on translating the structure of a problem into a computational model. This arises in applications as diver se as statistical physics and food recommender systems. The pattern of con nectivity in an undirected graphical model or the fact that datapoints in food recommendation are un- ordered collections of features can inform the structure of a model. First\, consider undi- rected graphical models from statistical physics like the ubiquitous Ising model. Basic research in ph ysics requires scalable simulations for comparing the behavior of a model to its experimental counterpart. The Ising model consists of binary random variables with local connectivity\; interactions between neighboring node s can lead to long-range correlations. Modeling these correlations is nece ssary to capture physical phenomena such as phase transitions. To mirror t he local structure of these models\, we use flow- based convolutional gen erative models that can capture long-range correlations. Com- bining flow -based models designed for continuous variables with recent work on hier- archical variational approximations enables the modeling of discrete rando m variables. Compared to existing variational inference methods\, this app roach scales to statistical physics models with tens of thousands of corre lated random variables and uses fewer pa- rameters. Just as computational choices can be made by considering the structure of an undirected graphica l model\, model construction itself can be guided by the structure of indi vidual datapoints. Consider a recommendation task where datapoints consist of un- ordered sets\, and the objective is to maximize top-K recall\, a c ommon recommendation metric. Simple results show that a classifier with z ero worst-case error achieves maxi- mum top-K recall. Further\, the unorde red structure of the data suggests the use of a permutation-invariant clas sifier for statistical and computational efficiency. We evalu- ate such a classifier on human dietary behavior data\, where every meal is an unord ered collection of ingredients\, and find that it outperforms probabilist ic matrix factorization methods. Finally\, we show that building problem s tructure into an approximate infer- ence algorithm improves the accuracy o f probabilistic modeling methods.\n\nZoom ID: 95950675768 Password: 246827 \n\n(Please disregard regular ID and password for the seminar series.)\n LOCATION:https://researchseminars.org/talk/PrincetonFPO/4/ END:VEVENT END:VCALENDAR