Scientific Series

I will explain that a geometric theory built upon the theory of complex surfaces can be used to understand wide variety of phenomena in five-dimensional supersymmetric theories, which includes the following:

1. Classification of 5d superconformal field theories (SCFTs).
2. Enhanced flavor symmetries of 5d SCFTs.
3. 5d N=1 gauge theory descriptions of 5d and 6d SCFTs.
4. Dualities between 5d N=1 gauge theories.
5. T-dualities between 6d N=(1,0) little string theories.

This relationship between geometry and 5d theories is based on M-theory and F-theory compactifications.

In our four-dimensional world supersymmetry is the only extension of the classical Poincaré invariance which laid the foundation of modern physics in the beginning of the 20th century. Supersymmetry, a new geometric symmetry extending Poincaré, was discovered in 1970 –– it was overlooked for decades because of its quantum nature. In the next 10 years or so supersymmetry 

assumed the role of a universal framework in which new models for natural phenomena and regularities (e.g. the concept of naturalness) have been developed. It gave rise to a powerful stream of theoretical phenomenology.

The fact that LHC at CERN produced no evidence for low-energy supersymmetry (and naturalness as well) was a powerful blow. However, despite its absence in experiments the less known second mission of supersymmetry is highly successful, with remarkable advances occurring on a regular basis. Supersymmetry proved its power and uniqueness for those who address hard questions in strongly coupled field theories, including Yang-Mills. Some supersymmetry-based exact results obtained in four dimensions are the main topics of my talk. In the past one could hardly dream that such results are possible.

Theories beyond the Standard Model of particle physics often predict new, light, feebly interacting particles whose discovery requires novel search strategies. A light particle, the QCD axion, elegantly solves the outstanding strong-CP problem of the Standard Model; cousins of the QCD axion can also appear, and are natural dark matter candidates.  First, I will discuss my experimental proposal based on thin films, in which dark matter can efficiently convert to detectable single photons. A prototype experiment is underway, and current techniques promise to reach significant new dark matter parameter space.

Second, I will show how rotating black holes turn into axionic and gravitational wave beacons, creating nature's laboratories for ultralight bosons. When an axion's Compton wavelength is comparable to a black hole size, energy and angular momentum from the black hole source exponentially-growing bound states of particles, forming `gravitational atoms'.  These `gravitational atoms' emit monochromatic gravitational wave signals, enabling current searches at gravitational wave observatories to discover ultralight axions. If the axions interact with one another, instead of gravitational waves, black holes populate the universe with axion waves.

Stacking two graphene layers twisted by the ‘magic angle’ 1.1º generates flat energy bands, which in turn catalyzes various strongly correlated phenomena depending on filling and sample details. While this system is most famous for the superconducting and insulating states observed at fractional fillings, I argue that charge neutrality presents an interesting interplay of disorder and interactions.

In scanning tunnelling microscopy (STM), the most striking signature of interactions occurs close to charge neutrality, where the splitting between the flat bands increases dramatically. In analogy with quantum Hall ferromagnetism, I show that this effect may be qualitatively understood as the result of an exchange energy gain. A low-energy manifold of gapped, symmetry-breaking states is identified, one of which possesses quantum valley Hall order. Transport measurements yield ostensibly conflicting information at charge neutrality: while some samples reveal semimetallicity (as expected when correlations are weak), yet others exhibit robust insulation. I reconcile these observations and those of STM by arguing that strong interactions supplemented by weak, smooth disorder stabilize a network of locally gapped quantum valley Hall domains with spatially varying Chern numbers determined by the disorder landscape—--even when an entirely different order is favored in the clean limit. I conclude with a discussion of experimental tests of this proposal via local probes and transport.

In this talk, we present a new outlook on canonical quantum gravity and its coupling to matter.

We will show how this fresh perspective combines the critical elements of holography, loop quantum gravity, and relative locality. I will first focus on the consequences of cutting open a portion of space and show that new symmetry charges and new degrees of freedom reveal themselves. 
I will explain the nature of this boundary symmetry algebra in metric gravity and then first-order gravity. We will see that a rich structure appears that explains from the continuum perspective the non-commutativity of geometric flux, the quantization of the area spectra, the nature of the simplicity constraints but also reveals the dual momentum observables and finally allow to reconcile the elements of canonical gravity with Lorentz invariance.
I will discuss the issue of quantization as a challenge of finding a representation of the boundary algebra and will give clues about where we are in this process.

Let G be a complex reductive group, and X be any smooth projective G-variety. In this talk, we will construct an algebra homomorphism from the G-equivariant homology of the affine Grassmannian Gr_G to the G-equivariant quantum cohomology of X. The construction uses shift operators in quantum cohomologies. We will also discuss the possible extension to the loop rotation equivariant setting and the relation with the Peterson isomorphism when X is the flag variety associated with G. This is based on joint work with Alexander Braverman.

There has long been interest in Beilinson-Bernstein localization for the affine Grassmannian (or affine flag variety). First, Kashiwara-Tanisaki treated the so-called negative level case in the 90's. Some ten years later, Frenkel-Gaitsgory (following work of Beilinson-Drinfeld and Feigin-Frenkel) formulated a conjecture at the critical level and made some progress on it. Their conjecture is more subtle than its negative level counterpart, but also more satisfying. We will review the necessary background from representation theory of Kac-Moody algebras at critical level, formulate the Frenkel-Gaitsgory conjecture, and outline a proof for GL_2. Time permitting, we will discuss how our result provides a test of the Frenkel-Gaitsgory proposal for local geometric Langlands.

The average entanglement entropy of subsystems of random pure states is (nearly) maximal. In this talk, we discuss the average entanglement entropy of subsystems of highly excited eigenstates of integrable and nonintegrable many-body lattice Hamiltonians. For translationally invariant quadratic models (or spin models mappable to them) we prove that, when the subsystem size is not a vanishing fraction of the entire system, the average eigenstate entanglement entropy exhibits a leading volume-law term that is different from that of random pure states. We argue that such a leading term is likely universal for translationally invariant (noninteracting and interacting) integrable models. For random pure states with a fixed particle number (random canonical states) away from half filling and normally distributed real coefficients, we prove that the deviation from the maximal value grows with the square root of the system's volume when the size of the subsystem is one half of that of the system. We then show that the average entanglement entropy of highly excited eigenstates of a particle number conserving quantum chaotic model is the same as that of random canonical states.

Defining a generic quantum system requires, together with a Hilbert space and a Hamiltonian, the introduction of an algebra of observables, or equivalently a tensor product structure. Assuming a background time variable, Cotler, Penington and Ranard showed that the Hamiltonian selects an almost-unique tensor product structure. This result has been advocated by Carrol and collaborators as supporting the Everettian interpretation of quantum mechanics and providing a pivotal tool for quantum gravity. In this talk I argue against this, on the basis of the fact that the Cotler-Penington-Ranard result does not hold in the generic background-independent case where the Hamiltonian is replaced by a Hamiltonian constrain. This reinforces the understanding that entropy and entanglement, that in the quantum theory depend on the tensor product structure, are quantities that are observable dependent. To conclude, I would like to pose the question of whether clocks can be thought as a resource, and how thinking of time in terms of physical clocks can inform our interpretation of quantum mechanics

I will describe how within eleven dimensional supergravity one can compute the logarithmic correction to the Bekenstein-Hawking entropy of certain magnetically charged asymptotically AdS_4 black holes with arbitrary horizon topology.  The result perfectly agrees with the dual field theory computation of the topologically twisted index in  ABJM theory and in certain theories obtained from M5 wrapping a hyperbolic 3-manifold. The extension to rotating, electrically charged AdS_4 black holes and the dual superconformal index will also be discussed. 

A scientific understanding of modern deep learning is still in its early stages. As a first step towards understanding the learning dynamics of neural networks, one can simplify the problem by studying limits that might have theoretical tractability and practical relevance. I’ll begin with a brief survey of our earlier body of work that has investigated the infinite width limit of deep networks, a topic of active study recently. With these results in hand, it nonetheless appears there is still a gap towards theoretically describing neural networks at finite width. I’ll argue that the choice of learning rate is one crucial factor in dynamics away from the infinite width limit and naturally classifies deep networks into two classes separated by a sharp transition. This is elucidated in a class of solvable simple models we present, which give quantitative predictions for the two classes. Quite remarkably, we test these predictions empirically in practical settings and find excellent agreement. 

Yasaman Bahri is a research scientist on the Google Brain team. Her current research program is to build a scientific understanding of deep learning using a combination of theoretical analysis and empirical investigation. Prior to Google, she was at the University of California, Berkeley, where she received her Ph.D. in physics in 2017, specializing in theoretical quantum condensed matter.