Search Talks

We study the problem of learning the Hamiltonian of a quantum many-body system given samples from its Gibbs (thermal) state. The classical analog of this problem, known as learning graphical models or Boltzmann machines, is a well-studied question in machine learning and statistics. In this work, we give the first sample-efficient algorithm for the quantum Hamiltonian learning problem. In particular, we prove that polynomially many samples in the number of particles (qudits) are necessary and sufficient for learning the parameters of a spatially local Hamiltonian in l_2-norm.
Our main contribution is in establishing the strong convexity of the log-partition function of quantum many-body systems, which along with the maximum entropy estimation yields our sample-efficient algorithm. Classically, the strong convexity for partition functions follows from the Markov property of Gibbs distributions. This is, however, known to be violated in its exact form in the quantum case. We introduce several new ideas to obtain an unconditional result that avoids relying on the Markov property of quantum systems, at the cost of a slightly weaker bound. In particular, we prove a lower bound on the variance of quasi-local operators with respect to the Gibbs state, which might be of independent interest.

Joint work with Srinivasan Arunachalam, Tomotaka Kuwahara, Mehdi Soleimanifar

In most materials, electrons fill bands, starting from the lowest kinetic energy states. The Fermi level is the boundary between filled states below and empty states above. This is the basis for our very successful understanding of how metals and semiconductors work. But what if all the electrons within a band had the same kinetic energy (this situation is called a "flat band")? Then electrons could arrange themselves so as to minimize their Coulomb repulsion, giving rise to a wide variety of possible states including superconductors and magnets. Until recently, flat bands were achieved only by applying large magnetic fields perpendicular to a 2D electron system; in this context they are known as Landau levels. Fractional quantum hall effects result from Coulomb-driven electron arrangement within a Landau level. Recently, Pablo Jarillo-Herrero of MIT and coworkers demonstrated flat minibands in graphene-based superlattices, discovering correlated insulators and superconductors at different fillings of these minibands. We have now discovered dramatic magnetic states in such superlattice systems. Specifically, in magic-angle twisted bilayer graphene which is also aligned with a hexagonal boron nitride (hBN) cladding layer, we observe a giant anomalous Hall effect as large as 10.4 kΩ, and signs of chiral edge states. This all occurs at zero magnetic field, in a narrow density range around an apparent insulating state at 3 electrons (1 hole) per moiré cell in the conduction miniband [1]. Remarkably, the magnetization of the sample can be reversed by applying a small DC current. Although the anomalous Hall resistance is not quantized, and dissipation is significant, we suggest that the system is essentially a "Chern insulator", a type of topological insulator similar to an integer quantum Hall state. In a quite different superlattice system, ABC-trilayer graphene aligned with hBN, again near 3 electrons (1 hole) per moiré cell a Chern insulator emerges [2]. This time the flat band is a valence miniband, and a magnetic field of order 100 mT is needed to quantize the anomalous hall signal. This trilayer system can be tuned in-situ to display superconductivity instead of magnetism [3]. We will discuss possible magnetic states, complementary probes to examine which state actually emerges as the ground state in each system, and what one might do with such states.

[1] A.L. Sharpe et al., “Emergent ferromagnetism near three-quarters filling in twisted bilayer graphene”, Science 365, 6453 (2019).
[2] G. Chen et al., “Tunable Correlated Chern Insulator and Ferromagnetism in Trilayer Graphene/Boron Nitride Moire Superlattice”, Nature 579, 56 (2020)
[3] G. Chen et al., “Signatures of tunable superconductivity in a trilayer graphene moiré superlattice”, Nature 572, 215 (2019).

Entanglement entropy quantifies the amount of uncertainty of a quantum state. For quantum fields in curved space, entanglement entropy of the quantum field theory degrees of freedom is well-defined for a fixed background geometry. In this work, we propose a generalization of the quantum field theory entanglement entropy by including dynamical gravity. The generalized quantity named effective entropy, and its Renyi entropy generalizations, are defined by analytic continuation of a gravitational path integral on replica geometry with a co-dimension-2 brane at the boundary of region we are studying. We discuss different approaches to define the region in a gauge invariant way, and show that the effective entropy satisfies the quantum extremal surface formula. When the quantum fields carry a significant amount of entanglement, the quantum extremal surface can have a topology transition, after which an entanglement island region appears. Our result generalizes the Hubeny-Rangamani-Takayanagi formula of holographic entropy (with quantum corrections) to general geometries without asymptotic AdS boundary, and provides a more solid framework for addressing problems such as the Page curve of evaporating black holes in asymptotic flat spacetime. We apply the formula to two example systems, a closed two-dimensional universe and a four-dimensional maximally extended Schwarzchild black hole. We discuss the analog of the effective entropy in random tensor network models, which provides more concrete understanding of quantum information properties in general dynamical geometries. By introducing ancilla systems, we show how quantum information in the entanglement island can be reconstructed in a state-dependent and observer-dependent map. We study the closed universe (without spatial boundary) case and discuss how it is related to open universe.

Understanding galaxy formation is an outstanding problem in Astrophysics. The feedback processes that drive it, exploding stars and accretion onto supermassive black holes, are poorly understood. This results in an order unity uncertainty in the distribution of the gas inside halos, the ``missing baryon problem''. Because baryons are 15% of the total mass in the universe, this baryonic uncertainty is the largest theoretical systematics for percent precision weak lensing surveys like DES, HSC, Rubin Observatory, Roman Observatory and Euclid.
By measuring the kinematic and thermal Sunyaev-Zel'dovich effects (kSZ and tSZ), high resolution and high sensitivity CMB experiments can solve these issues by measuring the gas thermodynamics in galaxy groups and clusters, at high redshift and out to the outskirts of the halo. I will present joint tSZ, kSZ and dust measurement of BOSS (CMASS) galaxy groups, for which clustering and lensing data is also available. Using data from the Atacama Cosmology Telescope (ACT), we produced the highest significance kSZ measurement to date. This measurement shows with high statistical confidence that the gas is more spread out than the dark matter. It informs the modeling of the CMASS galaxy-galaxy lensing data, and shows that the small-scale ``lensing is low'' tension is not entirely caused by baryonic effects. Finally, comparing the observed kSZ and tSZ to hydrodynamical simulations reveals insight about the modalities of feedback.

In this talk I will describe a new link "invariant" (with certain wall-crossing properties) for links L in a three-manifold M, where M takes the form of a surface times the real line. This link "invariant" is constructed via a map, called the q-nonabelianization map, from the
gl(N) skein algebra of M to the gl(1) skein algebra of a covering three-manifold M'. In the special case of M=R^3, this map computes well-known link invariants in a new way. As a physical application, the q-nonabelianization map computes protected spin character counting BPS ground states with spin for line defects in 4d N=2 theories of class-S. I will also mention possible extension to more general three-manifolds, as well as further physical applications to class-S theories. This talk is based on joint work with Andrew Neitzke, and ongoing work with Gregory Moore and Andrew Neitzke.

Quasars are the most luminous objects in the universe powered by accretion onto supermassive black holes (SMBHs). They can be observed at the earliest cosmic epochs, providing unique insights into the early phases of black hole, structure, and galaxy formation. Observations of these quasars demonstrate that they host SMBHs at their center, already less than ~1 Gyr after the Big Bang. It has been argued that in order to grow these SMBHs in such short amounts of cosmic time, they need to accrete matter over timescales comparable to the age of the universe, and thus the lifetime of quasars - the integrated time that galaxies shine as active quasars - is expected to be of order ~10^9 yr at a redshift of z~6, even if they accrete continuously at the theoretical maximum limit. 

I will present a new method to obtain constraints on the lifetime of high-redshift quasars, based on measurements of the sizes of ionized regions around quasars, known as proximity zones. The sizes of these proximity zones are sensitive to the lifetime of the quasars, because the intergalactic gas has a finite response time to the quasars’ radiation. Applying this method to quasar spectra at z>6, we discover an unexpected population of very young quasars, indicating lifetimes of only ~10,000 years, several orders of magnitude shorter than expected. I will discuss the consequences of such short lifetimes on the quasars' ionizing power, their black hole mass accretion rates, and highlight tensions with current theoretical models for black hole formation. Furthermore, I will present several modifications to the current SMBH formation paradigm that might explain our findings and show how we aim to disentangle the various scenarios by means of future observations with the upcoming James Webb Space Telescope, in order to shed new light onto the formation and growth of the first SMBHs in the universe.

We will argue that even with semiclassical gravity, it can be shown that a copy of  all the information on a Cauchy slice resides near the boundary of the slice. We will first demonstrate this in asymptotically global AdS, and then in four-dimensional asymptotically flat space. We will then describe a physical protocol that can be used to verify this property at low-energies and within perturbation theory. This property of gravity implies that information about the black-hole interior is always present "outside" the black hole, which leads to a fresh perspective on the information paradox.

References:
1) https://arxiv.org/abs/2002.02448
2) https://arxiv.org/abs/2008.01740

We reconsider the black hole firewall puzzle, emphasizing that quantum error-correction, computational complexity, and pseudorandomness are crucial concepts for understanding the black hole interior. We assume that the Hawking radiation emitted by an old black hole is pseudorandom, meaning that it cannot be distinguished from a perfectly thermal state by any efficient quantum computation acting on the radiation alone. We then infer the existence of a subspace of the radiation system which we interpret as an encoding of the black hole interior. This encoded interior is entangled with the late outgoing Hawking quanta emitted by the old black hole, and is inaccessible to computationally bounded observers who are outside the black hole. Specifically, efficient operations acting on the radiation, those with quantum computational complexity polynomial in the entropy of the remaining black hole, commute with a complete set of logical operators acting on the encoded interior, up to corrections which are exponentially small in the entropy. Thus, under our pseudorandomness assumption, the black hole interior is well protected from exterior observers as long as the remaining black hole is macroscopic. On the other hand, if the radiation is not pseudorandom, an exterior observer may be able to create a firewall by applying a polynomial-time quantum computation to the radiation.

Abstract: Large N matrix quantum mechanics are central to holographic duality but not solvable in the most interesting cases. We show that the spectrum and simple expectation values in these theories can be obtained numerically via a `bootstrap' methodology. In this approach, operator expectation values are related by symmetries -- such as time translation and SU(N) gauge invariance -- and then bounded with certain positivity constraints. We first demonstrate how this method efficiently solves the conventional quantum anharmonic oscillator. We then reproduce the known solution of large N single matrix quantum mechanics. Finally, we present new results on the ground state of large N two matrix quantum mechanics.