MIT Center for Theoretical Physics — a Leinweber Institute Publications

We explore the relationship between renormalization group (RG) flow and error correction by constructing quantum algorithms that exactly recognize 1D symmetry-protected topological (SPT) phases protected by finite internal Abelian symmetries. For each SPT phase, our algorithm runs a quantum circuit which emulates RG flow: an arbitrary input ground state wavefunction in the phase is mapped to a unique minimally-entangled reference state, thereby allowing for efficient phase identification. This construction is enabled by viewing a generic input state in the phase as a collection of coherent `errors’ applied to the reference state, and engineering a quantum circuit to efficiently detect and correct such errors. Importantly, the error correction threshold is proven to coincide exactly with the phase boundary. We discuss the implications of our results in the context of condensed matter physics, machine learning, and near-term quantum algorithms.

The geometry of 4-string contact interaction of closed string field theory is characterized using machine learning. We obtain Strebel quadratic differentials on 4-punctured spheres as a neural network by performing unsupervised learning with a custom-built loss function. This allows us to solve for local coordinates and compute their associated mapping radii numerically. We also train a neural network distinguishing vertex from Feynman region. As a check, 4-tachyon contact term in the tachyon potential is computed and a good agreement with the results in the literature is observed. We argue that our algorithm is manifestly independent of number of punctures and scaling it to characterize the geometry of \(n\)-string contact interaction is feasible.

We prove novel speed limits on the growth of entanglement, equal time correlators, and spacelike Wilson loops in spatially uniform time-evolving states in strongly coupled CFTs with holographic duals. These bounds can also be viewed as quantum weak energy conditions. Several of the speed limits are valid for regions of arbitrary size and with multiple connected components, and our findings imply new bounds on the effective entanglement velocity of small subregions. In 2d CFT, our results prove a conjecture by Liu and Suh for a large class of states. Key to our findings is a momentum-entanglement correspondence, showing that entanglement growth is computed by the momentum crossing the HRT surface. In our setup, we prove a number of general features of boundary-anchored extremal surfaces, such as a sharp bound on the smallest radius that a surface can probe, and that the tips of extremal surfaces cannot lie in trapped regions. Our methods rely on novel global GR techniques, including a delicate interplay between Lorentzian and Riemannian Hawking masses. While our proofs assume the dominant energy condition in the bulk, we provide numerical evidence that our bounds are true under less restrictive assumptions.

Recent applications of machine-learned normalizing flows to sampling in lattice field theory suggest that such methods may be able to mitigate critical slowing down and topological freezing. However, these demonstrations have been at the scale of toy models, and it remains to be determined whether they can be applied to state-of-the-art lattice quantum chromodynamics calculations. Assessing the viability of sampling algorithms for lattice field theory at scale has traditionally been accomplished using simple cost scaling laws, but as we discuss in this work, their utility is limited for flow-based approaches. We conclude that flow-based approaches to sampling are better thought of as a broad family of algorithms with different scaling properties, and that scalability must be assessed experimentally.

We study the properties of three-body resonances using a lattice complex scalar \(\varphi^4\) theory with two scalars, with parameters chosen such that one heavy particle can decay into three light ones. We determine the two- and three-body spectra for several lattice volumes using variational techniques, and then analyze them with two versions of the three-particle finite-volume formalism: the Relativistic Field Theory approach and the Finite-Volume Unitarity approach. We find that both methods provide an equivalent description of the energy levels, and we are able to fit the spectra using simple parametrizations of the scattering quantities. By solving the integral equations of the corresponding three-particle formalisms, we determine the pole position of the resonance in the complex energy-plane and thereby its mass and width. We find very good agreement between the two methods at different values of the coupling of the theory.

Gauge Theory plays a crucial role in many areas in science, including high energy physics, condensed matter physics and quantum information science. In quantum simulations of lattice gauge theory, an important step is to construct a wave function that obeys gauge symmetry. In this paper, we have developed gauge equivariant neural network wave function techniques for simulating continuous-variable quantum lattice gauge theories in the Hamiltonian formulation. We have applied the gauge equivariant neural network approach to find the ground state of 2+1-dimensional lattice gauge theory with U(1) gauge group using variational Monte Carlo. We have benchmarked our approach against the state-of-the-art complex Gaussian wave functions, demonstrating improved performance in the strong coupling regime and comparable results in the weak coupling regime.

Finding efficient optimization methods plays an important role for quantum optimization and quantum machine learning on near-term quantum computers. While backpropagation on classical computers is computationally efficient, obtaining gradients on quantum computers is not, because the computational complexity usually scales with the number of parameters and measurements. In this paper, we connect Koopman operator theory, which has been successful in predicting nonlinear dynamics, with natural gradient methods in quantum optimization. We propose a data-driven approach using Koopman operator learning to accelerate quantum optimization and quantum machine learning. We develop two new families of methods: the sliding window dynamic mode decomposition (DMD) and the neural DMD for efficiently updating parameters on quantum computers. We show that our methods can predict gradient dynamics on quantum computers and accelerate the variational quantum eigensolver used in quantum optimization, as well as quantum machine learning. We further implement our Koopman operator learning algorithm on a real IBM quantum computer and demonstrate their practical effectiveness.

We study Euclidean M5-branes wrapping vertical divisors in elliptic Calabi-Yau fourfold compactifications of M/F-theory that admit a Sen limit. We construct these Calabi-Yau fourfolds as elliptic fibrations over coordinate flip O3/O7 orientifolds of toric hypersurface Calabi-Yau threefolds. We devise a method to analyze the Hodge structure (and hence the dimension of the intermediate Jacobian) of vertical divisors in these fourfolds, using only the data available from a type IIB compactification on the O3/O7 Calabi-Yau orientifold. Our method utilizes simple combinatorial formulae (that we prove) for the equivariant Hodge numbers of the Calabi-Yau orientifolds and their prime toric divisors, along with a formula for the Euler characteristic of vertical divisors in the corresponding elliptic Calabi-Yau fourfold. Our formula for the Euler characteristic includes a conjectured correction term that accounts for the contributions of pointlike terminal \(\mathbb{Z}_2\) singularities corresponding to perturbative O3-planes. We check our conjecture in a number of explicit examples and find perfect agreement with the results of direct computations.

A number of proposals have been put forward for detecting axion dark matter (DM) with grand unification scale decay constants that rely on the conversion of coherent DM axions to oscillating magnetic fields in the presence of static, laboratory magnetic fields. Crucially, such experiments \(\unicode{x2013}\) including ABRACADABRA \(\unicode{x2013}\) have to-date worked in the limit that the axion Compton wavelength is larger than the size of the experiment, which allows one to take a magnetoquasistatic (MQS) approach to modeling the axion signal. We use finite element methods to solve the coupled axion-electromagnetism equations of motion without assuming the MQS approximation. We show that the MQS approximation becomes a poor approximation at frequencies two orders of magnitude lower than the naive MQS limit. Radiation losses diminish the quality factor of an otherwise high-\(Q\) resonant readout circuit, though this may be mitigated through shielding and minimizing lossy materials. Additionally, self-resonances associated with the detector geometry change the reactive properties of the pickup system, leading to two generic features beyond MQS: there are frequencies that require an inductive rather than capacitive tuning to maintain resonance, and the detector itself becomes a multi-pole resonator at high frequencies. Accounting for these features, competitive sensitivity to the axion-photon coupling may be extended well beyond the naive MQS limit.

Large language models with a huge number of parameters, when trained on near internet-sized number of tokens, have been empirically shown to obey neural scaling laws: specifically, their performance behaves predictably as a power law in either parameters or dataset size until bottlenecked by the other resource. To understand this better, we first identify the necessary properties allowing such scaling laws to arise and then propose a statistical model — a joint generative data model and random feature model — that captures this neural scaling phenomenology. By solving this model in the dual limit of large training set size and large number of parameters, we gain insight into (i) the statistical structure of datasets and tasks that lead to scaling laws, (ii) the way nonlinear feature maps, such as those provided by neural networks, enable scaling laws when trained on these datasets, (iii) the optimality of the equiparameterization scaling of training sets and parameters, and (iv) whether such scaling laws can break down and how they behave when they do. Key findings are the manner in which the power laws that occur in the statistics of natural datasets are extended by nonlinear random feature maps and then translated into power-law scalings of the test loss and how the finite extent of the data’s spectral power law causes the model’s performance to plateau.

Self-interacting dark matter (SIDM) offers the potential to mitigate some of the discrepancies between simulated cold dark matter (CDM) and observed galactic properties. We introduce a physically motivated SIDM model to understand the effects of self interactions on the properties of Milky Way and dwarf galaxy sized haloes. This model consists of dark matter with a nearly degenerate excited state, which allows for both elastic and inelastic scattering. In particular, the model includes a significant probability for particles to up-scatter from the ground state to the excited state. We simulate a suite of zoom-in Milky Way-sized N-body haloes with six models with different scattering cross sections to study the effects of up-scattering in SIDM models. We find that the up-scattering reaction greatly increases the central densities of the main halo through the loss of kinetic energy. However, the physical model still results in significant coring due to the presence of elastic scattering and down-scattering. These effects are not as apparent in the subhalo population compared to the main halo, but the number of subhaloes is reduced compared to CDM.

Large-time correlation functions have a pivotal role in extracting particle masses from Euclidean lattice field theory calculations, but little is known about the statistical properties of these quantities. In this work, the distributions of correlation functions at vanishing momentum are calculated exactly in the case of free real scalar field theory. It is further demonstrated that the results found for this simple case are valid for all local and hermitian operators in bosonic systems with a unique gapped vacuum at large Euclidean-time.

We propose a new mechanism to dynamically select the electroweak scale during inflation. An axion-like field \(\phi\) that couples quadratically to the Higgs with a large initial velocity towards a critical point \(\phi_c\) where the Higgs becomes massless. When \(\phi\) crosses this point, it enters a region where the Higgs mass is tachyonic and this results into an explosive production of Higgs particles. Consequently, a back-reaction potential is generated and the field \(\phi\) is attracted back to \(\phi_c\). After a series of oscillations around this point it is eventually trapped in its vicinity due to the periodic term of the potential. The model avoids transplanckian field excursions, requires very few e-folds of inflation and it is compatible with inflation scales up to \(10^5~\rm GeV\). The mass of \(\phi\) lies in the range of hundreds of GeV to a few TeV and it can be potentially probed in future colliders.

We use the recently developed methods of 2108.07810 to analyze vertical flux backgrounds and associated chiral matter spectra in the 4D universal \((\text{SU}(3) \times \text{SU}(2) \times \text{U}(1))/\mathbb{Z}_{6}\) model introduced in 1912.10991, which is believed to describe the most general generic family of F-theory vacua with tuned \((\text{SU}(3) \times \text{SU}(2) \times \text{U(}1))/\mathbb{Z}_{6}\) gauge symmetry. Our analysis focuses on a resolution of a particular presentation of the \((\text{SU}(3) \times \text{SU}(2) \times \text{U}(1))/\mathbb{Z}_{6}\) model in which the elliptic fiber is realized as a cubic in \(\mathbb{P}^2\) fibered over an arbitrary smooth threefold base. We show that vertical fluxes can produce nonzero multiplicities for all chiral matter families that satisfy 4D anomaly cancellation, which include as a special case the chiral matter families of the Minimal Supersymmetric Standard Model.

Understanding the detailed structure of energy flow within jets, a field known as jet substructure, plays a central role in searches for new physics, and precision studies of QCD. Many applications of jet substructure require an understanding of jets initiated by heavy quarks, whose description has lagged behind remarkable recent progress for massless jets. In this Letter, we initiate a study of correlation functions of energy flow operators on beauty and charm jets to illuminate the effects of the intrinsic mass of the elementary particles of QCD. We present a factorization theorem incorporating the mass of heavy quarks, and show that the heavy quark jet functions for energy correlators have a simple structure in perturbation theory. Our results achieve the very first full next-to-leading-logarithmic calculation of the heavy quark jet substructure observable at the LHC. Using this framework, we study the behavior of the correlators, and show that they exhibit a clear transition from a massless scaling regime, at precisely the scale of the heavy quark mass. This manifests the long-sought-after dead-cone effect and illustrates fundamental effects from the intrinsic mass of beauty and charm quarks in a perturbative regime, before they are confined inside hadrons. Our theoretical framework for studying energy correlators using heavy jets has many exciting applications for improving the description of mass effects in next generation parton shower event generators, probing the QGP, and studying heavy flavor fragmentation functions.

We explore machine learning-based jet and event identification at the future Electron-Ion Collider (EIC). We study the effectiveness of machine learning-based classifiers at relatively low EIC energies, focusing on (i) identifying the flavor of the jet and (ii) identifying the underlying hard process of the event. We propose applications of our machine learning-based jet identification in the key research areas at the future EIC and current Relativistic Heavy Ion Collider program, including enhancing constraints on (transverse momentum dependent) parton distribution functions, improving experimental access to transverse spin asymmetries, studying photon structure, and quantifying the modification of hadrons and jets in the cold nuclear matter environment in electron-nucleus collisions. We establish first benchmarks and contrast the estimated performance of flavor tagging at the EIC with that at the Large Hadron Collider. We perform studies relevant to aspects of detector design including particle identification, charge information, and minimum transverse momentum capabilities. Additionally, we study the impact of using full event information instead of using only information associated with the identified jet. These methods can be deployed either on suitably accurate Monte Carlo event generators, or, for several applications, directly on experimental data. We provide an outlook for ultimately connecting these machine learning-based methods with first principles calculations in quantum chromodynamics.

We compute the soft-drop jet-mass distribution from \(pp\) collisions to NNLL accuracy while including nonperturbative corrections through a field-theory based formalism. Using these calculations, we assess the theoretical uncertainties on an \(\alpha_s\) precision measurement due to higher order perturbative effects, nonperturbative corrections, and PDF uncertainty. We identify which soft-drop parameters are well-suited for measuring \(\alpha_s\), and find that higher-logarithmic resummation has a qualitatively important effect on the shape of the jet-mass distribution. We find that quark jets and gluon jets have similar sensitivity to \(\alpha_s\), and emphasize that experimentally distinguishing quark and gluon jets is not required for an \(\alpha_s\) measurement. We conclude that measuring \(\alpha_s\) to the 10% level is feasible now, and with improvements in theory a 5% level measurement is possible. Getting down to the 1% level to be competitive with other state-of-the-art measurements will be challenging.

We introduce a \(\mathbb{Z}_N\) stabilizer code that can be defined on any spatial lattice of the form \(\Gamma\times C_{L_z}\), where \(\Gamma\) is a general graph. We also present the low-energy limit of this stabilizer code as a Euclidean lattice action, which we refer to as the anisotropic \(\mathbb{Z}_N\) Laplacian model. It is gapped, robust (i.e., stable under small deformations), and has lineons. Its ground state degeneracy (GSD) is expressed in terms of a “mod \(N\)-reduction” of the Jacobian group of the graph \(\Gamma\). In the special case when space is an \(L\times L\times L_z\) cubic lattice, the logarithm of the GSD depends on \(L\) in an erratic way and grows no faster than \(O(L)\). We also discuss another gapped model, the \(\mathbb{Z}_N\) Laplacian model, which can be defined on any graph. It has fractons and a similarly strange GSD.

The quantum-mechanical description of assemblies of particles whose motion is confined to two (or one) spatial dimensions offers many possibilities that are distinct from bosons and fermions. We call such particles anyons. The simplest anyons are parameterized by an angular phase parameter \(\theta\). \(\theta = 0, \pi\) correspond to bosons and fermions respectively; at intermediate values we say that we have fractional statistics. In two dimensions, \(\theta\) describes the phase acquired by the wave function as two anyons wind around one another counterclockwise. It generates a shift in the allowed values for the relative angular momentum. Composites of localized electric charge and magnetic flux associated with an abelian U(1) gauge group realize this behavior. More complex charge-flux constructions can involve non-abelian and product groups acting on a spectrum of allowed charges and fluxes, giving rise to nonabelian and mutual statistics. Interchanges of non-abelian anyons implement unitary transformations of the wave function within an emergent space of internal states. Anyons of all kinds are described by quantum field theories that include Chern–Simons terms. The crossings of one-dimensional anyons on a ring are uni-directional, such that a fractional phase \(\theta\) acquired upon interchange gives rise to fractional shifts in the relative momenta between the anyons. The quasiparticle excitations of fractional quantum Hall states have long been predicted to include anyons. Recently the anyon behavior predicted for quasiparticles in the \(\nu = 1/3\) fractional quantum Hall state has been observed both in scattering and in interferometric experiments. Excitations within designed systems, notably including superconducting circuits, can exhibit anyon behavior. Such systems are being developed for possible use in quantum information processing.

Quantum neural networks have been widely studied in recent years, given their potential practical utility and recent results regarding their ability to efficiently express certain classical data. However, analytic results to date rely on assumptions and arguments from complexity theory. Due to this, there is little intuition as to the source of the expressive power of quantum neural networks or for which classes of classical data any advantage can be reasonably expected to hold. Here, we study the relative expressive power between a broad class of neural network sequence models and a class of recurrent models based on Gaussian operations with non-Gaussian measurements. We explicitly show that quantum contextuality is the source of an unconditional memory separation in the expressivity of the two model classes. Additionally, as we are able to pinpoint quantum contextuality as the source of this separation, we use this intuition to study the relative performance of our introduced model on a standard translation data set exhibiting linguistic contextuality. In doing so, we demonstrate that our introduced quantum models are able to outperform state of the art classical models even in practice.