2019 Simons Collaboration on Cracking the Glass Problem Annual Meeting
 Organized by

Giulio Biroli, Ph.D.CEA Saclay

David R. Reichman, Ph.D.Columbia University

Sidney Nagel, Ph.D.SteinFreiler Distinguished Service Professor in Physics, University of Chicago
The third annual meeting of the Cracking the Glass Problem collaboration took place March 7–8, 2019, at the Simons Foundation in New York. It gathered together a group of 92 researchers from the United States, Europe, Japan, Brazil and India. It featured stimulating talks by collaboration scientists and distinguished external speakers, and a poster session by the other members of the collaboration. The interactive setting provided by the Simons Foundation enabled an environment that fostered vibrant discussions among the diverse crowd of attending scientists.
The collaboration’s goal is to develop a complete and quantitative description of the glass transition, connecting the explicit and quantitative meanfield (i.e., infinite dimension, d=∞) and zerotemperature theories that have been pioneered by the collaborators. The aim is to develop a similarly quantitative and predictive theory of glassy dynamics in finite dimensions at finite temperatures. The tools developed in creating this framework are having important ramifications in a wide range of fields outside of physics, including computer science (e.g., in machine learning and optimization problems) and biology (e.g., evolutionary population dynamics and glassy behavior of tissues). Over the last two and a half years, the collaboration has obtained major breakthroughs: It obtained and explored the exact infinitedimensional solution of the dynamics of glassy hardsphere systems, provided a nearly complete characterization of a new type of amorphous phase transition — namely the Gardner transition — and introduced new powerful algorithms, such as SWAP Monte Carlo, that has opened the door to simulations of supercooled liquids in regimes completely out of reach until now. This latter breakthrough has allowed, among other advances, for the first simulation of the brittletoductile transition and provided convincing evidence of the existence of a finitetemperature ideal glass transition.
Read more of the report in the section below:

Sidney Nagel gave a “state of the collaboration” talk to kick off the meeting. He summarized the original motivation for the collaboration and the high excitement that surrounds this problem. He then gave an overview of the progress over the last two and a half years, highlighting successes of the collaboration as typified by the summary above. He also discussed new areas into which the collaboration has started to make inroads, including the glassy behavior of biological systems (highlighting, in particular, the confluent tissue rigidity transition), aspects of deep learning and material memory, and the design of novel metamaterials. In each case, he emphasized how the fundamental tools that have been invented and honed by the collaboration on the glass transition problem have enabled advances in these areas. Nagel also emphasized how the collaboration has successfully created a vibrant scientific community via the incorporation of students and postdocs into the collaboration and the creation of many workshops, tutorials, conferences and workinggroup gatherings.
Lisa Manning discussed flow and failure (rheology) in glassy systems, highlighting progress and the exciting new directions to be explored in the coming few years. After reminding the audience that rigidity is an emergent phenomenon, she discussed the collaboration’s findings concerning the unique universal vibrational properties of disordered systems, emphasizing the commonalities and distinctions between the mean field (d=∞) and behavior in two and three dimensions. She then discussed breakthroughs in formulating a meanfield theory of the jamming phase diagram and of yielding transitions. She discussed the formulation of elastoplastic models capable of describing the brittletoductile yielding transition (which was first captured in silico with the aid of SWAP Monte Carlo). Lastly, she emphasized exciting, new directions upon which the collaboration has embarked, including understanding the dynamics of yielding beyond meanfield theory and providing a microscopic understanding of localized excitations, such as twolevel systems (TLS) in lowtemperature glasses.
Andrea Liu discussed the new program within the collaboration related to harnessing our knowledge of glassy systems to understand natural phenomena, such as allostery, and to create novel metamaterials. In particular, she discussed how our understanding of the malleability of disordered systems can lead to the rational design of materials with unusual responses, such as auxetic materials that stretch in one direction when stretched in another. She discussed the implications of these ideas for biological systems, most notably proteins and blood flow in vascular systems. She discussed exciting directions for the collaboration, such as harnessing directed aging in metamaterial design and gaining a deeper understanding of nonlocal mechanical responses in proteins.
Jorge Kurchan highlighted progress in understanding the dynamics of glassy systems, emphasizing the foundation provided by the exact d=∞ solution for the dynamics of hard sphere liquids. He discussed the link with the energylandscape paradigm to understand the nature of activated processes in low space dimensions, as well as the potential use of cluster approaches to build in realspace features. He speculated on the role of reparametrization invariance in both the classical context, where it may be connected to dynamical heterogeneities, as well as in the quantum context, where it may be related to emergent properties that arise in the context of toy models of black holes.
Stimulating lectures were also given by noncollaboration members. Vincenzo Vitelli (University of Chicago) showed how driven, outofequilibrium systems generically exhibit ‘odd elasticity,’ which breaks the symmetric form of the elastic response matrix. These systems may do useful work under such conditions. Marc Mezard (École normale supérieure, Paris) discussed the connections between glassy physics and statistical inference problems, emphasizing the powerful role played by meanfield methods. Francis Hellman (University of California, Berkeley) discussed twolevel tunneling systems in glasses, a topic of much active interest within the collaboration. She showed intriguing experimental data suggesting that vapor deposition creates glasses with a paucity of such localized excitations. Following on this theme, Mark Ediger (University of WisconsinMadison) showed how vapordeposition techniques enable the synthesis of ultrastable glasses that are formed as close to the ideal glass transition temperature as possible. He outlined the synergy with the collaboration’s work where similar glasses have been created in silico using SWAP Monte Carlo.
The degree of success of the collaboration in terms of progress was aptly summarized by Vitelli at the start of his talk when he observed that he stopped working directly in the area of glass physics six years ago, and the progress made by the Cracking the Glass Problem team has been so great that he hardly recognizes the field as it stands today. Overall, the meeting highlighted the excitement generated, and the great progress made, during the tenure of the collaboration, as well as important, new directions that will pave the way for future discoveries.
Talks
Sidney Nagel
University of Chicago
The State of the Collaboration
Scientists are customarily taught to understand solid materials by treating them as ideal crystals. This approach, however, becomes untenable in an intrinsically disordered material such as a glass: A crystal is an abysmal starting point for understanding the rigidity or excitations of a glass with no obvious longrange order.
In this lecture, Sidney Nagel will discuss how physicists are working to understand glassy matter. He will explain jamming, an alternative starting point for describing solids where order, rather than disorder, is treated as a perturbation. He will then show how physicists have learned to exploit disorder to create solids with unique, varied, textured and tunable behavior including longrange interactions inspired by the allosteric behavior of proteins. Applying these techniques, however, requires computing the response of each bond between particles. Can such detailed computations be avoided? Because a material has a memory of the conditions under which it was prepared and subsequently aged, scientists can now direct aging using Nature’s (as distinct from a computer’s) socalled greedy algorithms to achieve novel kinds of mechanical functionality.
This talk will review the state of the collaboration Cracking the Glass Problem. Nagel will first review the fundamental issues about glasses, jamming and the nature of the amorphous state that initially inspired the collaboration’s work. The collaboration has made great strides in finding solutions to some of these problems, and Nagel will describe what the team has accomplished so far. These include finding new algorithms for extending simulations of glassy dynamics into new regimes, understanding the Garner phase of matter, deriving the dynamics in the meanfield limit and the extension to the creation of metamaterials. These accomplishments have opened up new opportunities for further work. Nagel will give a vision of our plans for the future and where we are going.
For further information on Professor Nagel’s Lecture, see its event page
Lisa Manning
Syracuse University
How Do Glasses Flow and Fail in Dimensions from Two to Infinity?
Solids resist shear forces, but under enough applied strain, they will yield and deform. We have recently established exact infinitedimensional solutions for amorphous solids under strain, which predict where a solid will remain stable as a function of strain and density and provide a strong foundation for understanding rheology in glasses. However, infinitedimensional theories cannot predict the temporal and spatial dynamics of deformation in amorphous solids, which vary dramatically from ductile and homogeneous flow in poorly annealed glasses to brittle fracture in deeply quenched glasses. We now understand that these processes are controlled by quasilocalized excitations that are common in 2D and 3D but disappear in higher dimensions. Analytically solvable elastoplastic models include a meanfield coupling between localized excitations. We have demonstrated that the dynamics of such models are quite similar to those in 2D and 3D numerical simulations and indicate that the brittletoductile transition is a random critical point. To make these theories quantitatively predictive, we are currently working to understand how material preparation and shear affect the population of low energy excitations, how the coupling between excitations differs from meanfield and how finite strain rates and temperatures impact these quantities.
Vincenzo Vitelli
James Franck Institute
Odd Elasticity
Hooke’s law states that the deformations or strains experienced by an elastic object are proportional to the applied forces or stresses. The number of coefficients of proportionality between stress and strain (i.e., the elastic moduli) is constrained by energy conservation. In this talk, we generalize continuum elasticity to active media with nonconservative (or nonreciprocal) microscopic interactions. This generalization, which we dub ‘odd elasticity,’ reveals that two additional elastic moduli can exist in a 2D isotropic solid with straindependent activity. Such an oddelastic solid can be regarded as a distributed engine: work is locally extracted, or injected, during quasistatic cycles of deformation. By coarse graining illustrative microscopic models, we show how odd elasticity emerges in active metamaterials composed of springs that actuate internal torques in response to strain. Our predictions, corroborated by simulations, uncover phenomena ranging from activityinduced auxetic behavior and buckling, to wave propagation powered by selfsustained active elastic cycles.
Andrea Liu
University of Pennsylvania
Mechanical Metamaterials and the Malleability of Disorder
Systems with complex landscapes have far more variation in their properties than those with simple ones. This natural variation can be pushed even further by design, allowing us to tune in unusual properties and novel functions into materials. For example, when most materials are stretched in one direction, they tend to shrink in the orthogonal directions. Materials that do the opposite and expand in the orthogonal directions when stretched are ‘auxetic’ and have attracted attention for applications such as high energy absorption. We have found that mechanical spring networks can be tuned easily to the extreme limit of auxetic behavior. Likewise, our collaboration has shown that properties common in living matter, such as the ability of proteins (e.g., hemoglobin) to change their conformations upon binding of an atom (oxygen) or molecule, the ability of the brain’s vascular network to send enhanced blood flow and oxygen to specific areas of the brain associated with a given task, or the ability to retain a memory, can be designed into disordered systems using similar principles. The ability to design properties and functions further gives new insight into the relation between microscopic structure and function that may help us both to understand living systems and to design new biologically inspired materials.
Marc Mézard
École Normale Supérieure
Statistical Physics and Statistical Inference
A major challenge of contemporary statistical inference is the largescale limit, where one wants to discover the values of many hidden parameters, using large amounts of data. In recent years, ideas from statistical physics of disordered systems, notably the cavity method, have helped to develop new algorithms for important inference problems, ranging from community detection to compressed sensing, machine learning (notably neural networks) and generalized linear regression. The talk will review these developments and explain how they can be used, together with the replica method, to identify phase transitions in benchmark statistical ensembles of inference problems.
Frances Hellman
University of California, Berkeley
Ideality and Tunneling Level Systems (TLS) in Amorphous Silicon Films
Heat capacity, sound velocity and internal friction of covalently bonded amorphous silicon (aSi) films with and without hydrogen show that lowenergy excitations, commonly called tunneling or two level systems (TLS), can be tuned over nearly three decades, from belowdetectable limits to the range commonly seen in glassy systems. This tuning is accomplished by growth temperature, thickness, growth rate, light soaking or annealing. We see a strong correlation with atomic density in aSi and in literature analysis of other glasses, as well as with dangling bond density, sound velocity and bond angle distribution as measured by Raman spectroscopy, but TLS density varies by orders of magnitude while these other measures of disorder vary by less than a factor of two. The lowest TLS films are grown at temperatures near 0.8 of the theoretical glass transition temperature of Si, similar to work on polymer films and suggestive that the high surface mobility at relatively low temperature of vapor deposition can produce materials close to an ideal glass, with higher density, lower energy and low TLS due to fewer nearby configurations with similarly low energy. The TLS measured by heat capacity and internal friction are strongly correlated for pure aSi, but not for hydrogenated aSi, suggesting that the standard TLS model works for aSi, but that aSi:H possess TLS that are decoupled from the acoustic waves measured by internal friction. Internal friction measures those TLS that introduce mechanical damping; Hellman is in the process of measuring low T dielectric loss, which yield TLS with dipole moments, in order to explore the correlation between different types of TLS. Additionally, a strong correlation is found between an excess T^{3 }term (well above the sound velocityderived Debye contribution) and the linear term in heat capacity, suggesting a common origin.
Mark Ediger
University of Wisconsin, Madison
Exploring the Limits of Amorphous Packing with VaporDeposited Glasses
Glasses formed by cooling a liquid inherit both their structure and their limited stability from the liquid state. In contrast, glasses prepared by vapor deposition can avoid both of these limitations. By utilizing the high mobility present near the free surface of many organic glasses, vapor deposition can build glasses with lowenthalpy, highdensity and highthermal stability. Based upon their position on the potential energy landscape, these materials approach “ideal glass” packing that otherwise could only be achieved by annealing a liquidcooled glass for thousands or millions of years. Vapor deposition of organic semiconductors produces glasses with improved properties for organic electronics, including the ability to produce anisotropic glasses with a wide range of structures. Remarkably, this “antiepitaxy” process uses the freesurface structure as its template, rather than the substrate structure. Recent work has shown that optimizing vapor deposition can produce organic light emitting diodes (OLEDs) that are more efficient and have extended lifetimes.
Jorg Kurchan
LPENS, Paris
Glassy Dynamics
Almost every meaningful scientific question we ask about supercooled liquids and glasses involves dynamics. If we want to control their behavior, we need to understand the way these systems flow, age or respond to external fields or violent stresses. However, there is another aspect about why the study of glassy dynamics is so basic: it has emergent and universal behavior — quantities and rules that transcend the immediate example being studied. These rules are independent of the microscopic structure of any particular material or example. Thus, what we learn by studying the dynamics in glasses and liquids teaches us lessons that are applicable far beyond the original manifestation.

Ada Altieri École Normale Supérieure Glassy features and collective phenomena in large ecosystems Marco Baity Jesi Columbia University Dynamics of glasses: from infinite to three dimensions HorstHolger Boltz University of Chicago Fluctuation distributions of energy minima in complex landscapes Carolina Brito Universidade Federal do Rio Grande do Sul Theory for Swap Acceleration near the Glass and Jamming Transition Davide Facoetti École Normale Supérieure Toy glasses and toy black holes Varda Faghir Hagh University of Chicago The Marginality Gap, From Low Dimensions to MeanField Elijah Flenner Colorado State University TBA Giampaolo Folena Università di Roma/LPTMS, Université Paris Sud Memories from the ergodic phase: the awkward dynamics of mixed pspin models Benjamin Guiselin Université de Montpellier Probing the hidden equilibrium phase transition in glassforming liquids Daniel Hexner University of Chicago Can a large packing be assembled from smaller ones? Yi Hu Duke University Dynamical transition of random Lorenz gas Harukuni Ikeda École Normale Supérieure Universality of jamming of nonspherical particles Atsushi Ikeda University of Tokyo Spatial structure of quasilocalized vibrations in nearly jammed amorphous solids Dmytro Khomenko École Normale Supérieure Statistics of TwoLevel systems in polydisperse amorphous solids. Robert Leheny Johns Hopkins University Microscopic Dynamics of Stress Relaxation in a Soft Glass Cathy Li University of Pennsylvania Marginal Stability in Meanfield and Lowdimensional Jammed Packings Peter Morse Syracuse Univesity The relationship between shear and random force in jammed systems Anshul Deep Singh Parmar Universite de Montpellier Knocking at the bottom of the energy landscape: Extended KobAndersen model Nidhi Pashine University of Chicago Directed aging and memory: Teaching an old foam new tricks Marko Popovic EPFL Does inertia kill criticality in depinning and friction? Mark Robbins Johns Hopkins University Probing large viscosities in glass formers with nonequilibrium simulations Valentina Ros École Normale Supérieure Geometrical properties of highdimensional landscapes: local minima, barriers and topology trivialization transitions Grant Rotskoff Courant Institute Global convergence of neuron birthdeath dynamics Miguel Ruiz Garcia University of Pennsylvania Tunning and jamming reduced to their minima Hajime Yoshino Osaka University A stabilityreversibility map unifies elasticity, plasticity and jamming in hardsphere glasses Ge Zhang University of Pennsylvania Interplay of softness and rearrangements during avalanche propagation