Simons Society of Fellows Retreat 2026

Hector Afonso Cruz,
New York Univeristy

The First Billion Years of Cosmic Sound

Future observations of 21-cm fluctuations will hold a wealth of information about the rise of the first luminous sources during cosmic dawn and their cosmological initial conditions. To probe the universe at such early times, we rely on mapping the 21-cm transition from neutral hydrogen, whose clumpiness against the background of star forming galaxies will track the state of the intergalactic medium (IGM) and the underlying dark matter density field. At such early times, two counteractive effects—density fluctuations and streaming velocities between cold dark matter and baryons—influence the spatial distribution of star formation. While matter density fluctuations dictate where galaxy-forming haloes eventually form, streaming velocities seeded at kinematic decoupling suppress star formation. Together, their spatial modulation and influence on the thermal and ionization evolution of the IGM yield two-point statistics with both baryon and velocity-induced acoustic features. I will show how to disentangle the complex astrophysics from cosmology at such early times with the first analytical decomposition of matter and velocity in the cosmic dawn 21-cm signal. As streaming velocities most impact lower mass haloes, I demonstrate how tracking the time evolution of acoustic amplitudes will unlock the ingredients of early galaxy evolution and the spectroscopic properties of the first stellar populations. Since both acoustic peaks can be used as standard rulers, I will show how assumptions of a density-only modulated 21cm signal may lead to erroneous estimates on cosmic expansion rates. As next generation experiments go online, these new analytical techniques will be paramount in connecting theory to data, which are poised to reveal how dark matter haloes foster the formation and evolution of their constituent galaxies across multiple epochs of cosmic time.

Andrea Alu,
City University of New York

Extreme Wave Phenomena with Metamaterials

Artificial materials engineered at the nanoscale, known as metamaterials, offer unprecedented control over light and sound propagation. In this talk, I discuss recent developments in metamaterials research, in particular emphasizing the role of broken symmetries in establishing emerging optical and acoustic material responses. Geometrical rotations and tailored perturbations in space and time can be engaged to engineer wave properties and enable new material responses ideally suited for practical technologies, from imaging, energy and sensing, to computing and communications.

Andrew Bahle,
New York University School of Medicine

Cortical Sequences Underlying Vocalization in Parrots

Complex vocal communication has arisen several times over the course of evolution and relies on conserved neural structures in the brain. Our own vocalizations are characterized by dazzlingly complex sets of syntactic, semantic, and statistical rules; properties that have been difficult or impossible to study in tractable model organisms, where the translational potential of comparative studies would be immense.

In an effort to bridge this gap, I will present work on the vocal behavior of freely behaving budgerigars (a small species of parrot) during their natural social interactions. We perform the first large-scale recordings of neural activity from a premotor area known as the central nucleus of the caudal nidopallium (NLc), an area previously implicated in the production of mimicked English words. Using unsupervised machine learning techniques, we identify a large repertoire of discrete vocal units and find that the usage of these vocal units is profoundly disrupted by lesions of NLc.

Recordings from populations of neurons in NLc further reveal that this brain region generates unique sequences of sparse activity that encode each vocal type. We propose that the dynamics within NLc provide a high-dimensional dynamical motor plan that specifies each sound in the bird’s repertoire and allows the rapid learning of new sounds. Our observations provide a promising path toward comparative neurophysiological studies of high-level features of human speech and toward understanding how vocal communication systems are shaped by the brain.

Angus Beane,
New York University

Rising from the Ashes: Deciphering the Milky Way’s Ancient past from Its Stars

Galaxies routinely die, and the Milky Way is no exception – at the present, the rate at which it is forming stars is steadily declining. However, we will argue that it suffered a violent episode in its distant past in which it died, abruptly ceasing star formation for ~300 million years (a brief period from the perspective of a galaxy), before quickly reviving. The evidence for this episode is found in an unlikely place – the distribution of the chemical compositions of stars. This distribution has a clear bimodality in elements produced in the explosions of the most massive stars. We will show that this chemical bimodality is a natural consequence of a brief halt in star formation about eight billion years ago.

Sydney Blattman,
Sloan Kettering Institute for Cancer Research

Scalable Genotyping in Fixed Transcriptomes Resolves Clonal Heterogeneity Via Single-Cell Sequencing

Despite mostly sharing the same genome, cells of the same organism are dynamic entities that express genes at different levels and constantly evolve genetic mutations. During both normal development and cancer progression, cells can assume dramatically different gene expression states through regulation of modules that determine cell type (e.g. immune or neuronal cell) and cell state, which reflects the active processes in a cell at a given time. Over the last decade, single-cell transcriptomics has revolutionized our understanding of these heterogenous cell populations by enable high-resolution measurement of gene expression across thousands to millions of single cells. However, it has been challenging to associate these transcriptional states, or phenotypes, with genotypes of these cells. Crucially, cancer is often driven by somatic mutations, genetic variants that emerge after conception and thus affect only a subset of an organism’s cells, so linking genetic

variation to gene expression in single cells is critical to understand disease progression and treatment resistance. To date, technical limitations of widely-used platforms have precluded association of cellular states with somatic mutations at scale. We introduce Genotyping in Fixed Transcriptomes (GIFT), a probe-based technology for highly accurate detection of hundreds of targeted genetic variants alongside whole single-cell transcriptomes. We demonstrate the ease and flexibility of custom probes to extend commercial single-cell platforms to profile mutations across diverse contexts from hematopoiesis to tumorigenesis and metastasis. We anticipate that GIFT will have broad utility in revealing genotype-to-phenotype relationships, including enabling tracking of genetic clones alongside comprehensive cell state measurements.

Adam Burnett,
New York Univeristy

Modern Climate Model Research: Extreme Events and Data-Driven Tuning

Climate models simulate the complex behavior of Earth’s atmosphere and oceans, and they project how this behavior evolves under ongoing anthropogenic climate change. Many aspects of climate change and its impacts are already clear. Others are less well understood, including how climate change influences damaging extreme weather events such as hurricanes. Models disagree on how the mean number of hurricanes per year may change, and we lack a theory predicting hurricane frequency based on large-scale atmospheric dynamics. I use simplified climate model simulations to investigate this relationship, and I find a few clues toward such a theory. We also can better assess extreme events by improving climate models directly. Some important phenomena, such as clouds and convection, occur at scales too small for climate models to simulate, so their effects must be approximated. These approximations include tunable parameter values which, though chosen based on observations and theory, harbor substantial uncertainty. I use statistical calibration methods to constrain uncertainty in convection parameters, which are relevant to hurricanes and other extreme precipitation patterns. This approach combines data-driven model tuning with the need to understand how climate change affects extreme weather, which will continue to be crucial in the coming decades.

Charles Dowell,
Rockefeller University

Where’s That Smell? How Fruit Flies Find Their Way Into Your Kitchen

Locating the origin of an odor source is vital for animals to find food, mates and sanctuary in an ever-changing environment. Odors are carried long distances by turbulent air streams and carry no directional information, meaning animals must integrate sensory inputs from multiple sources over varying timescales to structure their searches. Recent work in our lab has identified an olfactory navigation algorithm in the fruit fly Drosophila, whereby flies ascend a food odor plume for long distances (over 5 m) along its boundary by surging upwind while inside the plume and making memory guided returns while outside. By measuring activity in Drosophila navigational brain centers, I have identified neurons that encode a fly’s orientation when they last encountered odor, providing an instructive signal to guide searches back to the plume. Using the wide array of tools available in Drosophila, I aim to reveal the detailed circuit mechanisms by which these patterns of neural activity are formed and utilized by flies to navigate towards sources of food – like your kitchen. Thus, shedding light on how animals utilize their extraordinary sense of smell to structure search behaviors crucial to their survival and evolutionary success.

Asma Farhat,
Icahn School of Medicine at Mount Sinai

Decoding How Early-Life Neuroimmune Crosstalk at Barrier Tissues Shapes Sensory Function

Beyond serving as a protective barrier, the skin is the body’s primary sensory interface with the external environment, densely innervated by neurons that detect touch, temperature, pain, and itch. During early postnatal life, skin-innervating sensory neurons undergo rapid functional maturation alongside a developing immune system encountering its first environmental exposures, creating a period of dynamic neuroimmune interaction. Inflammatory skin disorders such as eczema arise predominantly during early childhood and are accompanied by heightened sensory symptoms, including chronic itch and pain, that substantially impair quality of life. However, whether inflammatory signals encountered during this developmental period shape sensory neuron maturation and long-term somatosensory function remains unknown. Using a neonatal model of allergic skin inflammation that recapitulates key features of early-life eczema, this work investigates how early-life inflammatory exposure shapes neuronal maturation and cutaneous innervation to impact touch sensitivity and sensory function. By defining how postnatal barrier inflammation programs lasting neuroimmune states, this research aims to uncover mechanisms linking early inflammatory experiences to altered sensory processing and neurodevelopmental outcomes.

Glennys Farrar,
New York Univeristy

The Highest Energy Particles in the Universe

Cosmic Rays are relativistic particles that rain down on the Earth from the cosmos. The most energetic ones arrive at a rate below 1 per square kilometer per century, and have millions of times higher energy than achieved by humans (in the Large Hadron Collider at CERN).   The origin and acceleration mechanism of these Ultrahigh Energy Cosmic Rays has been a mystery since their discovery over six decades ago.  Recently a compelling picture has emerged, and it turns out to be connected with another mystery only solved in recent times:  the origin of precious metals such as platinum and gold.  I will explain how merging neutron stars are responsible for both UHECRs and precious metals, and what will constitute definitive proof of this scenario.

Kevin Hu,
Columbia University

Propagation of Chaos in Disordered Systems

Large disordered systems are commonly used to model complex phenomena. Systems this large are not amenable to direct analysis, and often it is a good idea to average over the degrees of freedom to obtain a tractable macroscopic description. A mathematically precise version of this idea, which is known as ‘propagation of chaos’, has been tremendously useful in the study of highly symmetric systems with weak interaction. However the presence of disorder complicates matters, and in fact one expects to see phenomena not predicted by classical propagation of chaos results. In this talk, I will discuss recent mathematical progress on propagation of chaos in disordered systems in joint work with Arsene, Lacker, and Ramanan.

Ella King,
New York Univeristy

Emergent Activity: How Particles Get Out of Ded in the Morning

Active matter, a form of matter whose components consume energy and convert it to motion, can be viewed as a minimal model of living systems. Studying the properties of active matter has spurred the discovery of new materials and uncovered novel dynamical phase transitions. However, this form of matter lacks one crucial feature of life: active matter can’t die. Here, we uncover emergent activity, a new class of matter that can transition between passive and active states. While individual particles remain stationary, small clusters of particles gain the ability to transduce energy from the environment and use it to propel their motion.

Christopher Lafferty,
NYU Langone Health

TBA

Elise LePage,
Columbia University

Knot Invariants and String Theory

Usually, one expects new developments in mathematics to lead to breakthroughs in physics. However, it’s often the other way around. I will discuss one such example where a topological field theory originating from string theory has allowed us to define new invariants of mathematical knots. I’ll start by introducing string theory and the resulting topological field theory. Then, I’ll explain what knot invariants are, why they’re important, and how topological field theory naturally computes them.

Noam Mazor,
New York Univeristy

Computational Analogs of Randomness

Computational analogs of information-theoretic notions have given rise to some of the most intriguing phenomena in theoretical computer science. For example, pseudorandomness allows us to bypass Shannon’s lower bounds on the key length of encryption schemes. Moreover, computational analogs of entropy and randomness are key tools in the construction of pseudorandom generators and have become foundational concepts in complexity theory and cryptography. Despite its significance, fundamental questions about pseudorandomness remain open. In this talk, we will explore these questions along with recent advancements.

Leslie Sibener,
Rockefeller University

Investigating Selection Mechanisms Gating Long Term Memory

Each day we are posed with the opportunity to form new memories as we navigate environments and meet new people. But a critical question stands; how is a memory selected for long term storage in the brain, while others are forgotten? Recent work showed that the anterior thalamus (ANT) serves as a gate in selecting and stabilizing hippocampal memories to long-term cortical storage, but it is unclear what occurs locally in ANT to operate the gate. To dissect possible mechanisms, we designed a contextual memory-guided task to test how two strategies – value and repetition – may be used for memory selection. In the task, head-fixed mice navigate a virtual reality environment where they self-initiate trials to pass through four unique multisensory contexts. After learning occurs, mice are tested on memory recall at recent and remote timepoints to assess memory performance. With this task, we are exploring how brainstem neuromodulatory signaling in ANT plays a role in memory selection.

Alexis Stutzman,
Columbia University

Sniffing Out the Molecular Basis of Stress-Induced Olfactory Receptor Choice

Fear learning induces heritable changes to sensory systems that influences future behavior and increases chances of survival. Our work investigates how stressful experiences are interpreted by the body, and how the body passes on that information to ensure their offspring’s survival. Our lab recently demonstrated that fear learning in rodents results in changes to their main olfactory epithelium (MOE), the sensory organ responsible for smell. In particular, the olfactory sensory neurons that are activated by the fear-associated odor increase in number following fear learning. Moreover, this change is inherited by naïve offspring who have never encountered the fear-associated odor (Liff et al., 2025). This work is the first to suggest a heritable mechanism by which behavior is altered due to a stressor, whereby cell fate is biased during sensory system development. The molecular mechanism by which fear learning heritably biases olfactory sensory neuron identity remains unclear.

To investigate the molecular mechanism by which stressed animals induce development of new neurons that are activated by a fear-associated odor, we assessed transcriptional signatures in the MOE of stressed parents over a time course following fear learning. Our sequencing experiments indicate few changes to the olfactory sensory neurons that recognize the fear-associated scent but revealed key changes to the rest of the olfactory system that are important for maintaining memory of the fear-associated odor, including the previously reported development of new neurons that recognize the fear-associated scent. Our results provide foundational data that will shed light on an understudied mechanism of inheritance, which explains how a parent’s stressful experience can lead to biological changes in their offspring.

Hamish Swanson,
City University of New York

Developing a Universal Assembly Code for Peptide Materials Using Sequence Context and Molecular Conformations

Short peptides, which are sequence defined strings of amino acids, can be used to make materials with diverse properties, including humidity responsiveness, gel formation and adhesion; as well as perform manifold functions, including biomolecule encapsulation and molecular recognition. Peptide materials hold much promise as designed materials that are both biodegradable and fully renewable.

No robust framework currently exists with which to design peptides de novo for desired properties or functions and this represents a major bottleneck in their more widespread adoption as materials. This is in part because of the vastness of the sequence space obtained through the unique combination of the twenty naturally occurring amino acids (sequence) and controlled through their wider chemical environment (context, e.g., pH, buffer conditions, humidity).

To address this design challenge, we are developing a universal molecular syntax to rationally connect the discreet conformational ensembles of peptides, which are informed by amino acid sequence, to their collective intermolecular interactions and in turn their material properties. Practically this involves the development of simple statistical mechanics-based models of molecular conformational ensembles, as measured using molecular dynamics (MD) simulations, and close collaboration with experimental researchers to test specific hypotheses and rationalize observed functions.

In pursuing this strategy, we have found excellent agreement between conformational parameters derived from the simulation of all 400 dipeptides, and reported examples with distinct phase properties, demonstrating distinct regimes of assembly with varying degrees of order and dynamics. Recently, extension of these techniques towards tripeptides reveals a correlation between a sequence’s conformational diversity and the subsequent kinetics of crystallization. Moving forward we will leverage high throughput experimentation, in combination with statistical methods and machine learning, to generalize these learnings into a predictive framework for material design.

Thomas Werkmeister,
Columbia University

Electrons in Flatland: Anyons in Graphene

Electrons are essential to our everyday life, since modern civilization relies on our ability to manipulate electrons to store and process information via computation. However, we are now at the beginning of a new technological era in which we utilize the full quantum properties of electrons. My research focuses on creating high-quality quantum electronic devices to test our understanding of fundamental physics and develop new technologies, particularly those relevant to quantum computing and sensing. In this talk, I will describe the behavior of interacting electrons confined to a two-dimensional plane in large magnetic fields, where they form quantum states at low temperatures that support emergent particles called anyons, whose properties strikingly differ from the constituent electrons. Unlike conventional bosons and fermions, anyons can be used to manipulate and encode information through a pair-wise exchange process called “braiding”, which has been theoretically shown to enable fault-tolerant quantum computing. I will motivate and describe my recent experiments that demonstrated anyon braiding in graphene: a two-dimensional lattice of carbon atoms.

Jessica Zung,
Columbia University

A Fly’s-Eye View: Using Naturalistic Stimuli to Understand Visual Feature Processing 

The fundamental task of a brain is to take in sensory information from the environment and use it to drive an appropriate behavior. However, it would be far too energetically costly to encode the entirety of the sensory environment. Thus, a critical task of a sensory system is compression of information into compact, meaningful signals. In my work, I aim to uncover how a brain decomposes a complex sensory world into behaviorally relevant features. Despite the importance of feature processing, it is difficult to study in vertebrate brains because of their complexity and size. I address this problem in the much more tractable genetic model organism Drosophila melanogaster. Not only do we have the genetic tools in Drosophila to manipulate specific neurons, we now also have access to a whole-brain connectome—a comprehensive map of all neurons in the brain and the synaptic connections among them. These resources allow deep mechanistic insight into Drosophila neural circuits. My work focuses on a set of neurons called the LC (lobula columnar) neurons, which respond to complex visual features such as small moving shapes and rapidly approaching objects. LCs exist in several dozen distinct types, each of which appear to be tuned to different visual features. I will conduct a systematic survey of LC tuning, allowing us to understand for the first time the full repertoire of mid-level feature detectors used by an animal to process its visual environment. Importantly, my survey will use a set of stimuli that capture more of the complex visual characteristics of the natural world than the simple, mathematically defined shapes conventionally used to study fly vision. I have developed a pipeline to record flies and other animals interacting in a semi-natural environment and to then use motion-capture techniques to reconstruct the scene from a fly’s point of view. These “fly’s-eye view” reconstructions will serve as a feature-rich naturalistic stimulus set to probe LC tuning. Overall, my work will help reveal how brains compress complex sensory input to be encoded by a relatively small set of feature channels.