arXiv subjects

Marc Serra-Garcia

Publications and source records attributed to Marc Serra-Garcia.

At least 19 recordsLinked to original sources

arXiv2026

Local Learning Rules for Out-of-Equilibrium Physical Generative Models

We show that the out-of-equilibrium driving protocol of score-based generative models (SGMs) can be learned via local learning rules. The gradient with respect to the parameters of the driving protocol is computed directly from force measurements or from observed system dynamics. As a demonstration, we implement an SGM in a network of driven, nonlinear, overdamped oscillators coupled to a thermal bath. We first apply it to the problem of sampling from a mixture of two Gaussians in 2D. Finally, we train an oscillator network on the MNIST dataset to generate images of handwritten digits 0 and 1.

cs.LG↗

arXiv2026

Acoustic Cloning

Cloning refers to producing identical copies of existing objects. Here, we experimentally show how to clone acoustic scattering objects. We acquire a digital twin and bring it back to life - a simple two-step process. First, we use broadband speakers to illuminate the scattering object within a closed receiver aperture. From these recorded reverberative data, we retrieve the object's scattering Green's functions using multidimensional deconvolution. In the second step, the acoustic scatterer is holographically reconstructed using the acquired scattering Green's functions. The hologram scatters any wavefield in real-time exactly like the original object would. Low-latency feedback reproduces all orders of interactions between the physical wavefield and the numerically defined hologram. This two-step process is demonstrated by cloning and modifying several rigid scatterers in a two-dimensional acoustic waveguide. Applications range from fully realistic digital scattering models to efficient metamaterial experimentation.

physics.app-ph↗

arXiv2026

Programming nanomechanical computation with light

Looking at physical systems as computers allows us to regard physical properties, such as thermal noise, symmetry or topology, as unconventional resources for computation. However, harnessing these resources requires programming computational functionality through strong, controllable nonlinearities in the system. Here, we show that cavity optomechanical interactions allow laser-controlled computation with nanomechanical degrees of freedom. We demonstrate a set of basic digital logic gates with level restoration and controlled mechanical couplings as essential ingredients for arbitrary computing networks. Owing to the strong optomechanical nonlinearity and precise readout, the system operates close to thermal amplitudes, in the regime where thermodynamic stochasticity governs its behavior. This opens a new path for the realization of physical computing with controlled nonlinear resonators.

physics.optics↗

arXiv2026

Multimodal oscillator networks learn to solve a classification problem

We numerically demonstrate a network of coupled oscillators that can learn to solve a classification task from a set of examples -- performing both training and inference through the nonlinear evolution of the system. We accomplish this by combining three key elements to achieve learning: A long-term memory that stores learned responses, analogous to the synapses in biological brains; a short-term memory that stores the neural activations, similar to the firing patterns of neurons; and an evolution law that updates the synapses in response to novel examples, inspired by synaptic plasticity. Achieving all three elements in wave-based information processors such as metamaterials is a significant challenge. Here, we solve it by leveraging the material multistability to implement long-term memory, and harnessing symmetries and thermal noise to realize the learning rule. Our analysis reveals that the learning mechanism, although inspired by synaptic plasticity, also shares parallelisms with bacterial evolution strategies, where mutation rates increase in the presence of noxious stimuli.

cond-mat.mes-hall↗

arXiv2025

Synthesis of mass-spring networks from high-level code descriptions

Structural nonlinearity can be harnessed to program complex functionalities in robotic devices. However, it remains a challenge to design nonlinear systems that will accomplish a specific, desired task. The responses that we typically describe as intelligent -- such a robot navigating a maze -- require a large number of degrees of freedom and cannot be captured by traditional optimization objective functions. In this work, we explore a code-based synthesis approach to design mass-spring systems with embodied intelligence. The approach starts from a source code, written in a \emph{mechanical description language}, that details the system boundary, sensor and actuator locations, and desired behavior. A synthesizer software then automatically generates a mass-spring network that performs the described function from the source code description. We exemplify this methodology by designing mass-spring systems realizing a maze-navigating robot and a programmable lock. Remarkably, mechanical description languages can be combined with large-language models, to translate a natural-language description of a task into a functional device.

cs.OH↗

arXiv2025

Embodying computation in nonlinear perturbative metamaterials

Designing metamaterials that carry out advanced computations poses a significant challenge. A powerful design strategy splits the problem into two steps: First, encoding the desired functionality in a discrete or tight-binding model, and second, identifying a metamaterial geometry that conforms to the model. Applying this approach to information-processing tasks requires accurately mapping nonlinearity -- an essential element for computation -- from discrete models to geometries. Here we formulate this mapping through a nonlinear coordinate transformation that accurately connects tight-binding degrees of freedom to metamaterial excitations in the nonlinear regime. This transformation allows us to design information-processing metamaterials across the broad range of computations that can be expressed as tight-binding models, a capability we showcase with three examples based on three different computing paradigms: a coherent Ising machine that approximates combinatorial optimization problems through energy minimization, a mechanical racetrack memory exemplifying in-memory computing, and a speech classification metamaterial based on analog neuromorphic computing.

cond-mat.mes-hall↗

arXiv2025

Racetrack computing with a topological boundary ratchet

Multistable order parameters provide a natural means of encoding non-volatile information in spatial domains, a concept that forms the foundation of magnetic memory devices. However, this stability inherently conflicts with the need to move information around the device for processing and readout. While in magnetic systems, domains can be transported using currents or external fields, mechanisms to robustly shuttle information-bearing domains across neutral systems are scarce. Here, we experimentally realize a topological boundary ratchet in an elastic metamaterial, where digital information is encoded in buckling domains and transported in a quantized manner via cyclic loading. The transport is topological in origin: neighboring domains act as different topological pumps for their Bogoliubov excitations, so their interface hosts topological boundary modes. Cyclic loading renders these modes unstable through inter-domain pressure, which in turn drives the motion of the domain wall. We demonstrate that the direction of information propagation can be controlled through adjustable mechanical constraints on the buckling beams, and numerically investigate buckling-based domain-wall logic circuits in an elastic metamaterial network. The underlying tight-binding structure with low-order nonlinearities makes this approach a general pathway toward racetrack memories in neutral systems.

cond-mat.mes-hall↗

arXiv2024

Adjoint computation of Berry phase gradients

Berry phases offer a geometric perspective on wave propagation and are key to designing materials with topological wave transport. However, controlling Berry phases is challenging due to their dependence on global integrals over the Brillouin zone, making differentiation difficult. We present an adjoint-based method for efficiently computing the gradient of the Berry phase with respect to system parameters, involving only one forward and one adjoint calculation. This approach enables the use of advanced optimization techniques, such as topology optimization, to design new materials with tailored topological wave properties.

cond-mat.mes-hall↗

arXiv2024

Reprogrammable, in-materia matrix-vector multiplication with floppy modes

Matrix-vector multiplications are a fundamental building block of artificial intelligence; this essential role has motivated their implementation in a variety of physical substrates, from memristor crossbar arrays to photonic integrated circuits. Yet their realization in soft-matter intelligent systems remains elusive. Here, we experimentally demonstrate a reprogrammable elastic metamaterial that computes matrix-vector multiplications using floppy modes -- deformations with near-zero stored elastic energy. Floppy modes allow us to program complex deformations without being hindered by the natural stiffness of the material; but their practical application is challenging, as their existence depends on global topological properties of the system. To overcome this challenge, we introduce a continuously parameterized unit cell design with well-defined compatibility characteristics. This unit cell is then combined to form arbitrary matrix-vector multiplications that can even be reprogrammed after fabrication. Our results demonstrate that floppy modes can act as key enablers for embodied intelligence, smart MEMS devices and in-sensor edge computing.

cond-mat.soft↗

arXiv2024

On acoustic space-time media that compute their own inverse

We derive time reflection and transmission coefficients for 1D acoustic waves encountering a time boundary at which the properties of the medium change instantaneously. The time reflection and transmission coefficients are shown to be identical to so-called reverse-space reflection and transmission coefficients which appear in the recursive computation of focusing wavefields used in seismology. We establish a bijectivity between the focusing wavefields and the wavefields produced by time scattering and show how this can be used to construct a space-time medium where the time scattering anticipates the space scattering and "computes" the exact inverse for the space scattering. The construction is shown to be independent of the boundary conditions chosen to compute the reflection and transmission coefficients. We demonstrate the construction with a simple numerical example of a single pulse encountering a series of time boundaries before reaching a spatial inhomogeneity. The time boundaries scatter the single pulse into a focusing wavefield that subsequently focuses through the spatial inhomogeneity. Under certain conditions, the transmitted wave has both the same wave shape and amplitude as the original pulse, yielding a transmission coefficient of unity. The reflection coefficient of the space-time medium is always non-zero however.

physics.geo-ph↗

arXiv2023

Personalized identification, prediction, and stimulation of neural oscillations via data-driven models of epileptic network dynamics

Neural oscillations are considered to be brain-specific signatures of information processing and communication in the brain. They also reflect pathological brain activity in neurological disorders, thus offering a basis for diagnoses and forecasting. Epilepsy is one of the most common neurological disorders, characterized by abnormal synchronization and desynchronization of the oscillations in the brain. About one third of epilepsy cases are pharmacoresistant, and as such emphasize the need for novel therapy approaches, where brain stimulation appears to be a promising therapeutic option. The development of brain stimulation paradigms, however, is often based on generalized assumptions about brain dynamics, although it is known that significant differences occur between patients and brain states. We developed a framework to extract individualized predictive models of epileptic network dynamics directly from EEG data. The models are based on the dominant coherent oscillations and their dynamical coupling, thus combining an established interpretation of dynamics through neural oscillations, with accurate patient-specific features. We show that it is possible to build a direct correspondence between the models of brain-network dynamics under periodic driving, and the mechanism of neural entrainment via periodic stimulation. When our framework is applied to EEG recordings of patients in status epilepticus (a brain state of perpetual seizure activity), it yields a model-driven predictive analysis of the therapeutic performance of periodic brain stimulation. This suggests that periodic brain stimulation can drive pathological states of epileptic network dynamics towards a healthy functional brain state.

q-bio.NC↗

arXiv2023

Binary classification of spoken words with passive phononic metamaterials

Mitigating the energy requirements of artificial intelligence requires novel physical substrates for computation. Phononic metamaterials have a vanishingly low power dissipation and hence are a prime candidate for green, always-on computers. However, their use in machine learning applications has not been explored due to the complexity of their design process: Current phononic metamaterials are restricted to simple geometries (e.g. periodic, tapered), and hence do not possess sufficient expressivity to encode machine learning tasks. We design and fabricate a non-periodic phononic metamaterial, directly from data samples, that can distinguish between pairs of spoken words in the presence of a simple readout nonlinearity; hence demonstrating that phononic metamaterials are a viable avenue towards zero-power smart devices.

eess.SP↗

arXiv2021

Broadband acoustic invisibility and illusions

Rendering objects invisible to impinging acoustic waves (cloaking) and creating acoustic illusions (holography) has been attempted using active and passive approaches. While passive methods are applicable only to narrow frequency bands, active approaches attempt to respond dynamically, interfering with broadband incident or scattered wavefields by emitting secondary waves. Without prior knowledge of the primary wavefield, the signals for the secondary sources need to be estimated and adapted in real-time. This has thus far impeded active cloaking and holography for broadband wavefields. We present experimental results of active acoustic cloaking and holography without prior knowledge of the wavefield so that objects remain invisible and illusions intact even for broadband moving sources. This opens novel research directions and facilitates practical applications including architectural acoustics, and stealth.

physics.app-ph↗

arXiv2020

Experimental characterization of fragile topology in an acoustic metamaterial

Symmetries crucially underlie the classification of topological phases of matter. Most materials, both natural as well as architectured, possess crystalline symmetries. Recent theoretical works unveiled that these crystalline symmetries can stabilize fragile Bloch bands that challenge our very notion of topology: while answering to the most basic definition of topology, one can trivialize these bands through the addition of trivial Bloch bands. Here, we fully characterize the symmetry properties of the response of an acoustic metamaterial to establish the fragile nature of the low-lying Bloch bands. Additionally, we present a spectral signature in the form of spectral flow under twisted boundary conditions.

cond-mat.mes-hall↗

arXiv2020

Observation of a phononic quadrupole topological insulator

The modern theory of charge polarization in solids is based on a generalization of Berry's phase. Its possible quantization lies at the heart of our understanding of all systems with topological band structures that were discovered over the last decades. While based on the concept of the "charge" polarization, the same theory can be used as an elegant tool to characterize the Bloch bands of neutral bosonic systems such as photonic or phononic crystals. Recently, the theory of this quantized polarization was extended from the dipole- to higher multipole-moments. In particular, a two-dimensional quantized quadrupole insulator is predicted to have gapped yet topological one-dimensional edge-modes, which in turn stabilize zero-dimensional in-gap corner states. However, such a state of matter has not been observed experimentally. Here, we provide the first measurements of a phononic quadrupole insulator. We experimentally characterize the bulk, edge, and corner physics of a mechanical metamaterial and find the predicted gapped edge and in-gap corner states. We further corroborate our findings by comparing the mechanical properties of a topologically non-trivial system to samples in other phases predicted by the quadrupole theory. From an application point of view, these topological corner states are an important stepping stone on the way to topologically protected wave-guides in higher dimensions and thereby open a new design path for metamaterials.

cond-mat.mtrl-sci↗

arXiv2020

Axial-field-induced chiral channels in an acoustic Weyl system

Condensed-matter and other engineered systems, such as cold atoms, photonic, or phononic metamaterials, have proven to be versatile platforms for the observation of low-energy counterparts of elementary particles from relativistic field theories. These include the celebrated Majorana modes, as well as Dirac and Weyl fermions. An intriguing feature of the Weyl equation is the chiral symmetry, where the two chiral sectors have an independent gauge freedom. While this freedom leads to a quantum anomaly, there is no corresponding axial background field coupling differently to opposite chiralities in quantum electrodynamics. Here, we provide the experimental characterization of the effect of such an axial field in an acoustic metamaterial. We implement the axial field through an inhomogeneous potential and observe the induced chiral Landau levels. From the metamaterials perspective these chiral channels open the possibility for the observation of non-local Weyl orbits and might enable unidirectional bulk transport in a time-reversal invariant system.

cond-mat.other↗

arXiv2020

Discovery of topological metamaterials by symmetry relaxation and smooth topological indicators

Robustness against small perturbations is a crucial feature of topological properties. This robustness is both a source of theoretical interest and a drive for technological applications, but presents a challenge when looking for new topological systems: Small perturbations cannot be used to identify the global direction of change in the topological indices. Here, we overcome this limitation by breaking the symmetries protecting the topology. The introduction of symmetry-breaking terms causes the topological indices to become non-quantized variables, which are amenable to efficient design algorithms based on gradient methods. We demonstrate this capability by designing discrete and continuous phononic systems realizing conventional and higher-order topological insulators.

cond-mat.mes-hall↗

arXiv2019

Turing complete mechanical processor via automated nonlinear system design

Nanomechanical computers promise a greatly improved energetic efficiency compared to their electrical counterparts. However, progress towards this goal is hindered by a lack of modular components, such as logic gates or transistors, and systematic design strategies. This article describes a universal logic gate implemented as a nonlinear mass-spring-damper model, followed by an automated method to translate computations, expressed as source code of arbitrary complexity, into combinations of this basic building block. The proposed approach is validated numerically in two steps: First, a set of discrete models are generated from code. The models implement computations with increasing complexity, starting by a simple adder and ending in a 8-bit Turing complete mechanical processor. Then, the models are forward integrated to demonstrate their computing performance. The processor is validated by executing the Erathostenes' sieve algorithm to mechanically compute prime numbers.

cs.ET↗