Search for: All records

Recent advances in ferroic materials have identified topological defects as promising candidates for enabling additional functionalities in future electronic systems. The generation of stable and customizable polar topologies is needed to achieve multistates that enable beyond-binary device architectures. In this study, we show how to autonomously pattern on-demand highly tunable striped closure domains in pristine rhombohedral-phase BiFeO3 thin films through precise scanning of a biased atomic force microscopy tip along carefully designed paths. By employing this strategy, we generate and manipulate closed-loop structures with high spatial resolution in an automated manner, allowing the creation of highly tunable and intricate topological domain structures that exhibit distinct polarization configurations without the need for electrode deposition or complex heterostructure growth. As a proof-of-concept for ferroelectric beyond-binary memory devices, we use such topological domains as multistates, engineering an alphabet and automating the symbolic writing/reading process using autonomous microscopy. The resulting information density is compared with that of current commercially available memory devices, demonstrating the potential of ferroelectric topological domains for multistate information storage applications. 

An increasingly popular machine learning paradigm is to pretrain a neural network (NN) on many tasks offline, then adapt it to downstream tasks, often by re-training only the last linear layer of the network. This approach yields strong downstream performance in a variety of contexts, demonstrating that multitask pretraining leads to effective feature learning. Although several recent theoretical studies have shown that shallow NNs learn meaningful features when either (i) they are trained on a single task or (ii) they are linear, very little is known about the closer-to-practice case of nonlinear NNs trained on multiple tasks. In this work, we present the first results proving that feature learning occurs during training with a nonlinear model on multiple tasks. Our key insight is that multi-task pretraining induces a pseudo-contrastive loss that favors representations that align points that typically have the same label across tasks. Using this observation, we show that when the tasks are binary classification tasks with labels depending on the projection of the data onto an r-dimensional subspace within the d k r-dimensional input space, a simple gradient-based multitask learning algorithm on a two-layer ReLU NN recovers this projection, allowing for generalization to downstream tasks with sample and neuron complexity independent of d. In contrast, we show that with high probability over the draw of a single task, training on this single task cannot guarantee to learn all r ground-truth features. 

Hierarchical assemblies of ferroelectric nanodomains, so-called super-domains, can exhibit exotic morphologies that lead to distinct behaviours. Controlling these super-domains reliably is critical for realizing states with desired functional properties. Here we reveal the super-switching mechanism by using a biased atomic force microscopy tip, that is, the switching of the in-plane super-domains, of a model ferroelectric Pb0.6Sr0.4TiO3. We demonstrate that the writing process is dominated by a super-domain nucleation and stabilization process. A complex scanning-probe trajectory enables on-demand formation of intricate centre-divergent, centre-convergent and flux-closure polar structures. Correlative piezoresponse force microscopy and optical spectroscopy confirm the topological nature and tunability of the emergent structures. The precise and versatile nanolithography in a ferroic material and the stability of the generated structures, also validated by phase-field modelling, suggests potential for reliable multi-state nanodevice architectures and, thereby, an alternative route for the creation of tunable topological structures for applications in neuromorphic circuits. 

In undoped lead zirconate titanate films of 1–2  μm thick, domain walls move in clusters with a correlation length of approximately 0.5–2  μm. Band excitation piezoresponse force microscopy mapping of the piezoelectric nonlinearity revealed that niobium (Nb) doping increases the average concentration or mobility of domain walls without changing the cluster area of correlated domain wall motion. In contrast, manganese (Mn) doping reduces the contribution of mobile domain walls to the dielectric and piezoelectric responses without changing the cluster area for correlated motion. In both Nb and Mn doped films, the cluster area increases and the cluster density drops as the film thickness increases from 250 to 1250 nm. This is evident in spatial maps generated from the analysis of irreversible to reversible ratios of the Rayleigh coefficients. 

Deep neural networks have shown the ability to extract universal feature representations from data such as images and text that have been useful for a variety of learning tasks. However, the fruits of representation learning have yet to be fully-realized in federated settings. Although data in federated settings is often non-i.i.d. across clients, the success of centralized deep learning suggests that data often shares a global feature representation, while the statistical heterogeneity across clients or tasks is concentrated in the labels. Based on this intuition, we propose a novel federated learning framework and algorithm for learning a shared data representation across clients and unique local heads for each client. Our algorithm harnesses the distributed computational power across clients to perform many local-updates with respect to the low-dimensional local parameters for every update of the representation. We prove that this method obtains linear convergence to the ground-truth representation with near-optimal sample complexity in a linear setting, demonstrating that it can efficiently reduce the problem dimension for each client. Further, we provide extensive experimental results demonstrating the improvement of our method over alternative personalized federated learning approaches in heterogeneous settings.