Broadband giant nonlinear response using electrically tunable polaritonic metasurfaces.

Intersubband transitions in semiconductor heterostructures offer a way to achieve large and designable nonlinearities with dynamic modulation of intersubband energies through the Stark effect. One promising approach for incorporating these nonlinearities into free space optics is a nonlinear polaritonic metasurface, which derives resonant coupling between intersubband nonlinearities and optical modes in nanocavities. Recent work has shown efficient frequency mixing at low pumping intensities, with the ability to electrically tune the phase, amplitude, and spectral peak of it.

Electrically tunable third-harmonic generation using intersubband polaritonic metasurfaces.

Nonlinear intersubband polaritonic metasurfaces, which integrate giant nonlinear responses derived from intersubband transitions of multiple quantum wells (MQWs) with plasmonic nanoresonators, not only facilitate efficient frequency conversion at pump intensities on the order of few tens of kW cm but also enable electrical modulation of nonlinear responses at the individual meta-atom level and dynamic beam manipulation. The electrical modulation characteristics of the magnitude and phase of the nonlinear optical response are realized through Stark tuning of the resonant intersubband nonlinearity. In this study, we report, for the first time, experimental implementations of electrical modulation characteristics of mid-infrared third-harmonic generation (THG) using an intersubband polaritonic metasurface based on MQW with electrically tunable third-order nonlinear response.

Mechanism for feature learning in neural networks and backpropagation-free machine learning models.

Understanding how neural networks learn features, or relevant patterns in data, for prediction is necessary for their reliable use in technological and scientific applications. In this work, we presented a unifying mathematical mechanism, known as average gradient outer product (AGOP), that characterized feature learning in neural networks. We provided empirical evidence that AGOP captured features learned by various neural network architectures, including transformer-based language models, convolutional networks, multilayer perceptrons, and recurrent neural networks.

Quartz-Enhanced Photoacoustic Sensor Based on a Multi-Laser Source for In-Sequence Detection of NO, SO, and NH.

In this work, we report on the implementation of a multi-quantum cascade laser (QCL) module as an innovative light source for quartz-enhanced photoacoustic spectroscopy (QEPAS) sensing. The source is composed of three different QCLs coupled with a dichroitic beam combiner module that provides an overlapping collimated beam output for all three QCLs. The 3λ-QCL QEPAS sensor was tested for detection of NO, SO, and NH in sequence in a laboratory environment.

Efficient Second-Harmonic Generation from Dielectric Inter-subband Polaritonic Metasurfaces Coupled to Lattice Resonance.

Nonlinear optical metasurfaces offer a possibility to perform frequency mixing without the phase-matching constraints of bulk nonlinear crystals and with control of the local nonlinear response at a sub-wavelength scale. Nonlinear inter-subband polaritonic metasurfaces created by combining the semiconductor heterostructures with quantum-engineered inter-subband nonlinear response and electromagnetically engineered metal-clad nanoresonators offer by far the largest second-order nonlinear response of all condensed matter systems reported to date. However, the nonlinear optical response of these metasurfaces is limited by optical intensity saturation in the nanoresonator hot spots that prevented the achievement of power conversion efficiencies over 0.

Electrical Phase Modulation Based on Mid-Infrared Intersubband Polaritonic Metasurfaces.

Electrically reconfigurable metasurfaces that overcome the static limitations in controlling the fundamental properties of scattered light are opening new avenues for functional flat optics. This work proposes and experimentally demonstrates electrically phase-tunable mid-infrared metasurfaces based on the polaritonic coupling of Stark-tunable intersubband transitions in semiconductor heterostructures and electromagnetic modes in plasmonic nanoresonators. In the applied voltage range of -3 to +3 V, the local phase tuning of the light reflects from the metasurface, which enables the electrical control of the polarization state and wavefront of the reflected wave.

Wide and deep neural networks achieve consistency for classification.

While neural networks are used for classification tasks across domains, a long-standing open problem in machine learning is determining whether neural networks trained using standard procedures are consistent for classification, i.e., whether such models minimize the probability of misclassification for arbitrary data distributions.

Control of second-harmonic generation in all-dielectric intersubband metasurfaces by controlling the polarity of χ.

All-dielectric metasurfaces have recently led to a paradigm shift in nonlinear optics as they allow for circumventing the phase matching constraints of bulk crystals and offer high nonlinear conversion efficiencies when normalized by the light-matter interaction volume. Unlike bulk crystals, in all-dielectric metasurfaces nonlinear conversion efficiencies primarily rely on the material nonlinearity, field enhancements, and the modal overlaps, therefore most efforts to date have only focused on utilizing these degrees of freedom. In this work, we demonstrate that for second-harmonic generation in all-dielectric metasurfaces, an additional degree of freedom is the control of the polarity of the nonlinear susceptibility.

Terahertz difference-frequency-generation quantum cascade lasers on silicon with wire grid current injectors.

We propose the concept and experimentally verify the operation of terahertz quantum cascade laser sources based on intra-cavity Cherenkov difference-frequency generation on a silicon substrate with the current injection layer configured as a metal wire grid. Such a current injector configuration enables high transmission of TM-polarized terahertz radiation into the silicon substrate while simultaneously providing a low-resistivity metal contact for current injection.

Simple, fast, and flexible framework for matrix completion with infinite width neural networks.

Matrix completion problems arise in many applications including recommendation systems, computer vision, and genomics. Increasingly larger neural networks have been successful in many of these applications but at considerable computational costs. Remarkably, taking the width of a neural network to infinity allows for improved computational performance.

An All-Dielectric Polaritonic Metasurface with a Giant Nonlinear Optical Response.

Enhancing the efficiency of second-harmonic generation using all-dielectric metasurfaces to date has mostly focused on electromagnetic engineering of optical modes in the meta-atom. Further advances in nonlinear conversion efficiencies can be gained by engineering the material nonlinearities at the nanoscale, however this cannot be achieved using conventional materials. Semiconductor heterostructures that support resonant nonlinearities using quantum engineered intersubband transitions can provide this new degree of freedom.

Overcoming Intensity Saturation in Nonlinear Multiple-Quantum-Well Metasurfaces for High-Efficiency Frequency Upconversion.

Engineered intersubband transitions in semiconductor heterostructures featuring multiple quantum wells (MQWs) are shown to support record-high second-order nonlinear susceptibilities. By integrating these materials in metasurfaces with tailored optical resonances, it is possible to further enhance photonic interactions, yielding giant nonlinear responses in ultrathin devices. These metasurfaces form a promising platform for efficient nonlinear processes, including frequency upconversion of low-intensity thermal infrared radiation and harmonic generation, free of phase-matching constraints intrinsic to bulk nonlinear crystals.

Strong Coupling in All-Dielectric Intersubband Polaritonic Metasurfaces.

Mie-resonant dielectric metasurfaces are excellent candidates for both fundamental studies related to light-matter interactions and for numerous applications ranging from holography to sensing to nonlinear optics. To date, however, most applications using Mie metasurfaces utilize only weak light-matter interaction. Here, we go beyond the weak coupling regime and demonstrate for the first time strong polaritonic coupling between Mie photonic modes and intersubband (ISB) transitions in semiconductor heterostructures.

Giant Nonlinear Circular Dichroism from Intersubband Polaritonic Metasurfaces.

Nonlinear metasurfaces are advancing into a new paradigm of "flat nonlinear optics" owing to the ability to engineer local nonlinear responses in subwavelength-thin films. Recently, attempts have been made to expand the design space of nonlinear metasurfaces through nonlinear chiral responses. However, the development of metasurfaces that display both giant nonlinear circular dichroism and significantly large nonlinear optical response is still an unresolved challenge.

Overparameterized neural networks implement associative memory.

Identifying computational mechanisms for memorization and retrieval of data is a long-standing problem at the intersection of machine learning and neuroscience. Our main finding is that standard overparameterized deep neural networks trained using standard optimization methods implement such a mechanism for real-valued data. We provide empirical evidence that 1) overparameterized autoencoders store training samples as attractors and thus iterating the learned map leads to sample recovery, and that 2) the same mechanism allows for encoding sequences of examples and serves as an even more efficient mechanism for memory than autoencoding.

Reconciling modern machine-learning practice and the classical bias-variance trade-off.

Breakthroughs in machine learning are rapidly changing science and society, yet our fundamental understanding of this technology has lagged far behind. Indeed, one of the central tenets of the field, the bias-variance trade-off, appears to be at odds with the observed behavior of methods used in modern machine-learning practice. The bias-variance trade-off implies that a model should balance underfitting and overfitting: Rich enough to express underlying structure in data and simple enough to avoid fitting spurious patterns.

Back to the Future: Radial Basis Function Network Revisited.

Radial Basis Function (RBF) networks are a classical family of algorithms for supervised learning. The most popular approach for training RBF networks has relied on kernel methods using regularization based on a norm in a Reproducing Kernel Hilbert Space (RKHS), which is a principled and empirically successful framework. In this paper we aim to revisit some of the older approaches to training the RBF networks from a more modern perspective.

Quantum Confinement in Oxide Heterostructures: Room-Temperature Intersubband Absorption in SrTiO/LaAlO Multiple Quantum Wells.

The Si-compatibility of perovskite heterostructures offers the intriguing possibility of producing oxide-based quantum well (QW) optoelectronic devices for use in Si photonics. While the SrTiO/LaAlO (STO/LAO) system has been studied extensively in the hopes of using the interfacial two-dimensional electron gas in Si-integrated electronics, the potential to exploit its giant 2.4 eV conduction band offset in oxide-based QW optoelectronic devices has so far been largely ignored.

Spectral purity and tunability of terahertz quantum cascade laser sources based on intracavity difference-frequency generation.

Terahertz sources based on intracavity difference-frequency generation in mid-infrared quantum cascade lasers (THz DFG-QCLs) have recently emerged as the first monolithic electrically pumped semiconductor sources capable of operating at room temperature across the 1- to 6-THz range. Despite tremendous progress in power output, which now exceeds 1 mW in pulsed and 10 μW in continuous-wave regimes at room temperature, knowledge of the major figure of merits of these devices for high-precision spectroscopy, such as spectral purity and absolute frequency tunability, is still lacking. By exploiting a metrological grade system comprising a terahertz frequency comb synthesizer, we measure, for the first time, the free-running emission linewidth (LW), the tuning characteristics, and the absolute center frequency of individual emission lines of these sources with an uncertainty of 4 × 10.

Electrical tuning of the polarization state of light using graphene-integrated anisotropic metasurfaces.

Plasmonic metasurfaces have been employed for moulding the flow of transmitted and reflected light, thereby enabling numerous applications that benefit from their ultra-thin sub-wavelength format. Their appeal is further enhanced by the incorporation of active electro-optic elements, paving the way for dynamic control of light's properties. In this paper, we realize a dynamic polarization state generator using a graphene-integrated anisotropic metasurface (GIAM) that converts the linear polarization of the incident light into an elliptical one.

Thermopile detector of light ellipticity.

Polarimetric imaging is widely used in applications from material analysis to biomedical diagnostics, vision and astronomy. The degree of circular polarization, or light ellipticity, is associated with the S Stokes parameter which is defined as the difference in the intensities of the left- and right-circularly polarized components of light. Traditional way of determining this parameter relies on using several external optical elements, such as polarizers and wave plates, along with conventional photodetectors, and performing at least two measurements to distinguish left- and right-circularly polarized light components.

Experimental Demonstration of Phase Modulation and Motion Sensing Using Graphene-Integrated Metasurfaces.

Strong interaction of graphene with light accounts for one of its most remarkable properties: the ability to absorb 2.3% of the incident light's energy within a single atomic layer. Free carrier injection via field-effect gating can dramatically vary the optical properties of graphene, thereby enabling fast graphene-based modulators of the light intensity.