AMO Parallel Talks

Deep learning has ubiquitous applications in many fields of science including optics, biomedical imaging, computer vision, drug discovery, fundamental scientific research, and much more. Despite its widespread success, deep learning depends on high quality and abundant data, a requirement which is rarely met in ultrafast optical systems. Moreover, deep learning suffers from overfitting to noise which is only amplified when data is sparse. In most ultrafast systems, measurements can only be conducted through integrating over long periods of time compared to an optical cycle or pulse duration. For these reasons, optical systems are often sparsely sampled and noisy. In this regime, Physics Informed Neural Networks (PINNs) have emerged as a promising platform for intelligently interpreting noisy and low volume data sets by incorporating governing physics into the learning process. In this work, we introduce a Physics-Informed Frequency Resolved Optical Gating (Pi-FROG) method that allows us to discover the hidden propagation and parameters of an ultrafast optical pulse by measuring the system at only the input and output. Pi-FROG utilizes the Runge Kutta method to connect the input and output through discrete steps of the hidden propagation. By linking the input and output Pi-FROG is able to unambiguously retrieve the magnitude and sign of the optical phase. The inclusion of the underlying physics allows Pi-FROG to perform similarly to conventional retrieval algorithms but at much lower SNR’s. We experimentally demonstrate Pi-FROG by discovering the nonlinear coefficient of single-mode optical fiber by measuring the spectrogram at only the input and output. With Pi-FROG, we enable intelligent discovery of pulse propagation physics in a low SNR and data-sparse regime. Our work has the potential to vastly improve the sensitivity of conventional FROG systems, real-time monitoring of optical systems at sparsely sampled locations, and paves the way towards automated physics discovery in ultrafast optics.

Fiber optics have revolutionized precision frequency and time transfer between locations for applications such as communication, ranging, and timing. Such methods are redefining these application capabilities previously limited by microwave electronics. However, signals transmitted through optical fibers are subject to the noise environment in which the fiber is laid. This is particularly problematic for precision optical frequency transfer over long distances and for precision timing across large baselines. A method for cancellation of phase noise must be implemented to establish a stable, low-phase-noise link between locations. This becomes more difficult over larger separation distances. 

Mechanical frequency sensors are widely used in various applications. By coupling an external effect onto a mechanical resonator, changes in that effect produce shifts in the resonator’s frequency, transducing otherwise difficult-to-measure signal into simple motion measurements. One advantage of this platform is that its signal-to-noise ratio (SNR) can be boosted by actuating the resonator's motion to higher amplitudes. However, this approach is limited by the nonlinearity of the mechanics: when the resonator is driven beyond the harmonic oscillator regime, amplitude-to-frequency noise conversion occurs due to the Duffing nonlinearity. This excess noise offsets the SNR enhancement. In this work, we demonstrate an experimentally straightforward method that mitigates this noise conversion on a tensioned thin-film resonator driven into nonlinear regimes. By characterizing the resonator’s Duffing characteristics and performing concurrent amplitude measurements during signal readout, we surpass the sensitivity limits set by the mechanical nonlinearity. We show that operation beyond the nonlinear regime is advantageous, in contrast to long-standing perspectives in the field.

By exciting alkali atoms to a high principle quantum number we can create Rydberg atoms that are very susceptible to incident electric fields.  By using this concept and applying another probe laser beam, we can create a so called dark state in the atom and read out an electromagnetically Induced Transparency signal, which can then be used to measure incident electric field.  This technology has been used to accurately measure electric fields to well defined quantum constants, to build an atom based receiver that works in a broad bandwidth, and to directly sense and image electric fields.  

Squeezed light, a nonclassical state characterized by reduced quantum fluctuations in one field quadrature, is a key resource for improving precision measurements and quantum metrology. In the past decades, the most well-known application of squeezed light is gravitational wave detection, realized by the Laser Interferometer Gravitational- Wave Observatory (LIGO). There, squeezing has pushed the strain precision of a 4-km-long interferometer to an astounding $10^{-23}$ level. An interesting question researchers are pursuing is whether squeezing and related precision can be realized in cm-size photonic chips. Recently, the Kerr nonlinearity on integrated photonics chips have opened a new window for squeezed light generation in a nanophotonic platform. In this work, we seek to extend these advances to single-mode squeezed state (SMS) generation in Silicon Nitride (SiN) Kerr microresonators with a coherent-laser-based dual-pump SMS scheme. This is the first step towards a scalable approach to on-chip squeezed-light sources for a new class of integrated photonics sensors.

Many materials exhibit heterogeneous electronic and vibrational dynamics on nanometer spatial and picosecond temporal scales. To gain a better understanding of these elementary processes and how they relate to observed macroscopic optical properties of interest, a technique with sufficient temporal, spatial, and spectral resolutions is required. Additionally, many transient effects in materials are long lived, and thus require low-repetition rate lasers (~ 1 MHz) for photo-excitation and probing, giving rise to difficulties in signal isolation. Previously established methods, such as ultrafast transmission electron microscopy (TEM), photoemission electron microscopy (PEEM), and X-ray microscopy are not yet capable of the required excited state contrast. This talk will discuss an experimental implementation of heterodyne pump-probe infrared scattering-scanning near-field optical microscopy (HPP IR s-SNOM), which provides simultaneous spatial, temporal, and spectral resolutions with high sensitivity, to resolve phenomena on nanometer length and picosecond time scales. Finally, there will be a brief discussion of an application of HPP IR s-SNOM applied to a perovskite film to better understand how structural and compositional characteristics control the emerging photovoltaic performance.

The pathways and origins of many magnetic phenomena are not well understood on their fundamental timescales. Understanding these ultrafast spin dynamics is vital for the development of faster and more efficient technologies such as spintronic devices. Using extreme ultraviolet (EUV) light in combination with the transverse magneto optical effect (TMOKE), direct element specific measurements of ultrafast spin dynamics can be obtained for magnetic thin films and multilayer samples.