Our work in free-space optical communication focuses on wavefront correction using techniques such as 1) electronic wavefront correction via coherent detection and digital signal processing and 2) few-mode optical pre-amplification.

Our lab was the first to demonstrate long-distance transmission of using few-mode fibers, which enables space-division multiplexing to increase capacity beyond the single-mode fiber capacity limit. Our lab developed the theory of the few-mode EDFA and experimentally demonstrated the first amplified mode-division multiplexed transmission in few-mode fibers in collaboration with NEC and Corning. We introduced the concept of supermode fibers, which is also referred to as strongly-coupled multi-core fibers. We developed techniques for frequency-domain equalization for electronic mode demultiplexing.

Coherent optical communication experienced a remarkable revival in recent years. Our contributions include the first introduction of the MIMO concept to coherent optical communication for polarization demultiplexing, use of infinite-impulse response filters for digital dispersion compensation, orthogonal wavelength-division multiplexing and electronic wavefront correction for free-space coherent optical communications. Our lab's most noteworthy contribution is fiber nonlinearity compensation using digital backward propagation. With digital backward propagation, all deterministic linear and nonlinear impairments can be compensated, pushing the envelope for fundamental capacity at high transmission powers. A large number of recording-setting transmission experiments employ digital backward propagation.

Our group takes advantage of nonlinear dynamics in multi-section semiconductor lasers for application in analog fiber-optic links and subcarrier-multiplexed (SCM) optical networks. Particular areas that we have made significant contributions include optical generation of microwave/millimeter-wave (MMW) signals, high-gain and low noise figure analog fiber-optic links using few-mode fibers, and optical phased-array beamforming for massive MIMO.

To realize all-optical retiming, reshaping and reamplification (3R regeneration), clock recovery at the line rate is a necessary function that must be realized in the optical domain. Our approach to all-optical clock recovery also exploits nonlinear dynamics in multi-section semiconductor lasers, namely, coherent injection locking of self-pulsation two-section gain-coupled DFB lasers. Our group demonstrated record speed in all-optical clock recovery at 180 Gb/s, which still remain as the highest speed in all-optical clock recovery at the line rate.

In the area of optical regeneration, our group has made fundamental contributions especially for regeneration of phase-modulated signals. Prior to our pioneering work on phase regeneration, all-optical regeneration was limited to regeneration of intensity levels. It was not clear whether signals encoded in the optical phase could be regenerated from a fundamental perspective. Using the phase-sensitive amplification processes, we proposed and demonstrated regeneration of the two phase levels of a DPSK signal.

The laser is one of the greatest inventions in history. Because of its ubiquitous applications and profound societal impact, the concept of the laser has been extended to other physical domains including phonon lasers and atom lasers. Quite often, a laser in one physical domain is pumped by energy in another. However, all lasers demonstrated so far have only lased in one physical domain. We have experimentally demonstrated simultaneous photon and phonon lasing in a two-mode silica fiber ring cavity via forward intermodal SBS mediated by long-lived flexural acoustic waves. This two-domain laser may find potential applications in optical/acoustic tweezers, optomechanical sensing, microwave generation, and quantum information processing. Furthermore, we believe that this demonstration will usher in other multi-domain lasers and related applications.

The TDPM algorithm, inspired by its spatial-domain counterpart, the wavefront-matching algorithm, reaches high fidelity with fewer modulation stages than simulated annealing. For a one-to-one mode conversion (HG_0 to HG_5), five modulation steps achieve >99% overlap and four steps reach ~98% at 40 GHz modulation bandwidth constraint, while the simulated annealing baseline requires more modulation stages for a comparable fidelity under the same conditions. Furthermore, at low bandwidth (5 GHz), the TDPM algorithm still converges with a certain number of steps, whereas the simulated annealing baseline shows notable fidelity degradation. These results indicate that TDPM algorithm reduces required cascade depth for temporal mode conversion/(de)multiplexing and thus associated loss and calibration limits.