Demonstrating Programmable, On-Chip Nonlinear Photonics Transcending the One-Device-One-Function Paradigm

By Ryotatsu Yanagimoto, Ph.D.
Research Scientist, NTT Research Physics & Informatics (PHI) Lab
NTT Postdoctoral Fellow, Cornell University

A new paper from NTT Research, Cornell University and Stanford University published in Nature, “Programmable On-Chip Nonlinear Photonics,” details the successful fabrication of a programmable nonlinear waveguide, demonstrating the capability to implement and switch among various nonlinear-optical functions on a single device.

The developed technology, for the first time, marks a departure of nonlinear optics from the conventional one-device-one-function paradigm, expanding the applications to situations where fast device reconfigurability and high yields are essential.

Research Background

Nonlinear optics (NLO) harness nonlinear interactions among lightwaves, enabling functions that are inaccessible in purely linear optics. Since its birth in 1961, NLO has become a backbone of photonics technologies, where applications include light sources, ultrafast optics, frequency metrology and ranging, to name a few. Furthermore, NLO enables generation and manipulation of quantum states of light. As such, NLO has played an essential role in the foundation of quantum mechanics and is expected to be a central workhorse for future quantum photonic technologies (e.g., quantum computers and networks).

Importantly, such complicated functions of NLO are not present in raw materials in their naturally available forms. Instead, these functions are designed and “sculpted” to the structure of nonlinear-photonic devices through nanofabrication processes. For instance, one of the most powerful means to engineer a nonlinear-optical function is via tailoring quasi-phase matching (QPM) grating, which is formed by periodic spatial modulation of nonlinearity. A QPM grating is conventionally realized by inverting domain walls of ferroelectric materials using nano-fabricated electrodes or by epitaxially growing materials on orientation-patterned substrate.

The requirement for such “sculpting” steps means the function of a device is fixed during the fabrication and cannot be changed or added afterwards, leading to the present paradigm that one device can fulfill only one function. This also means the device functions are highly sensitive to fabrication imperfections and environmental drifts, reducing the yield. Overall, the one-device-one-function paradigm has limited the applicability of nonlinear optics to situations where such inflexibility and low-yield are tolerable.

Operating Principles of the Device

Figure 1(a) shows how a programmable nonlinear waveguide operates. Researchers utilized a planar optical waveguide whose core was made of silicon nitride to which pump light was coupled. Researchers then projected a structured light as programming illumination, which induced optical nonlinearity inside the core with the same spatial pattern as the programming illumination. This allowed the researchers to set and update an arbitrary distribution of nonlinearity, with which researchers could engineer nonlinear-optical functions via a mechanism known as quasi-phase matching (QPM).

Figure 1(b) shows the structure of a programmable nonlinear waveguide in detail. On top of the silicon nitride waveguide was a layer of photoconductive material and a transparent electrode. A strong bias voltage was applied through the entire stack of the films via the electrode and the substrate, which was made of conductive silicon. When programming illumination shone on the photoconductive layer, it turned the material conductive only where the illumination was strong. This lets the bias electric field penetrate through the photoconductive layer and impinge on the silicon nitride core. Then, this bias electric field broke the inversion symmetry of the core material and induced nonlinearity. Consequently, researchers realized a system where the programming illumination directly maps to the two-dimensional distribution of nonlinearity.

Experimental Results

For each “image” projected on the waveguide, one function is realized. In this paper, researchers demonstrated a wide range of functions, showing the device could tailor the dynamics of nonlinear optics in spectral, spatial, and spatio-spectral domains, all on a single chip of programmable nonlinear waveguide. A few of the demonstrations are highlighted below.

Figure 2(a) illustrates the demonstration of arbitrary spectral shaping of second-harmonic generation (SHG), a process that doubles the frequency of the pump light. Researchers pumped a programmable nonlinear waveguide with a broadband pulsed laser, measured the output SHG spectrum and fed back the results to the programming illumination to tailor the spectrum to a desired target form.

Such a real-time optimization based on real-time feedback was made possible by the programmability of the device. As shown in Figure 2(b), this optimization found highly nontrivial programming illumination patterns that generate desired target output spectra.

Furthermore, as shown in Figure 3(a), researchers can dynamically change the output spectrum by playing back the optimized programming illumination in real time as a “video”. With this, researchers portraited “NTT” and “CORNELL” patterns in the time trace of the SHG spectrum as shown in Figure 3(b).

The engineering that a programmable nonlinear waveguide enables goes beyond spectral domain. By leveraging the full two-dimensional programmability of the device, researchers demonstrated simultaneous control of spatial and spectral features of NLO. In the data shown in Figure 4, the SHG output took a highly complicated shape, exhibiting different numbers of localized spatial peaks depending on the wavelength.

In other words, the device can programmably “route” different colors of light to designated locations.

Application to Nanophotonics

The technique to implement programmable nonlinearity is not limited to a planar waveguide geometry.

In fact, one could “augment” programmable nonlinearity to various existing nanophotonic structure. As a proof of this concept, researchers fabricated a programmable nonlinear channel waveguide, where the core was a one-dimensional ridge of silicon nitride as shown in Figure 5. This enhanced the transverse confinement of light, and research demonstrated 40x enhancement of conversion efficiency compared to the case of the planar waveguide geometry.

Future Outlook

It is expected that the developed technology in this research could unlock various potential applications that have been inaccessible with conventional nonlinear optics that do not support programmability. Specifically, it is found that programmable nonlinear waveguide could find particularly compelling applications in the following four areas, even assuming the core performances demonstrated in this work: (i) on-chip arbitrary pulse shapers, (ii) reconfigurable quantum frequency converters, (iii) widely wavelength-tunable integrated light sources, and (iv) quantum light sources with programmable entanglement structure.

Explorations of materials with larger electric-field induced nonlinearity would further expand the scope of applications.