E.E. Umukoro, M. Mishra, P. Flanigan, B. Vlahovic
North Carolina Central University,
United States
Keywords: quantum entanglement, quantum photonics, silicon photonics, photonics
Summary:
The generation of entangled triphotons via spontaneous six-wave mixing (SSWM) has been demonstrated experimentally, but improving the generation rate would be greatly advantageous to pushing this technology forward and advancing the fields of quantum communication and quantum computing. By coupling the generative medium (Rb-87 vapor) to a silicon nitride microring resonator (MRR), it will be possible to generate triphotons at a higher rate compared to the current method of direct laser illumination. The goal then becomes a matter of designing the controllable properties of the microring resonator (ring radius, width, etc.) to maximize coupling efficiency at the three wavelengths where the relevant electronic transitions in the Rb atom are known to occur. This project uses electromagnetic simulations techniques (such as FDTD) and Machine Learning-based optimization to identify the ideal configuration and to predict the microring’s optical behavior when the simplified analytical equations fail. The systems being simulated are quite large in terms of required memory. This is unavoidable given the operating wavelength, the size of the ring, and the resolution (mesh fineness) required to obtain accurate results. Using high-performance computing becomes necessary, but even then, this project exceeds the limit of what can be done on a shared academic cluster, because not only are the systems large, but the multi-variable parameter optimization space is as well. To deal with this, Machine Learning algorithms (such as random forest) are being employed to predict how the high-memory cases will behave without having to actually run them. This requires identifying the underlying trendlines based on physical intuition and some simple analytical formulae. The location of the spectral maxima is particularly important here, since it is predetermined by the choice of generative medium (which shows where the peaks need to occur here). One challenge of this method is that three spectral maxima must be induced simultaneously to produce entangled triphotons. The goal is to obtain Q factors of at least 10^2 million for these resonances, which has been demonstrated experimentally for a ring multiple orders of magnitude greater than the operating wavelength. Certain properties such as the free spectral range show excellent agreement with the predictions based on analytical equations. However, the precise location of the peaks show disagreement between the simulated results and the predicted values in some cases. This is a result of the simplifying assumptions that go into calculations of the optical mode’s propagation constant (beta). Thus, the Machine Learning analysis has focused on not only the resultant spectral maximum, but also the nature of the eigenmodes in a realistic curved geometry. It has been found that carefully identifying the connections between data points along the independent variable axis will result in more accurate predictions when going into the high-radius regime, but how this will scale to very high radii remains to be seen.