Research Track Leader: Dr. Frank Brückerhoff-Plückelmann.

Different physical computing platforms offer distinct strengths that can be combined within a single architecture. We therefore combine photonic processors with complementary analog computing platforms, assigning different parts of a computation to the hardware that performs them most efficiently. Fast and parallel photonic processing can, for example, be coupled with electronic in-memory or nonlinear computing. Our goal is to develop tightly integrated hybrid architectures that exploit the strengths of different physical platforms to achieve advantages at the level of the complete computing system.

Research Track Leaders:

Dr. Frank Brückerhoff-Plückelmann and Dr. Rongyang Xu.

We develop integrated photonic processors that bring optical computing from individual devices towards complete, programmable systems. This requires scalable photonic circuits as well as high-speed interfaces, light generation and detection, electronic control and software that allow them to operate together. We investigate their performance on applications ranging from deep neural network inference to solving partial differential equations, with the aim of translating the speed and parallelism of photonics into practical system-level computational advantages.

Research Track Leader: Dr. Shabnam Taheriniya.

We investigate how the physical properties of materials, photonic devices and light itself can be used directly for computation. This includes materials that can store information, devices that mimic the nonlinear response of neurons, and tailored light fields whose fluctuations enable probabilistic processing. By exploiting these physical behaviours, we develop new building blocks for neuromorphic hardware that can perform computations efficiently without reproducing every operation in conventional digital electronics.

Research Track Leader: Dr. Zhe Zhao. 

Optical neural networks offer massive parallelism across multiple degrees of freedom (DoFs), including wavelength, frequency, polarization, phase, amplitude, mode, and time. We therefore investigate the multiplexing of these multiple DoFs to enhance optical computing performance, achieving ultra-high computing capacity and density together with ultra-low power consumption and system latency. To this end, we develop novel architectures based on both free-space diffractive layers and integrated photonic chips, enabling high-performance multidimensional optical computing. By exploring hybrid multiplexing technologies, we aim to improve the key performance metrics of optical computing cores by more than tenfold.

Research Track Leaders: Dr. Shabnam Taheriniya and Dr. Rongyang Xu.

Larger-scale photonic integrated circuits (PICs) are enabling increasingly complex optical systems for information processing and other emerging applications. PICs consist of cascaded photonic devices, including directional couplers, ring resonators, and Mach-Zehnder interferometers. However, fabrication variations in individual devices result in device response deviations. These response deviations accumulate throughout cascaded photonic circuits during light propagation. Such accumulation can degrade the system’s performance and reliability. We are working on developing post-fabrication trimming methods based on focused ion beam (FIB) techniques that introduce localized refractive index modifications to calibrate device and system responses. Our goal is to further reduce trimming-induced excess loss while improving the scalability of post-fabrication trimming, ultimately extending these methods from individual device calibration to chip-level calibration.