Packaging and Interconnect Considerations in Neuromorphic Photonic Accelerators

Nezami, Mohammadreza Sanadgol; Lima, Thomas Ferreira de; Mitchell, Matthew; Yu, Shangxuan; Wang, Jing; Bilodeau, Simon; Zhang, Weipeng; Al-Qadasi, Mohammed; Taghavi, Iman; Tofini, Alexander; Lin, Stephen; Shastri, Bhavin J.; Prucnal, Paul R.; Chrostowski, Lukas; Shekhar, Sudip

doi:10.1109/jstqe.2022.3200604

Cited by 9 publications

(6 citation statements)

References 64 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These breakthroughs demonstrate a promising path for manufacturing large-scale silicon photonic-electronic chips based on the rapid MPW prototyping service. By leveraging photonic packaging [16], such as photonic wire bonding [17] and flip-chip die bonding for photonic-electronic co-integration [18], silicon photonic chips can be calibrated, reprogrammed, and trained efficiently by using electronic microcontrollers with high reproducibility for neuromorphic computing [19,20].…”

Section: Introductionmentioning

confidence: 99%

Fabrication and integration of photonic devices for phase-change memory and neuromorphic computing

Zhou,

Shen,

Yang

et al. 2024

Int. J. Extrem. Manuf.

View full text Add to dashboard Cite

In the past decade, there has been tremendous progress in integrating chalcogenide phase-change materials (PCMs) on the silicon photonic platform for non-volatile memory to neuromorphic in-memory computing applications. Especially, these non von Neumann computational elements and systems benefit from mass manufacturing of silicon photonic integrated circuits (PICs) on 8-inch wafers using 130-nm complementary metal-oxide semiconductor (CMOS) line. Chip manufacturing based on the deep-ultraviolet (DUV) lithography and electron-beam lithography (EBL) enable rapid prototyping of PICs, which can be integrated with high-quality PCMs based on the wafer-scale sputtering technique as a back-end-of-line (BEOL) process. In this article, we overview recent advances of waveguide integrated PCM memory cells, functional devices, and neuromorphic systems, with an emphasis on fabrication and integration processes to attain the state-of-the-art device performance. After a short overview of PCM based photonic devices, we discuss the materials properties of the functional layer as well as the progress on the light guiding layer, namely, the silicon and germanium waveguide platforms. Next, we discuss the cleanroom fabrication flow of waveguide devices integrated with thin films and nanowires, silicon waveguide and plasmonic microheaters for electrothermal switching of PCMs and mixed-mode operation. Finally, the fabrication of photonic and photonic-electronic neuromorphic computing systems is reviewed. These systems consist arrays of PCM memory elements for associative learning, matrix-vector multiplication, and pattern recognition. With large-scale integration, neuromorphic photonic computing paradigm holds the promise to outperform digital electronic accelerators by taking the advantages of ultra-high bandwidth, high speed, and energy efficient operation in running machine learning algorithms.

show abstract

Section: Introductionmentioning

confidence: 99%

Fabrication and integration of photonic devices for phase-change memory and neuromorphic computing

Zhou,

Shen,

Yang

et al. 2024

Int. J. Extrem. Manuf.

View full text Add to dashboard Cite

show abstract

“…This distinction renders the SimTEC approach a viable method for precisely adjusting the chip temperature, complementing broader system-level thermal management techniques. A parallel scenario arises in the neuromorphic photonic accelerator module, 15 where specific components, including the core photonic processor, DFB lasers, and SOAs, are susceptible to thermal fluctuations. This susceptibility presents an opportunity for the SimTEC device to provide a temperature stabilization solution seamlessly integrated into the module.…”

Section: Introductionmentioning

confidence: 99%

Substrate integrated micro-thermoelectric coolers in glass substrate for next-generation photonic packages

Gupta,

Tanwar,

et al. 2024

J. Optical Microsystems

View full text Add to dashboard Cite

Managing the temperature of photonic chips within intricate electro-optic packages poses a notable challenge concerning the thermal crosstalk between the photonic chip, electronic chip, and the chip-fiber connection point. This is a multifaceted problem and requires packaging solutions that cannot only address high-performance thermal management but must also be scalable to high volumes. Glass has long been thought of as a suitable platform for next-generation photonic packaging due to its low thermal conductivity, which minimizes unwanted heat transfer between electronic and photonic components. Achieving proper thermal isolation between the chips and the chip-fiber interface necessitates a microscale thermal solution that guarantees accurate temperature regulation of the photonic circuitry without disrupting the optical coupling interface with the fiber array, due to the presence of epoxy used for fiber attachment. We propose a technique for the development of a substrate-integrated microthermoelectric cooler (SimTEC) for the effective temperature control of the electronic and photonic integrated devices. The proposed device uses glass substrate vias that are half-filled with p and n-type thermoelectric materials and the other half with copper. A COMSOL multiphysics model is developed to study the variations in the cooling performance of this SimTEC device based on changes in the via parameters. Interestingly, the maximum range of temperature gradient variation for SimTEC is 6 times greater compared to that of equivalent free-standing micro-TEC pillars. However, there are some challenges associated with implementing this method, as the temperature gradient (or cooling effect) achieved by SimTEC still falls short of that achieved by the free-standing micro-TEC pillars.

show abstract

“…[24] Furthermore, PWBs have been employed to address the complex packaging requirements of neuromorphic photonic accelerator architectures. [25,26] PWBs are 3D waveguides created by using direct-write twophoton polymerization. [27,28] Therefore, they provide potential superior scalability, making them an attractive alternative to the existing methods.…”

Section: Introductionmentioning

confidence: 99%

“…[ 24 ] Furthermore, PWBs have been employed to address the complex packaging requirements of neuromorphic photonic accelerator architectures. [ 25,26 ]…”

Section: Introductionmentioning

confidence: 99%

Plug‐and‐Play Fiber‐Coupled Quantum Dot Single‐Photon Source via Photonic Wire Bonding

De Gregorio,

Yu,

Witt

et al. 2023

Adv Quantum Tech

Self Cite

View full text Add to dashboard Cite

The collection of single‐photon emission from a quantum dot (QD) in a Bragg waveguide through a photonic wire bond (PWB) via free‐space resonant frequency pumping at 1.6 K is demonstrated. The in‐fiber single photons show a small multiphoton contribution, quantified by a low second order photon autocorrelation value of (background‐corrected) or (raw data). The decay time of the QD is measured to be ps. The PWB obviates the need for in‐cryostat alignment of the single‐photon source with an optical fiber and thus offers a route to scalable integration of quantum photonic devices in a cryogenic environment. Uniquely, the approach combines the QD‐waveguide technique, enabling resonant driving of individual QDs without the need for cross‐polarization filtering, and the PWB for deterministic, alignment‐free coupling of single‐photon sources to optical fibers.

show abstract

Packaging and Interconnect Considerations in Neuromorphic Photonic Accelerators

Cited by 9 publications

References 64 publications

Fabrication and integration of photonic devices for phase-change memory and neuromorphic computing

Fabrication and integration of photonic devices for phase-change memory and neuromorphic computing

Substrate integrated micro-thermoelectric coolers in glass substrate for next-generation photonic packages

Plug‐and‐Play Fiber‐Coupled Quantum Dot Single‐Photon Source via Photonic Wire Bonding

Contact Info

Product

Resources

About