Skip to main content

Weighted Spintronic-Nano-Oscillator-based Neuromorphic Computing System Assisted by laser for Cognitive Computing

Periodic Reporting for period 1 - SpinAge (Weighted Spintronic-Nano-Oscillator-based Neuromorphic Computing System Assisted by laser for Cognitive Computing)

Reporting period: 2020-10-01 to 2022-03-31

SpinAge's goal is to design and implement a high-speed and low-power spintronic-based (spin oscillator) neuromorphic computing system where neurons are implemented using spin oscillators where weights are in close proximity or even co-integrated with the neurons. To ease the processing of the data (meaning the oscillation of the spin oscillators), we use laser not only through heating but also other processes (for example Seebeck effect). The main challenges we have been facing are the variations of the spintronic devices, the mismatches between different technologies (spintronic devices and memristors, or spintronic devices and CMOS circuits, and so on), and probably in a later stage integration. Although the effect of laser on SHNOs are destructive, there are potentials for positive effect of laser on vortex oscillators which requires more investigation.
The main objectives of the project to reach the main goal of the project are:
1.Design and implement novel low-energy and compact WSHNOs and all-spin WSTNOs assisted by an integrated low-energy laser pulse (<50 fJ) through temperature increase, Seebeck effect or super-diffusive current.
2.Design and implement an on-chip integrated laser chip controlled by CMOS chip enabling the ease of oscillation with a higher speed and lower energy.
3.Design and implement novel memristors to improve the performance of the WSHNOs.
4.Design low power CMOS interface for WSHNOs and WSTNOs and laser control.
5.Integrate spintronics, electronics, and photonics on a single platform.
We have done tremendous effort towards the final goal of the project by carefully addressing the challenges of the project (envisioned by the project) with the entailed risks. The effort towards different objectives is elaborated in the 1RP report, however in short, we have designed two read-out circuitries for WSHNOs and WSTNOs (designed, fabricated and tested). New WSHNOs and WSTNOs have been designed, fabricated and tested, and the next step will be heterogenous integration of the CMOS chips and the spin oscillators which will give two distinct NCSs with different frequency range (medium and high frequency). In a close collaboration between spintronic partners and material study partner (UG), the device characterization through THz has started and it is going on, which is hoped to bring us a platform for quick test of the samples exposed to laser. In a close collaboration between NanOsc and POLIMI, also progress in building new materials for memristors to be co-integrated with the SHNOs has been made. laser arrays are ready to be characterized and in a later stage integrated with the spintronic devices to study the effect of laser (with the maximum available power from the VCSELs) on spintronic and probably memristor devices. Integrating of all these devices are challenging, but good progress with a clear plan for monolithic integration is in place.
We expect to have (probably) the first ever monolithic spintronic-cmos-photonic integrated NCS that paves the road for more complex system combining all these technologies into a single platform. Although the project is a high risk project especially when it comes to power consumption (energy), if the integration of the whole system is successful and we achieve an operational system even with a comparable energy consumption with its counterparts, we can expect to see more integrated systems with other functionality with a focus on high-performance processing.