Silicon photonics integration roadmap for applications in computing systems Bert Jan Offrein Neuromorphic Devices and Systems Group © 2016 IBM Corporation Outline . Photonics and computing? – The interconnect bottleneck – The Von Neumann Bottleneck . Optical interconnects for computing systems – Optical interconnects roadmap – CMOS Silicon Photonics – Novel functionalities by adding new materials . Photonic synaptic elements for Neural Networks – Motivation – Photonic Synaptic Processor – Non-volatile optical memory elements . Conclusions Why Optics – The interconnect bottleneck . Physics of electrical links - going towards higher bandwidth – Increased loss – Increased crosstalk – Resonant effects . Physics of optical links – 190 THz EM waves – 50 THz Bandwidth available (1300-1600 nm) . Larger bandwidth X length product of optics – Electrical coax cable: ~ 100 MHz km – Multimode fiber: ~ 500 MHz km – Single mode fiber: > 5000 MHz km . Lower propagation loss of optical cables – Electrical coax cable: ~ 1 dB/m – Multimode fiber: ~ 3 dB/km – Single mode fiber: > 0.3 dB/km . Power efficiency . Larger density of optical links 3 Electrical and Optical Communication between two processors Electrical Physics of RF EM waves The signal is the carrier Transceiver V laser Optical Physics of optical EM waves The signal is modulated on an optical carrier driver amplifier modulator V • 1000 x Larger bandwidth Scalability & Optical communication: • 1000 x Lower loss • 100 x Larger distance Power efficiency !!! However, many more components and assembly steps required !!! 4 Where to transition from electrical to optical ? On package On board edge board At On processor memory processor board backplane Better performance, more disruptive, more development required 5 2011: IBM Power P775, High Performance Supercomputing System Memory Processor Fibers Fibers Processor Package Avago MicroPODTM 6 Integration? Looking back, electronics Pictures taken at: Whirlwind, MIT, 1952 EAI 580 patch panel, Electronic Associates, 1968 Today’s state of computing is based on: - Integration and scaling of the logic functions (CMOS electronics) - Integration and scaling of the interconnects (PCB technology & assembly) For optical interconnects, this resembles: - Electro-optical integration and scaling of transceiver technology - Integration of optical connectivity and signal distribution 7 Could one INTEGRATE the electrical and optical functions into the system? Transmit & receive optical signals ? Distribute optical signals Vision: Electrical and optical communication embedded in a computing system 8 CMOS Silicon photonics Integrate electrical & optical functions in silicon 9 4 λ x 25 Gb/s optical transceiver demonstration Rx3 Tx0 Rx2 Tx1 Rx1 Tx2 Rx0 Tx3 as transmitted from Tx0 as measured on Rx3 Demonstration of a flip chip mounted 100G transceiver with four wavelength multiplexing at 25 Gb/s each. 10 CMOS Embedded III-V on silicon technology Electrical SiO2 contacts BEOL III-V Front-end Si SiO2 FEOL Si wafer CMOS Si Photonics … + III-V functionality . Overcome discrete laser and assembly cost . New functions, directly combining electronics, passive and active photonics 11 Silicon photonics integration roadmap V V Directly laser modulated laser laser driver driver driver modulator modulator amplifier amplifier amplifier CMOS Siliconphotonics V V V Laser specs Wavelength 1300 nm Optical Power 20 mW 10 mW 2.5 mW Electrical Power 200 mW 100 mW 50 mW Total link power Laser: 200 mW Laser: 100 mW Laser: 50 mW @ 25 Gb/s Modulator: 900 mW Modulator: 900 mW Laser driver: 100 mW TIA: 100 mW TIA: 100 mW TIA: 100 mW Total: 1200 mW Total: 1100 mW Total: 250 mW 12 IBM Research - Zurich Processing scheme 5 InAlGaAs quantum wells SiO 2 (MOCVD) III-V epi layer epi layer SiO2 SiPh wafer SiPh wafer Wafer bonding Feedback grating SiO2 SiO2 Substrate removal III-V structuring MQW section SiO2 Metallization 13 Bert Jan Offrein © 2017 IBM Corporation IBM Research - Zurich Optically pumped ring laser Measured FSR: 0.194 nm Estimated FSR from ring: 0.203 nm Estimated FSR from III-V: 0.266 nm Lasing with feedback from silicon photonics Directional coupler output Gain section 14 Bert Jan Offrein © 2017 IBM Corporation IBM Research - Zurich Electrically pumped lasing Optical spectrum at 110 K 5000 4000 Lasing modes 3000 2000 Counts (a.u.) Counts Spontaneous emission Laser devices: 10 dB optical loss at room 1000 temperature Cooling down increases gain Increased gain can overcome loss 0 Pulsed electrical pumping 1160 1180 1200 1220 1240 1260 Wavelength (nm) 15 Bert Jan Offrein © 2017 IBM Corporation Could one INTEGRATE the optical functions into the system? Transmit & receive optical signals ? Distribute optical signals Vision: Electrical and optical communication embedded in a computing system 16 Outline . Photonics and computing? – The interconnect bottleneck – The Von Neumann Bottleneck . Optical interconnects for computing systems – Optical interconnects roadmap – CMOS Silicon Photonics – Novel functionalities by adding new materials . Photonic synaptic elements for Neural Networks – Motivation – Photonic Synaptic Processor – Non-volatile optical memory elements . Conclusions Neuromorphic hardware for big data analytics GPU . Today’s status on deep neural networks • Software based on Von Neumann systems • Training is the bottleneck – HPC required • GPU accelerators – processing - memory bottleneck . Fast and efficient neural network data processing 1. Analog approximate signal processing Metal electrode 2. Tight integration of processing and memory – 10’000x improvement using crossbar arrays Tunable resistance . Hardware implementations – Electrical crossbar arrays Metal electrode – Photonic crossbar arrays Synaptic element Crossbar array © 2017 IBM Corporation 18 Accelerated learning: Analog crossbar arrays Information flow Update weight proportional to signals on crossbar row and column Feedforward • Increase and decrease of weight Deep Neural • Symmetric behavior for positive and negative updates Network • High weight resolution (~1000 levels) required Physical challenge: Identify material systems fulfilling those Input layer Hidden layers Output layer requirements Synaptic weight Training Inference symmetry 훿 # of levels 푥 target Resistance non-ideal 푥 푊푥 Training cycle © 2017 International Business Machines Corporation Photonic crossbar unit - operating principle Electrical crossbar Photonic crossbar 훿 Forward propagation 훿 Backward propagation Weights 푥 푊푇훿 푥 푊푇훿 푊푥 • Electrical wires 푊푥 • Planar waveguiding • Local weights • Distributed weights • Resistance tuning • Refractive index tuning Writable photorefractive gratings provide the same functionality as the tunable resistive elements in a crossbar unit 20 Copyright © 2017 Photonic crossbar unit . Alternative crossbar physical principle leveraging the photorefractive effect – Demonstrated in 3D free space photonic neural networks in the 90s . i.e. Hughes Research Laboratories – New developments we can leverage . Integrated optic technology . Co-integration of new materials . Non-volatile weights applying the photorefractive effect . Grating writing by interference of optical plane waves in an electro-optic material Strength proportional to product of the amplitudes of the writing beams . Written grating acts as the synaptic interface between plane optical waves Weighted and Weighted combined signal 2 source signals Copyright © 2017 Diffraction from a photorefractive grating . Measurement on a thick GaAs layer Two-wave mixing in bulk GaAs ≈ single synapse Mirror Detection e s r a to r h la e P u b d i f o GaAs m M Beam- S splitter Polarizer Collimator Incident light Detection Mirror 22 IBM Research - Zurich CMOS Embedded III-V on silicon technology Electrical SiO2 contacts BEOL III-V Front-end Si SiO2 FEOL Si wafer CMOS Si Photonics … + III-V functionality • Overcome discrete laser and assembly cost • New functions, directly combining electronics, passive and active photonics • & photorefractive materials 23 Bert Jan Offrein © 2017 IBM Corporation MOCVD-based growth of epitaxial GaAs core and InGaP cladding Development of low-temperature grown GaAs on (110) GaAs n ~ 4.45 1014 cm-3 2 wafers μn ~ 414 cm /Vs . Tg = 440°C with 600°C pre-bake under As for oxideIII-V desorption. Atomically flat layers obtained. AFM Decreasing growth temperature using high V/III ratio results in a 14 -3 semi-insulating (10 cm ) material containing electron trapsR ~ 0.3 nm (EL2 ?). Pt RMS RRMS ~ 0.2 nm I. II. Development of InGaP cladding layer GaAs 107 106 105 104 103 2434 102 1 . Tuning the content of InxGa1-xP allows to grow layers lattice [cps] signal XRD 10 100 matched to GaAs. 106 105 104 103 XRD 102 2437 1 XRD signal [cps] signal XRD 10 . Further development steps include the growth of a thick 100 Scratches – holes ~ 40-50 nm deep 107 InGaP cladding with a low-temperature grown EL2-containing 106 105 104 GaAs. 103 2 2442 XRD signal [cps] signal XRD 10 101 65.0 65.5 66.0 66.5 67.0 2[°] | © 2017 IBM Corporation 24 Conclusions . Photonics technology helps to overcome interconnect bottlenecks – Communication at scale – Von Neumann bottleneck . Applications for computing – Optical interconnects – Training of synaptic weights in neural networks . Extend silicon technology – III-V but also other types such as ferroelectric – Photorefractive, optical gain, switching, optical weights . + large variety of other opportunities and applications © 2017 IBM Corporation 25 Acknowledgments IBM Research – Zurich, Switzerland Stefan Abel, Folkert Horst, Marc Seifried, Gustavo Villares, Roger Dangel, Felix Eltes, Jacqueline Kremer, Jean Fompeyrine, D. Caimi, L. Czornomaz, M. Sousa, H. Siegwart, C. Caer, Y. Baumgartner, D. Jubin, N. Meier, A. La Porta, J. Weiss, V. Despandhe, U. Drechsler Co-funded by the European Union Horizon 2020 Programme and the Swiss National Secretariat for Education, Research and Innovation (SERI).