Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Lecture course "Integrated Nanophotonics" by Dmitry Fedyanin

DmitryFedyanin
October 05, 2017

Lecture course "Integrated Nanophotonics" by Dmitry Fedyanin

Lecture 2:

DmitryFedyanin

October 05, 2017
Tweet

More Decks by DmitryFedyanin

Other Decks in Science

Transcript

  1. 3 THE ORIGIN Three revolutionary inventions 1) First laser in

    1960 (T.H. Maiman) 2) Breakthrough in the semiconductor industry a) Semiconductor integrated circuits, 1958-59 (J. Kilby & R. Noyce)
  2. 4 THE ORIGIN Three revolutionary inventions 1) First laser in

    1960 (T.H. Maiman) 2) Breakthrough in the semiconductor industry a) Semiconductor integrated circuits, 1958-59 (J. Kilby & R. Noyce) b) GaAs LED, 1961 (J.R. Biard & G. Pittman)
  3. 5 THE ORIGIN Three revolutionary inventions 1) First laser in

    1960 (T.H. Maiman) 2) Breakthrough in the semiconductor industry a) Semiconductor integrated circuits, 1958-59 (J. Kilby & R. Noyce) b) Electrically pumped semiconductor laser, 1962 (R. N. Hall et al.)
  4. 6 THE ORIGIN Three revolutionary inventions 1) First laser in

    1960 (T.H. Maiman) 2) Breakthrough in the semiconductor industry a) Semiconductor integrated circuits, 1958-59 (J. Kilby & R. Noyce) b) Electrically pumped semiconductor laser, 1962 (R. N. Hall et al.) b)* Continuous wave operation at room temperature using double heterostructures, 1970 (Z.I. Alferov et al.) Microelectron. J. 36, 950 (2005)
  5. 7 THE ORIGIN Three revolutionary inventions 1) First laser in

    1960 (T. H. Maiman) 2) Breakthrough in the semiconductor industry a) Semiconductor integrated circuits, 1958-59 (J. Kilby & R. Noyce) b) Electrically pumped semiconductor laser, 1962 (R. N. Hall et al.) 3) Optical fiber – loss reduction, 1965 (C. K. Kao & G. A. Hockham )
  6. 8 THE ORIGIN Optics Optics Photonics Electro-optics Optoelectronics Acousto-optics Quantum

    electronics Quantum optics Nonlinear optics Optical Communications Optical sensing Integrated photonics
  7. 9 THE ORIGIN What is “integrated photonics”? • The term

    “integrated optics” was introduced in 1960. • The term emphasizes the similarity to integrated electronics. • The idea behind the term is to fabricate an optical integrated circuit, which consists of various elements (both passive and active) fabricated on the same substrate and connected with optical transmission lines, called optical waveguides.
  8. 10 THE ORIGIN What is integrated photonics? Electro-optics Optoelectronics Acousto-optics

    Quantum optics Nonlinear optics Waveguide technology Integrated photonics
  9. 11 MOTIVATION Light at wavelength of λ = 1550 nm

    can give the maximum bandwidth of up to 194x1012 Hz! 1% of this value is 2 THz, which is roughly equal to a bit rate of 2 Tbit/s.
  10. 12 MOTIVATION Light at wavelength of λ = 1550 nm

    can give the maximum bandwidth of up to 194x1012 Hz! 1% of this value is 2 THz, which is roughly equal to a bit rate of 2 Tbit/s. 2 THz = 4x105 analog video channels at a bandwidth of 5 MHz = 4x108 telephone calls at 5 kHz per call
  11. 13 MOTIVATION Light at wavelength of λ = 1550 nm

    can give the maximum bandwidth of up to 194x1012 Hz! 1% of this value is 2 THz, which is roughly equal to a bit rate of 2 Tbit/s. 2 THz = 4x105 analog video channels at a bandwidth of 5 MHz = 4x108 telephone calls at 5 kHz per call How to use this potential? Problems: • Electronics components did not operate at such speeds (+ remember that this was in 1960s). • A new medium was needed to carry light.
  12. 14 MOTIVATION Optical fiber: optics versus electronics Twisted-pair cable Coaxial

    cable - R.G. Hunsperger, Integrated optics. Theory and technology. Springer, 2009. Only very short-distance commutations or operation at very low frequencies. Coaxial cables can be used for data transmission over relatively short distances, at frequencies up to about 100 MHz.
  13. 15 MOTIVATION Reasons: 1. Higher bandwidth 2. Higher latency (lower

    delays) 3. Lower power consumption 4. Lower crosstalk 5. Smaller dimensions 6. Less number of guides
  14. 18 FIRST CONCEPTS Transmitter based on GaAs - F.A. Blum

    et al., Naval Res. Rev. 28, 1 (Feb. 1975). - T. Tamir, Integrated Optics, Springer, 1979. Lithium niobate modulator Now, fiber-coupled LiNbO 3 optical modulators are commercially available at a very low price. The typical modulation bandwidth is 10-40 Gb/s.
  15. 19 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Figure: Nature 466, 690-691 (2010) Since 80s and early 90s
  16. 20 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Since the late 90s Figure: http://www.furukawa.co.jp Figure: http://www.fibernetworx.com 2003 – IEEE 802.3ae 10 Gbit Ethernet
  17. 21 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm
  18. 22 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Dr. Bert Jan Offrein, Manager Photonics, IBM Research – Zurich At Swisslasernet Workshop, October 2010 At IBM Technology Forum 2010 – «Photonics», November 2010
  19. 23 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm March 2012 – IBM “Holey Optochip” Distance: up to 200 m (fiber limit) Chip size: 5.2 mm x 5.8 mm | 24 receiver and 24 transmitter based on 850 nm VCSEL Bandwidth: 960 Gbit/s (480 Gbit/s in each direction simultaneously) Power consumption: 4.7 W < 5 pJ/bit for the WHOLE CHIP http://www-03.ibm.com/press/us/en/pressrelease/37095.wss Blue Gene/Q was fully deployed in 2012 at Lawrence Livermore National Laboratory (LLNL), the system, named “Sequoia”, is expected to achieve 16+ PFLOPS. Number 1 in top500 list in 2012.
  20. 24 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm State-of-the-Art VCSEL Transmitters5 [1] Y.-C. Chang, et al., Electron. Lett. 43, 1022 (2007). [2] P. Westbergh, et al., Electron. Lett. 45, 366 (2009). [3] J.-W. Shi, et al., IEEE Photon. Journal 2, 960 (2010). [4] S. Imai, et al., Proc. IEEE ISLC, 152 (2010). [5] P. Moser, et al., Appl. Phys. Lett. 98, 231106 (2011). Un. California, Santa Barbara1 Chalmers Un. of Technology2 Un. Taiwan/ IQE Ltd., UK3 Furukawa Electric Co.4 Technische Un. Berlin/ VI Systems5 Supported by IBM, Agilent Year 2007 2009 2010 2010 2011 Bandwidth 35 Gbit/s 32 Gbit/s 15.5 Gbit/s 10 Gbit/s 17 Gbit/s Energy per bit 360 fJ 460 fJ5 270 fJ5 180 fJ 83 fJ Wavelength 980 nm 850 nm 850 nm 1060 nm 850 nm VCESL output power per bit 70 fJ 130 fJ5 80 fJ5 ~40 fJ 14 fJ
  21. 25 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Figures: www-03.ibm.com 11 Sep 2003 – IBM and Agilent Technologies announced “Terabus”, multi-year $30 million project with the Defense Advanced Research Projects Agency (DARPA) The same principle as for node- to-node interconnects, which is based on VCSELs and photodiodes, but shorter distances and polymer waveguides instead of optical fibers.
  22. 26 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Figures: www.intel.com The 50G Silicon Photonics Link 2010, Intel Si / III-V laser LOW power efficiency HIGH energy per bit ratio
  23. 27 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Cell BE processor (IBM/Toshiba/Sony) Developed in 2001 – 2004 Released in 2005 8 (SPE) + 1 (PPE) cores Clock rate > 4 GHz Performance > 256 GFLOPS Figure: ibm.com
  24. 28 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Conventional Cache Conventional CPUs perform all their operations in registers which are directly read from or written to main memory. Caches work by storing part of the memory the processor is working on. Synergistic Processor Element Local Store SPE uses 256 KB “local store”. The SPE operates on registers which are read from or written to the local stores. 16 Byte (128 bit) can be moved per cycle to or from the local store giving 64 GB/s = 512 Gbit/s. The local store can access main memory in blocks of 1Kb minimum (16Kb maximum) but the SPE cannot act directly on main memory (they can only move data to or from the local stores). SPE performance was 32 GFLOPS, while Pentium D 840 could give < 13 GFLOPS and Nvidia GeForce 6800 Ultra < 40 GFLOPS.
  25. 29 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Cell BE processor (IBM/Toshiba/Sony) 8 (SPE) + 1 (PPE) cores Clock rate 3.2 GHz Data bus bandwidth: 2 x 4 x 25.6 GB/s = 1.6 Tbit/s 2.4 Tbit/s (peak) Data bus width: 2 x 4 x 128 bit = 1024 bit Command bus: 0.8 Tbit/s for coherent commands 1.6 Tbit/s for non-coherent Figures: ibm.com
  26. 30 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Figures: http://www.ibm.com Optical layers not only connect different electronic elements but also ROUTE THE TRAFFIC.
  27. 31 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Optical interconnect layers Electrical interconnect layer Processing and memory layers
  28. 32 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Optical interconnect layers Electrical interconnect layer Processing and memory layers
  29. 33 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Optical interconnect layers Electrical interconnect layer Processing and memory layers
  30. 34 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Optical interconnect layers Electrical interconnect layer Processing and memory layers
  31. 35 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Optical interconnect layers Electrical interconnect layer Processing and memory layers
  32. 36 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Yurii Vlasov IBM Research Division Tutotial given at ECOC, September 22, 2008
  33. 37 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Year Who Band width Energy pet bit Comments 2000 Swiss Federal Institute of Technology, Lausanne 10 MHz N/A [link] Mach–Zehnder p+–p–n+ First fully integrated Mach– Zehnder interferometer in silicon-on-insulator technology. p+-n+ period: 30 μm Waveguide height: h = 10 μm Waveguide width: W = 10 μm Modulator length: 20 mm Modulating current density: 2 kA/cm2 Modulator based on Mach– Zehnder Interferometer
  34. 38 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Year Who Band width Energy pet bit Comments 2000 Swiss Federal Institute of Technology, Lausanne 10 MHz N/A [link] Mach–Zehnder p+–p–n+ 2004 Intel 1 Gb/s N/A [link] Mach–Zehnder for Chip-to-Chip 2005 Intel 10 Gb/s >30 pJ [link] Mach–Zehnder for Chip-to-Chip 2005 M. Lipson group, Cornell University 1.5 Gb/s N/A (~1 pJ) [link] Ring–resonator Ring diameter 12μm Modulation depth 9dB 2007 IBM 10 Gb/s 5 pJ [link] Mach–Zehnder 2007 M. Lipson group, Cornell University 12.5 Gb/s ~300 fJ [link] Ring–resonator Ring diameter 12μm 2008 Sandia National Labs 10 Gb/s 85 fJ [link] Microdisk Reverse-biased vertical p–n junction Modulation depth 8dB
  35. 39 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Year Who Band width Energy pet bit Comments 2008 Microphotonics Center in MIT and BAE Systems 1.2 GHz 50 fJ [link] SiGe absorption modulator 2009 Kotura Inc. and Sun Microsystems (ORACLE) + DARPA 10 Gb/s 50 fJ [link] Ring-resonator p++–p–n–n++ Modulation depth 6.5dB 2010 HP 6 Gb/s 45 fJ [link] Ring-resonator Light wavelength ~1320nm Ring diameter 5μm Modulation depth 7dB
  36. 40 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm -L. Chen, K. Preston, S. Manipatruni, M. Lipson, Integrated GHz silicon photonic interconnect with micrometer-scale modulators and detectors, Opt. Express 17, 15248 (2009). Bandwidth: 3 Gb/s Modulator voltage swing: 0.5 V Q c V pp (0–1): 55 fJ Average switching energy: 14 fJ DC energy consumption: 72 fJ Total modulator consumption: 86 fJ Detector length: 40 μm Detector bias: 1.0 V Detector dark current: 1.4 μA Detector capacitance: ~4 fF Detector responsivity: 0.8–0.9 A/W Detector consumption, with an averaged output DC current 0.1 mA: 33 fJ Total energy per bit: ~120 fJ (excluding an external laser, which has an optical power of 0.6 mW)
  37. 41 OPTICAL INTERCONNECTS Wide Area Local Area Node-to-Node Board-to-Board Chip-to-Chip

    On-Chip > 10 km 10 – 2000 m 1 – 100 m 0.2 – 1 m 5 – 100 mm < 20 mm Prith Banerjee Director HP Labs At DesignCon 2012, January 2012 HP “Corona” Date: 2019 Number of cores: 256 Process: 16 nm 3D packaging with 64 four-core clusters interconnected by an all-optical DWDM crossbar Per-core: 4 Threads 5 GHz Frequency http://www.hpl.hp.com/personal/Robert_Schreiber/papers/2008%20ISCA/npa.pdf
  38. 42 INTEGRATED PHOTONICS There are many other modern application of

    integrated (nano)photonics, which are out of the scope of this lecture: • Chemical and biosensing • Graphics processors • Ultrafast encryption • Quantum optics: from quantum cryptography to quantum computers • Deep learning • All-optical microprocessor
  39. 44 NEXT CLASS Light propagation in different media. Complex dielectric

    function. Electromagnetic energy and its relation to the dielectric function. Group velocity and energy velocity. Transmission/reflection at the interface.