OPEN Lab Prof. Sorger Volker Sorger 14 th NSF-Korea Nanotechnology Forum 2017 aJ/bit Modulators and Photonic Neuromorphic Computing
OPEN Lab Prof. Sorger
Volker Sorger
14th NSF-Korea Nanotechnology Forum 2017
aJ/bit Modulators and Photonic Neuromorphic Computing
OPEN Lab Prof. Sorger
Orthogonal Physics Enabled Nanophotonics (OPEN) lab
compared to any photonic (non-plasmonic) mode and cavity structures where any placement of the metallic contact
close to the optical mode will introduce intolerable losses. This is different for our plasmonic mode, which is
inherently lossy, but the polaritonic (matter-like) mode allows to scale-down the device into a few micrometer small
device (a reduction of a factor of 100) compared to traveling-wave Silicon-based modulators. This ‘in-the-device-
basing’, as suppose to biasing the device few to 10’s of micrometer away from the active region. As such, the overall
design allows for a more compact overall footprint. Lastly, reducing the dielectric thickness (tox = 5 nm), improves the
electrostatics enabling a sub-1 Volt modulation performance. Despite this increasing the capacitance, the lower Rc and
small device area (5 mm2) the energy consumption is relatively low 18 fJ/bit. However, this can be further improved by
only biasing the device in the steepest region of the transfer function (e.g. 0-0.3V) for a small-signal modulation, and
narrowing the waveguide width from (currently 1 mm) to a SOI diffraction-limited waveguide width of 250 nm
reducing the E/bit to low 160aJ (Fig. 1d).
In summary, we have experimentally shown a graphene-based silicon-photonics integrated electro-absorption
modulator operating at telecom frequencies. Utilizing a hybrid-photon-plasmon modes high group index, field-
concentration to increase the optical overlap factor with the thin graphene active material, and improving the
electrostatics by reducing the gate oxide allows for a steep switching transfer function to enable sub-1 Volt
modulation, which has positive impacts on energy-efficiency. Further improvements allow for a 160aJ/bit efficient
modulator using this approach platform, thus paving the way for next-generation nanophotonics devices [13]. !
References [1] R. A. Soref, D. L. McDaniel, B. R. Bennett, “Guided-Wave Intensity Modulators Using Amplitude and-Phase Perturbations”
IEEE Journal of Lightwave Technology, vol. 6, no. 3, (1988).
[2] K. Liu, S. Sun, A. Majumdar, V. J. Sorger, “Fundamental Scaling Laws in Nanophotonics” Sci. Rep., 6, 37419 (2016).
[3] S. K. Pickus, S. Khan, C. Ye, Z. Li, and V. J. Sorger, “Silicon Plasmon Modulators: Breaking Photonic Limits” IEEE Phot.
Soc., 27, 6 (2013).
[4] M. Liu, X. Yin, E. Ulin-Avila, B. Geng, T. Zentgraf, L. Ju, F. Wang, and X. Zhang, “A graphene-based broadband optical
modulator,” Nature, vol. 474, no. 7349, pp. 64–67, (2011).
[5] M. Liu, X. Yin and X. Zhang, "Double-Layer Graphene Optical Modulator", Nano Letters, 12, no. 3, pp. 1482-1485, (2012).
[6] C. Ye, S. Khan, Z. R. Li, E. Simsek, V. J. Sorger (2014). “λ-size ITO and graphene-based electro-optic modulators on
SOI”. Selected Topics in Quantum Electronics, IEEE Journal of, 20(4), 40-49 (2014).![7] C. Ye, K. Liu, R. Soref, V. J. Sorger, “3-Waveguide 2x2 Plasmonic Electro-optic Switch” Nanophot., 4, 1, pp. 261-268 (2015).
[8] Z. Ma, M. Tahersima, S. Khan, V. J. Sorger, IEEE J. Sel. Topics in Quantum Electronics, 23(1), 3400208 (2017).
[9] R.Amin, C.Suer, Z.Ma, J. B. Khurgin, R. Agarwal, V. J. Sorger, “Active Material, Optical Mode and Cavity Impact on electro-
optic Modulation Performance”, arXiv:1612.02494, (2017)
[10] V. J. Sorger, R. Ma, C. Huang, Z. Li, M. Liu, and X. Zhang, "Graphene, Plasmonic and Silicon Optical Modulators,"
Conference on Lasers and Electro-Optics - IQEC,OSA, (2013).
[11] Z. Ma, Z. Li, K. Liu, C. Ye, V. J. Sorger, “Indium-Tin-Oxide for High-performance Electro-optic Modulation”, Nanophot., 4, 1
(2015).
[12] Ansell, D. et al. Hybrid graphene plasmonic waveguide modulators. Nature Commun. 6, 8846 (2015). [13] A. Fratalocchi, C. M Dodson, R. Zia, P. Genevet, et al. “Nano-optics gets practical”, Nature Nanotech. 10, 11-15 (2015).
Figure 1. a, Schematic of a hybrid-photon-plasmon
graphene-based electro-absorption modulator. The
modulation mechanism is based on Pauli-blocking upon
gating the Fermi level of graphene. b, Silicon waveguide-
integrated modulator. A cw laser (l = 1.55 mm) is fiber
coupled into the SOI waveguide via grating couplers.
Device length, L = 8 mm, tox = 5 nm. c, Electric field
density across the active MOS region of the modulator
showing an enhanced field strength coinciding with the
active graphene layer. This improves the optical overlap
factor by about 25%. Taking into consideration the grain
boundaries introduced during the metal deposited creates
in-plane field vectors inside the graphene layer. d
Modulator transfer function; normalized modulation depth
at different drive voltages (VD). The modulator
performance yields a high extinction ratio of 0.25dB/mm,
due to the combination of the plasmonic MOS mode
enhancing the electroabsorption in the active region (see
text for details).
!
Modulator
(i) (ii)
(iii) (iv)bias
input
output
Electro-OpticNonlinearity
ElectronicNonlinearity
pump
output
input
bias
output
input
output
pump
input
inputs
(c)
T(λ) output
input
t t t
λ1 λ2 λ3 λ1 λ2 λ3 λ4
photo-detector
spectralfilter
I(t) I(t) i(t) I(t)
t
weights summation nonlinearity output
(a)
(b)
!x1
x2
x3
w1
w2
w3
x⌃
f (x⌃ )
X
i
wi x ix4
modulator/laser
Laserelectrical wire
lightwave
Atto-Joule Optoelectronics
Photonic Functions
Analogue Computing
Today
Sorger, Zhang lab, Nature Photonics (2008) Sorger, Zhang lab, Nature (2009) Sorger lab , IEEE Photonics (2013) Sorger lab, Altug lab, Nature Nanotech. (2015) Sorger lab, Majumdar Lab, Sci. Reports (2016) Sorger lab, Optics Letters (2016) Sorger lab, IEEE STQE (2014 & 2017)
Sorger lab, IEEE Photonics (2015) Sorger lab, Nanophotonics (2016) Sorger lab, Optics Letters (2016) Sorger lab, El-Ghazawi lab, IPCC (2017) Sorger lab, El-Ghazawi lab, Mircoprocess. & MS (2017) Sorger lab, Frontiers in Optics (2017)
Sorger lab, Nanophotonics (2017) Sorger Lab, Grace lab, Biofabrication (2017) Sorger lab, IEEE Rebooting Computing (2017) Prucnal lab, Sorger lab, (in preparation)
OPEN Lab Prof. Sorger OPEN Lab
Prof. Sorger
5cm x 1cm
Modulators = Optical Transistors
EO Modulator
(=SS@FET)
OPEN Lab Prof. Sorger
Power-BW Roadmap
V. J. Sorger, Nature Nanotech., 10, 11-15 (2015)
Power = E bit( ) × Bitrate( )
J
bit
bit
s=
J
s= W
é
ëê
ù
ûú
OPEN Lab Prof. Sorger
EO Modulators
Sorger, Nanoph. (2012) Huang, IEEE Phot. (2013)
Ye, IEEE STQE (2014) Khan, (under review)
Ma, IEEE STQE (2017)
LASER &PHOTONICSREVIEWS
Vol. 9 | March 2015
Review and perspective on ultra-fast wavelength-size electro-optic modulators
Ke Liu, Chenran Ye, Sikandar Khan, Volker J. Sorger
www.lpr-journal.org Vol. 9 | March 2015 www.lpr-journal.org
Selective switching of individual multipole resonances in single dielectric nanoparticles
Pawel Wozniak, Peter Banzer, Gerd Leuchs
LASER &PHOTONICSREVIEWS
LPOR_9_2_cover.indd 2 11/03/15 5:05 PM
Sarpkaya, (in prep)
Grating period: 1um Fill Factor: 50%
Slot Gap < 50nm achieved Single EBL Exposure
ER = 1.2dB/um (single layer graphene)
ON
OFF 20 mm
5 mm
Control
2H 2H
1T’ Tr
ansm
issi
on
Control-Voltage
15 15
E/bit (fJ)
3.4
0.3
15
2.4 (0.16)
OPEN Lab Prof. Sorger
E/bit Scaling: FET vs. EAM
300K = 0.004aJ
= 4aJ
= 4fJ
15fJ
3fJ
0.3J
Boltzmann Approximation (=Thermal Smearing)
Sorger Group EAMs (Time not to scale)
10’s aJ Cavity
Finesse = 10
Mode Material
HPP ITO
HPP Graphene Slot Gr or TMD
SIS TMD
Technology Gap With equal E/bit, Photonics performance over Electronics:
1) <10x speed (<100GHz EAM vs. <2GHz link) 2) <1-5x WDM (depending on link length)
Sorger Nanophot. (2012)
Khan, (in review)
Ma, IEEE STQE (2017)
Cavity design based on Liu, J. Nanoph.(2015)
•
Thermally Unattainable Without Cooling
Transistors by (Intel)
Sorger Group, J. Opt., special issue (submitted)
100x
OPEN Lab Prof. Sorger
HyPPI: Hybrid Plasmon Photonics Interconnect
Sun, Badaway, Narayana, El-Ghazawi, Sorger, IEEE Photonics, 7,6 (2015).
High BFD
Mid BFD
low BFD
Electrical
Plasmonic Photonic
HyPPI (this work)
* BFD: Bit Flow Density [Gbps/um2]
Physics & Material (E/bit)Device SNR @ Rx Desired BER
OPEN Lab Prof. Sorger
Photonic Reconfigurable Computing
2015 2016 2017
MorphoNoC
DEVICE
LINK
NETWORK
SYSTEM & APPLICATION
HyPPI BFD
MoDetector
Universal CLEAR
For P
eer Review
Fig. 2 Capability-to-Length-Energy-Area-Resistance (CLEAR) based performance-cost comparison
between electrical (red) and hybrid photon-plasmon (blue) on-chip interconnect links as a function of link
length and technology evolution time. The chip scale (CS = 1 cm) link length and current year (2016) are
denoted in red. The following models are deployed; a) A capacity-area model based on the number of
transistors and on-chip optical devices, which can be regarded as the original Moore’s Law model; b) An
energy efficiency model is derived based on Koomey’s law, which is bounded by the kBT ln(2) ≈ 2.75zJ/bit,
Landauer limit (kB is the Boltzmann constant; T is the temperature); c) A the economic resistance model
based on technology-experience models and at the year 2016, the electronic link cost less than one billionth
to one millionth of the cost of the hybrid link; and d) A model for parallelism (after year 2006) capturing
multi-core architecture and the limitation from ‘dark’ silicon concepts in electrical link interconnects. The
yellow data point represents the actual CMOS silicon photonic chip that IBM fabricated in 2015.
Page 5 of 5 Spectrum
Link-CLEAR
Dynamic-CHyPPI HyPPI NoC
D-CHyPPI based Camera Sensor
Noise HyPPI
5x5 Optical Router
Fundamental Scaling Law
G. Pomrenke F. Darema
El-Ghazawi & Sorger Groups
OPEN Lab Prof. Sorger
Optical on-chip FFT
Hillerkus OE (2011)
Cooley-Tukey Method Addition
Multiplication
Sorger Group, Frontiers in optics (2017)
OPEN Lab Prof. Sorger
Convolutional Neural Networks based-on Optical FTT
Sorger Group, IEEE Computing (2017)
OPEN Lab Prof. Sorger
Mirror Symmetry Density with Delay in Spiking Neural Networks
Sorger Group. (2017) submitted
OPEN Lab Prof. Sorger
Prof. Volker Sorger sorger.seas.gwu.edu [email protected]
OPEN Sorger Team
Post Docs Grad Students U-grads Collaborators
Dr. I. Sarpkaya Dr. Ke Liu Dr. Hasan Goktas Dr. Elnaz Akbari
Mohammad Tahersima Sikandar Khan Matt Zhizhen Shuai Sun Jonathan George Rubab Amin Hani Nejadriahi Seyed Haghshenas Rohit Hemnani
D. D’Hemmecout T. Weinshel J. Crandall
Prof. Majumdar (UW) Prof. Agarwal (UPenn) Prof. Kimerling (MIT) Prof. Reed (Stanford) Prof. Bartels (UCR) Prof. Lee (KU) Prof. Prucnal (Princeton ) Prof. El Ghazawi (GWU) Dr. Sadana (IBM Watson) Prof. Cesare (NTU)