Graphene-Based Integrated Photonics For Next-Generation Datacom And Telecom

M. Romagnoli; V. Sorianello; M. Midrio; F. H. L. Koppens; C. Huyghebaert; D. Neumaier; P. Galli; W. Templ; A. D'Errico; A. C. Ferrari

Source-linked AI summary

Graphene-Based Integrated Photonics For Next-Generation Datacom And Telecom

M. Romagnoli, V. Sorianello, M. Midrio, F. H. L. Koppens, C. Huyghebaert, D. Neumaier, P. Galli, W. Templ, A. D'Errico, A. C. Ferrari

arXiv:1906.00353v1Jun 2, 2019physics.app-ph

TL;DR

The paper addresses how integrated photonics can meet rising datacom and telecom demands while improving power consumption, manufacturability, and wafer-scale integration. It reviews graphene-based modulators, detectors, switches, and transceivers against established technologies, reporting compact, low-energy operation and identifying integration constraints. The authors conclude that graphene offers an evolutionary integration pathway whose competitiveness depends strongly on carrier mobility and process maturity.

Problem
Rising datacom and telecom bandwidth requirements motivate integrated-photonics technologies that improve power consumption, manufacturability, and wafer-scale integration.
Method
The paper analyzes graphene photonic modulators, detectors, and switches, compares them with established technologies, and discusses integration and manufacturing strategies.
Results
Graphene modulation is reported at 0.1 pJ bit–1 with phase-shifter length <0.5 mm and FOMPM ~0.1 V dB, while detectors provide at least ~0.5 A W−1 bolometric responsivity or 10 V W−1 photothermal responsivity.
Takeaways & Limitations
Graphene photonics offers an evolutionary pathway in which active functions are applied to passive waveguides, with competitive performance expected at mobility >10,000 cm2 V−1 s−1 and carrier concentration ~10^12 cm−2.
Takeaways & Limitations
Wafer-scale adoption remains limited by the lack of a standardized SLG–CMOS-compatible contacting scheme and by direct-growth conditions involving processed wafers and catalytic metal.

Abstract

from arXiv · show

Graphene is an ideal material for optoelectronic applications. Its photonic properties give several advantages and complementarities over Si photonics. For example, graphene enables both electro-absorption and electro-refraction modulation with an electro-optical index change exceeding 10$^{-3}$. It can be used for optical add-drop multiplexing with voltage control, eliminating the current dissipation used for the thermal detuning of microresonators, and for thermoelectric-based ultrafast optical detectors that generate a voltage without transimpedance amplifiers. Here, we present our vision for grapheme-based integrated photonics. We review graphene-based transceivers and compare them with existing technologies. Strategies for improving power consumption, manufacturability and wafer-scale integration are addressed. We outline a roadmap of the technological requirements to meet the demands of the datacom and telecom markets. We show that graphene based integrated photonics could enable ultrahigh spatial bandwidth density , low power consumption for board connectivity and connectivity between data centres, access networks and metropolitan, core, regional and long-haul optical communications.

Graphene-based modulators

Graphene modulators use carrier-controlled changes in absorption and refractive index to provide compact electro-absorption and electro-refractive operation. Double-SLG designs can improve modulation efficiency and reduce silicon-related loss, while remaining subject to material, voltage, and competing-technology constraints.

2.3% absorption at normal incidence increases to thousands of dB cm−1 when graphene is integrated over a silicon waveguide.Shifting EF reduces or cancels absorption through Pauli blocking, while defects and scattering introduce background loss.
Carrier accumulation in SLG–insulator–SLG capacitors shifts EF, with voltage determined by dielectric and quantum-capacitance contributions.The double-SLG configuration places two graphene layers on an undoped silicon waveguide.
At 1,550 nm, interband absorption occurs below |EF| = Eph/2, whereas Pauli blocking suppresses it above that threshold; shorter τ increases residual absorption.For single and double SLG, interband absorption can reach 0.1 and 0.2 dB μm−1, respectively, when |EF| < Eph/2.
2 × 10−3 effective-index modulation produces ~8.1 rad mm−1 phase modulation, exceeding comparable silicon modulation by more than tenfold.This operation uses |EF| > Eph/2, where absorption is small.
Double-SLG devices provide the largest and steepest Δneff variation, exceeding other technologies over the 2.5–4 V range and enabling short phase-shifter sections above EF ~0.45 eV.The transparency region above approximately 5 V allows steep voltage-dependent index modulation.
SLG phase-modulator FOMPM is about ten times higher than silicon's at high mobility, with VπL below 2.8 V mm for single SLG and 1.6 V mm for double SLG.A double-SLG device embedded in the waveguide core can reach VπL < 0.7 V mm.
Double-SLG modulators avoid silicon doping and can reduce loss, whereas silicon-insulator-silicon and depleted p–n designs incur approximately 5 dB mm−1 and 0.55 dB mm−1, respectively.The double-SLG configuration can maintain small αloss because silicon doping is not required.
Theoretical double-SLG performance reaches VπL ~1.6 V mm, FOMPM < 2 V dB at τ = 100 fs, and FOMPM < 1 V dB at τ = 300 fs.These estimates are comparable to or better than several established technologies, although InGaAsP membranes on silicon can achieve similar performance.

Graphene-based switching

Graphene-based switching uses voltage-controlled absorption in micro-ring resonators to route wavelengths directly between DROP and THROUGH ports. This approach targets lower-latency, lower-power optical switching for aggregated data streams.

Optical switching can avoid opto-electro-optical conversion and packet handling, addressing latency and energy costs when large data streams are aggregated.
Graphene switches can handle aggregated 100 GBE and beyond data streams and be operated remotely on demand through interconnected data centres.
In the ON state, low ring absorption enables coherent interference and DROP transmission, whereas high absorption directs light to the THROUGH port.
Graphene electro-absorption varies micro-ring losses from ~1,000 dB cm−1 to <10 dB cm−1, replacing resonance tuning as the switching mechanism.
The four-port micro-ring filter has a 1.2 THz free spectral range and 20 GHz bandwidth, with τ ~300 fs achieving isolation from adjacent channels.

Graphene photodetectors

Graphene photodetectors offer broadband, fast optical-to-electrical conversion through several mechanisms, including photothermoelectric generation. Waveguide integration improves absorption, while device designs target higher responsivity and reduced dark current.

Graphene’s zerogap band enables frequency-independent absorption from the UV to far-IR region, with fast photo-switching reported up to ~270 GHz.
Photothermoelectric detection converts photon absorption and ultrafast carrier scattering into a local voltage through the Seebeck effect rather than a current.
Evanescent coupling in waveguide-integrated detectors raises absorption above 2.3% and, with increased interaction length, can approach 100%.
A speed-optimized CVD graphene detector reported ~50 Gb s−1 operation, while a separate device achieved Rph~0.36 A W−1 and a 40 GHz 3 dB bandwidth.
A double-gated thermoelectric design is predicted to achieve Rph > 0.8 A W−1 or voltage responsivity Rph > 100 V W−1 under realistic parameters.
Reverse-biased graphene–Si Schottky detection produced Rph~85 mA W−1 with Idark~20 nA at 1 V, improving responsivity over standard metal–Si detection.

Wafer-scale integration

Wafer-scale graphene photonics requires CMOS-compatible contacts, controlled fabrication, and integration schemes that preserve device performance while enabling heterogeneous packaging.

Laboratory-scale metal lift-off is unsuitable for very large-scale integration because it causes redeposition, metal-edge ears, and partial metal retention.
Cu damascene processing patterns dielectric trenches or vias, deposits overfilling metal, and removes excess metal by chemical mechanical polishing.
35°C?

Conclusions and outlook

The paper presents graphene-integrated photonics as an evolutionary platform for telecom and datacom, combining graphene-based active functions with passive waveguides. It reports low-power, compact modulation and simplified detection, while identifying mobility, contact resistance, and integration maturity as practical requirements.

Conclusions and outlook: Graphene-based active functionalities can be post-processed onto passive waveguides, separating the guiding circuit from modulation, detection, and switching.This avoids the full-integration approach strongly tied to CMOS processing in Si photonics.
Conclusions and outlook: 0.1 pJ bit–1 energy consumption, phase-shifter length <0.5 mm, and FOMPM ~0.1 V dB are reported for graphene modulation.The same conclusion reports electro-absorption length <0.1 mm, FOMEA ~3 dB, and drive voltage ≤1 V.
Conclusions and outlook: At least ~0.5 A W−1 responsivity is reported for graphene bolometric detectors, while photothermal detectors provide at least ~0.4 A W−1 or 10 V W−1 with negligible dark current.Photothermal voltage operation can avoid the transimpedance amplifier required for conventional current detection and simplify receiver design.
Conclusions and outlook: Graphene switches use voltage-driven charge accumulation in an SLG–insulator–SLG capacitor to enable or disable output ports.This contrasts with conventional Si thermo-optic switches operated under continuous electrical current flow.
Conclusions and outlook: Carrier mobility >10,000 cm2 V−1 s−1 and carrier concentration ~10^12 cm−2 are identified as conditions for competitive modulation, detection, and switching.The paper also highlights single crystals, optimized transfer or encapsulation, and minimized contact resistance as development needs.

Appendix 1 | Basic concepts of optical modulation.

Optical communication encodes electrical information in light by modulating amplitude or phase, then uses a photodetector to recover the signal. Datacom and telecom differ mainly in link length and corresponding priorities.

Appendix 1 | Basic concepts of optical modulation.: Digital communication sends messages as sequences of bits by modulating a source into electrical or optical signals.The communication goal is transferring a message between points, with unpredictability determining whether transmission conveys new information.
Appendix 1 | Basic concepts of optical modulation.: Integrated photonics modulates amplitude or phase through the electro-optical waveguide material or a single-layer graphene layer placed on the waveguide core.A receiver photodetector discriminates the encoded signal against channel noise and converts the optical signal into an electrical signal.
Appendix 1 | Basic concepts of optical modulation.: Datacom covers data-centre links of approximately 2 km, whereas telecom spans tens of kilometres to transoceanic distances.Shorter datacom links can tolerate smaller extinction ratios when reducing size, insertion loss, and power consumption is prioritized.

Appendix 2 | Modulators and detectors for optical transceivers

Optical transceivers combine modulators, detectors, and link-specific specifications. Their key evaluation criteria include energy, bandwidth, insertion loss, phase-shifter efficiency, responsivity, and dark current.

Appendix 2 | Modulators and detectors for optical transceivers: InP electro-absorption modulators serve mainly links up to ~80 km, Si MZI modulators serve pluggable interconnections below 10 km, and LiNbO3 MZIs serve links above 100 km.These platforms represent established modulator technologies for different optical-link ranges.
Appendix 2 | Modulators and detectors for optical transceivers: VπL measures the voltage–length product needed for a π phase shift, while FOMPM = VπLαloss combines phase-shifting efficiency with component loss.VπL is expressed in dB cm in the passage, and αloss is the loss introduced by the phase-shifting section.
Appendix 2 | Modulators and detectors for optical transceivers: InP photodetectors are commercially available and widely adopted, while Ge photodetectors are epitaxially grown on Si waveguides.Ge devices can receive waveguide light through edge coupling or evanescent coupling.
Appendix 2 | Modulators and detectors for optical transceivers: Responsivity is electrical output per optical input, bandwidth is limited by circuit design, and dark current is leakage in the absence of illumination.The passage gives responsivity units of A W−1 or V W−1 for waveguide photodetectors, excluding input coupling losses.
Appendix 2 | Modulators and detectors for optical transceivers: 100GE CLR4, LR4, and ER4 specify different interconnection lengths, while LR4 and ER4 use four 25 Gb s−1 optical channels over 1,260–1,360 nm.The transceiver specifications are presented as optical interface requirements for these Ethernet standards.
Appendix 2 | Modulators and detectors for optical transceivers: For a selected optical link, the figure of merit determines transmitter specifications such as laser power and modulator driving voltage.The passage illustrates this with a 100GE ER4 link carrying 100 Gb s−1 over 40 km of optical fibre.

Appendix 3 | the quality of graphene required for integrated photonic devices

Integrated photonic graphene quality is primarily tied to maximizing carrier mobility, which is governed by scattering time and affected by fabrication and material engineering. The relevant mobility range is typical for large-area CVD-grown and wet-transferred graphene, while phonon scattering imposes an upper bound and high-doping estimates require validation.

Integrated photonic devices require graphene carrier mobility μ to be maximized.The paper distinguishes photonic-device quality from graphene quality requirements for other applications.
Scattering time τ is limited by phonon, short-range, and charge-impurity scattering contributions.The inverse scattering time is modeled as the sum of inverse contributions from these mechanisms and others.
τ values of 10 fs, 100 fs, and 300 fs correspond to μ values of approximately 220, 2,200, and 6,600 cm2 V−1 s−1 at EF = 0.4 eV.
These mobility values are typical for large-area graphene grown by CVD and wet transferred.
Phonon scattering sets an upper bound on τ and therefore on mobility, while high-doping estimates need experimental validation.Substrate engineering and encapsulation may change phonon scattering and increase mobility at a given doping.

Appendix 4 | Mechanisms of optical modulation

Optical modulation can vary either absorption or refractive index, and graphene supports both mechanisms through Fermi-energy-dependent electronic transitions. This enables electro-absorption and phase modulation, with phase modulation convertible to amplitude modulation using an MZI.

Electro-absorption modulation varies optical absorption, whereas electro-refractive modulation varies refractive index and therefore optical phase.
A Mach–Zehnder interferometer converts phase modulation into amplitude modulation.
In silicon, plasma dispersion changes absorption and refractive index through carrier concentration, but carrier-induced absorption can produce loss.
Electrical fields can modulate semiconductor absorption through the Franz–Keldysh effect or quantum-confined Stark effect.In GeSi integrated on silicon waveguides, Franz–Keldysh modulation has reached rates up to 100 Gb s−1.
In graphene, changing EF controls absorption through Pauli blocking and phase through EF-dependent refractive index.When EF exceeds half the photon energy, interband excitation is inhibited and graphene becomes transparent; longer τ lowers intraband absorption.

Appendix 5 | The evolution of mobile communications: the 5G network

The 5G evolution is framed as a network-wide shift involving fixed and wireless links, multiple network segments, scalable connectivity, and increasingly customized services. Its applications span mobile cloud, dedicated premises networks, smart infrastructure, and industrial systems.

5G concepts are considered across telecom and datacom network segments rather than only at the component level.
Telecommunication networks comprise access, aggregation, and core segments with different cost, performance, quality, and device-count requirements.
5G combines wireless access with fixed copper-wire or fibre links, including connections among cell sites and network infrastructure.
Virtual access nodes provide dedicated cells for industrial or private applications with customized connectivity, mobility, and policy requirements.
5G wireless access supplements large high-power towers with more numerous small-cell stations to improve efficiency.
Mobile wireless cloud networks target Smart City, Smart Industry, and Smart Life applications, while dedicated premises networks combine optical interconnections with pervasive wireless networks.
Dedicated end-user premises networks are described as providing average capacity of 1Gbit/s per user, compared with 100Mbit/s per user in mobile wireless cloud networks.
The 5G ecosystem supports customized services and connected applications involving media, sensors, vehicles, smart grids, manufacturing, robotics, and remote-controlled machinery.