Can graphene on silicon chips detect 2 μm light fast and well?
A thin silicon waveguide topped by graphene and a wide metal strip absorbs light mostly in the graphene, giving fast chip-based detectors beyond the usual telecom band.
Source
High-performance silicon-graphene hybrid plasmonic waveguide photodetectors beyond 1.55 μm
Study at a glance
- Design
- Other — Mode-solver design of a thin-silicon hybrid plasmonic waveguide, then fabrication and measurement of photocurrent maps, responsivity and frequency response for three devices on one chip.
- N
- Device experiment; results reported for three devices (A, B, C) on the same chip, not a sample.
- Population
- Monolayer-graphene photodetectors on silicon-on-insulator hybrid plasmonic waveguides operating at 2 and 1.55 micrometres
- Outcome
- Graphene absorption fraction, photoresponse mechanism versus gate and bias voltage, responsivity, and 3 dB bandwidth
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The authors simulated how silicon ridge width, silicon thickness and metal-strip size change how much guided light is absorbed by graphene versus wasted in the metal, and chose a thin, wide silicon ridge. They fabricated detectors with a top metal signal electrode and metal-graphene-metal ground contacts, mapped photocurrent against gate and bias voltage, and measured responsivity and frequency response at 2 μm, plus a companion device at 1.55 μm.
What they found
Thinning the silicon core roughly doubled absorption, and the design put about 70% of absorbed light into graphene. At zero bias the photocurrent flipped sign with gate voltage, the signature of the photothermoelectric effect; with bias, bolometric or photoconductive effects dominated depending on doping. At 2 μm a device reached about 70 mA/W with bandwidth above 20 GHz, and at 1.55 μm a device reached 396 mA/W with bandwidth above 40 GHz and received 30 Gbit/s data.
The limits
What it doesn't show
Both bandwidth figures are limited by the lab equipment, so the true speed is unknown. The metal-graphene-metal design has a high dark current, so signal-to-dark ratio and noise-equivalent power are modest, and graphene mobility was low, probably due to fabrication defects. Only three devices on one chip were characterised, with some random variation in contact resistance, so reproducibility is not established. Responsivity at 2 μm remains well below the 1.55 μm device.
Key terms
- Hybrid plasmonic waveguide
- A waveguide whose mode is partly a dielectric (photonic) mode and partly a surface-plasmon mode on metal, concentrating light near an interface.
- Photothermoelectric effect
- Light heats electrons, and a temperature difference across regions with different Seebeck coefficients drives a current without bias.
- Bolometric effect
- Light heating changes the material's conductance, so under bias the current changes; in highly doped graphene this gives a current opposite in sign to the bias.
- Photoconductive effect
- Extra photogenerated carriers increase conductivity, so the photocurrent has the same sign as the bias.
- 3 dB bandwidth
- The modulation frequency at which the detector's output falls to half power; a measure of speed.
Flashcards
0 of 9 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
What mechanism dominates photocurrent at zero bias?
Common questions
Why make the silicon so thin?
A thinner core confines light less vertically, so more of the evanescent field reaches the graphene and absorption coefficients more than doubled.
How can you tell which mechanism generates the photocurrent?
From signs: a zero-bias current that flips with gate is photothermoelectric; with bias, a current opposite to the bias indicates bolometric and one matching the bias indicates photoconductive.
Why not just make the metal strip very narrow to cut metal loss?
A narrow strip reduces graphene absorption and raises contact resistance, hurting responsivity and bandwidth, so a medium width was chosen.
More on Plasmonics