Bie and colleagues (Nature Nanotechnology 12, 1124–1129, 2017) demonstrated that a single bilayer MoTe2_2 p–n junction transferred onto a silicon photonic-crystal waveguide can serve both as an electrically driven near-infrared light-emitting diode and as a waveguide-coupled photodetector, establishing two-dimensional transition-metal dichalcogenides (TMDs) as a candidate active material for back-end-of-line integration with silicon photonics. The demonstration left five linked gaps between the proof of concept and a practical on-chip optical transceiver: only about 5% of the electroluminescence was collected from the grating couplers, detector bandwidth was verified only to 200 MHz, external quantum efficiency was 0.5% and responsivity 4.8 mA W−1^{-1} at 1160 nm, electrically pumped lasing of a TMD gain medium in a high-Q photonic-crystal nanocavity was not shown, and the transfer-based fabrication had no demonstrated path to reproducible, CMOS-backend-compatible scaling. The 2017–2026 literature has discharged parts of this agenda — gigahertz-class waveguide-integrated MoTe2_2 detectors with responsivities up to hundreds of mA W−1^{-1}, and waveguide–emitter coupling efficiency up to 21.2% (with 67% system-level emission-to-extraction efficiency) — while the surviving open core remains: net-gain lasing under electrical carrier injection in a TMD integrated with a high-Q cavity on silicon, reproducible wafer-scale/CMOS-backend fabrication of the van der Waals stack, and a complete TMD-based on-chip link meeting practical link budgets.

Background

Silicon photonics routes optical signals through waveguides, interferometers and modulators fabricated directly on silicon chips, and is a leading architecture for overcoming the electrical interconnect bottleneck in high-performance computing. Its two persistent weak components are the light source and the photodetector: lasers are usually provided by one or a few off-chip or wafer-bonded III–V chips, while detectors in advanced silicon-photonic processes rely on additional germanium epitaxy, and both options complicate the vision of many low-power transceivers distributed at the electrical interface. Two-dimensional transition-metal dichalcogenides (TMDs, layered compounds such as MoTe2_2, MoS2_2 and WSe2_2) were proposed as an alternative active material class: because they are bonded in-plane by van der Waals forces, flakes can be transferred onto finished photonic circuits without epitaxy or lattice matching, an integration route compatible with back-end-of-line (BEOL) processing — the late, low-temperature fabrication steps performed after the transistor metal layers of a complementary metal-oxide-semiconductor (CMOS) wafer are complete. Bilayer MoTe2_2 has a near-infrared bandgap just below the silicon transparency window, making it suitable for silicon-photonic wavelengths around 1.1–1.3 µm. In the 2017 demonstration of Bie et al., an exfoliated bilayer MoTe2_2 flake encapsulated in hexagonal boron nitride (h-BN) was placed on a holey silicon photonic-crystal (PhC) waveguide — a periodically perforated silicon membrane guiding a transverse-electric-like mode — with split graphite gates electrostatically inducing p-type and n-type regions, forming a lateral p–n junction. Under forward bias the junction emitted electroluminescence near 1175 nm, a fraction of which coupled evanescently into the guided mode and was extracted by grating couplers (diffractive gratings converting between guided and free-space light); under illumination through a grating the same junction produced photocurrent. Key reported figures: diode ideality factor 2.3 (a measure of how closely the junction follows ideal diode behavior, where 1 is ideal), series resistance 0.25–0.6 MΩ, reverse saturation current about 10 pA; grating-extracted electroluminescence fraction γ≈5%\gamma\approx 5\%; detector responsivity (photocurrent per unit incident optical power) 4.8 mA W−1^{-1} and external quantum efficiency (EQE, the ratio of photogenerated carriers to incident photons) of 0.5% at 1160 nm; photocurrent response verified only to 200 MHz, the limit of the current amplifier, with drift-velocity estimates suggesting gigahertz capability. The authors themselves noted that the EQE and responsivity were modest compared with incumbent germanium photodetectors, that coupling losses and MoTe2_2 absorption stood between the demonstration and practical on-chip link budgets (the end-to-end accounting of optical power launched, coupled, guided and detected on a chip), that narrow-linewidth electrically pumped lasing would require integrating the electrically driven TMD gain medium with high-Q (high quality factor, i.e. long photon lifetime and correspondingly narrow resonance linewidth) photonic-crystal nanocavities and demonstrating net gain under electrical injection, and that the transfer-based, exfoliation-dependent fabrication had not been shown to scale with reproducible yields and industrial CMOS backend compatibility. Independent context for the lasing item: optically pumped TMD nanolasers (e.g. monolayer WSe2_2 on photonic-crystal cavities, reported from 2015 onward) exist, but a 2025 community assessment states that whether true lasing has been achieved in TMD devices at all remains heavily debated, and injection current density in van der Waals junctions remains far below what lasing thresholds are expected to require.

Problem Statement

The problem is: turn the 2017 proof-of-concept — one exfoliated bilayer MoTe2_2 p–n junction, transferred in a van der Waals stack onto a silicon photonic-crystal waveguide, acting as both a near-infrared light-emitting diode and a waveguide-coupled photodetector — into a practical building block for silicon photonic integrated circuits, one overarching objective with five jointly linked targets. (i) Route a substantially larger fraction of the electrically generated emission into the guided mode and out through couplers — the source estimated only γ≈5%\gamma\approx 5\% collected via its grating couplers — so that on-chip link budgets (the end-to-end accounting of optical power from source to detector) become practical. (ii) Demonstrate, rather than estimate, the achievable electrical detection bandwidth including contact and circuit parasitics, beyond the 200 MHz amplifier-limited verification of the source. (iii) Raise the detector's external quantum efficiency (0.5% at 1160 nm) and responsivity (photocurrent per unit incident optical power, 4.8 mA W−1^{-1}) to values competitive with incumbent photodetectors. (iv) Demonstrate net optical gain and lasing of a TMD gain medium under electrical carrier injection, integrated with a high-Q photonic-crystal nanocavity on the silicon photonic platform. (v) Establish a reproducible, high-yield fabrication route for this van der Waals stack that is compatible with industrial CMOS back-end-of-line processing. Later work (2017–2026) has already discharged the bandwidth uncertainty and much of the detector-efficiency gap, and has raised waveguide-coupling efficiency to 21.2% (with 67% system-level emission-to-extraction efficiency) by alternate architectures; the surviving open core that a solution must still deliver is: (a) a TMD laser showing net gain under electrical carrier injection — not AC-field-driven, optically pumped, or photoluminescence-based operation — integrated with a high-Q cavity on a silicon photonic circuit; (b) a scalable, CMOS-backend-compatible, reproducible fabrication flow for the transfer-based TMD stack; and (c) a complete on-chip emitter-to-detector link, based on such TMD components, whose measured coupling, speed, efficiency and dark current jointly satisfy practical on-chip link budgets. The three surviving-core components are jointly required: the objective is resolved only when all three are established. The lasing claim (iv)/(a) is net optical gain and lasing of a TMD gain medium under electrical carrier injection; forward-bias electroluminescence alone does not establish lasing. 'Practical on-chip link budgets' and 'CMOS-backend-compatible' are qualitative criteria with no fixed numeric value.

References

  1. Ya-Qing Bie, Gabriele Grosso, Mikkel Heuck, Marco M. Furchi, Yuan Cao, Jiabao Zheng, Darius Bunandar, Efren Navarro-Moratalla, Lin Zhou, Dmitri K. Efetov, Takashi Taniguchi, Kenji Watanabe, Jing Kong, Dirk Englund, Pablo Jarillo-Herrero, 'A MoTe2-based light-emitting diode and photodetector for silicon photonic integrated circuits', Nature Nanotechnology 12 (12), 1124–1129 (2017), DOI 10.1038/nnano.2017.209, https://doi.org/10.1038/nnano.2017.209.
  2. Yanhao Tang, Kin Fai Mak, '2D materials for silicon photonics', Nature Nanotechnology 12 (12), 1121–1122 (2017), DOI 10.1038/nnano.2017.230, https://doi.org/10.1038/nnano.2017.230.
  3. Chen Li, Ruijuan Tian, Ruixuan Yi, Siqi Hu, Yuxin Chen, Qingchen Yuan, Xutao Zhang, Yan Liu, Yue Hao, Xuetao Gan, Jianlin Zhao, 'MoTe2 PN Homojunction Constructed on a Silicon Photonic Crystal Cavity for High-Performance Photodetector', ACS Photonics 8 (8), 2431–2439 (2021), DOI 10.1021/acsphotonics.1c00628, https://doi.org/10.1021/acsphotonics.1c00628.
  4. Hao Wang, Behrouz Movahhed Nouri, Hamed Dalir, Volker J. Sorger, '30 GHz plasmonic slot MoTe2 photodetector integrated with silicon photonic circuits at telecom wavelength', Proc. SPIE 12486 (Ultrafast Phenomena and Nanophotonics XXVII), paper 38 (2023), DOI 10.1117/12.2651013, https://doi.org/10.1117/12.2651013.
  5. Changming Yang, Zeyi Liu, Hongjun Cai, Dehui Li, Yu Yu, Xinliang Zhang, 'High-Speed and High-Responsivity Vertical van der Waals Heterostructure Waveguide Photodetector Operating in Telecom Band', ACS Nano 19 (9), 8661–8666 (2025), DOI 10.1021/acsnano.4c14937, https://doi.org/10.1021/acsnano.4c14937.
  6. Zheng-Zhe Chen, Hsiang-Ting Lin, Chiao-Yun Chang, Muhammad Adil, Po-Cheng Tsai, Tsung Sheng Kao, Chi Chen, Shih-Yen Lin, Min-Hsiung Shih, 'Electrically-Driven 2D Semiconductor Microcavity Laser', Advanced Materials 37 (42) (2025), DOI 10.1002/adma.202509861, https://doi.org/10.1002/adma.202509861.
  7. Teresa López-Carrasco, Marcos H. D. Guimarães, 'Atomically thin transition metal dichalcogenide nanolasers: Challenges and opportunities', Applied Physics Letters 126 (22) (2025), DOI 10.1063/5.0267478, https://doi.org/10.1063/5.0267478.
  8. Chen Li, Yongzhuo Li, Yutong Zhong, Yuqian Tang, Cun-Zheng Ning, '2D Material light–emitting transistor with a dynamically controllable emission location for optimized waveguide coupling on silicon', Science Advances 12 (14) (2026), DOI 10.1126/sciadv.aeb8783, https://doi.org/10.1126/sciadv.aeb8783.
  9. Rei Usami, Koshi Oi, Keisuke Yamada, Jiang Pu, Hao Ou, Takahiko Endo, Yasumitsu Miyata, Taishi Takenobu, 'Stepwise Engineering of Van der Waals Heterostructures for High Current Density in Light Emitting Devices', Advanced Electronic Materials 12 (16) (2026), DOI 10.1002/aelm.202400930, https://doi.org/10.1002/aelm.202400930.
  10. Wei Huang, Jianbi Le, Rong Chen, Zhixian Lin, Xiongtu Zhou, Yongai Zhang, Chaoxing Wu, 'Research Progress and Applications of Non-Carrier-Injection Electroluminescence', Advanced Electronic Materials 12 (10) (2026), DOI 10.1002/aelm.202500889, https://doi.org/10.1002/aelm.202500889.
  11. Jinlong Lu, Fang Yang, Jian Feng, Xing Zhang, Lunchun Guo, Shuangchen Ruan, Hui Li, Bingxi Xiang, 'Direct Integration of MoTe2 Onto Thin-Film Lithium Niobate Grating Couplers for On-Chip Photodetection', Advanced Optical Materials 13 (31) (2025), DOI 10.1002/adom.202502148, https://doi.org/10.1002/adom.202502148.
  12. Zikang Li, Xiukai Chen, Dehao Kong, Guodong Sun, Shuyan Tian, Yinuo Sun, Chengyue Xiong, Xiaoguo Song, Jicai Feng, Hong Bian, 'Two-Dimensional Molybdenum Ditelluride (MoTe2): Synthetic Strategies, Physical Properties, and Future Perspectives', Small 22 (35) (2026), DOI 10.1002/smll.73709, https://doi.org/10.1002/smll.73709.