Technical Paper Analysis | One Modulator, 800 nm of Spectrum: TFLN Pushes from the O-Band All the Way to 2 μm

2 days ago

7 min read

Optical communications is running out of spectrum, so the industry is looking for new room beyond the traditional C-band — and pushing the long-unused 2-μm window into the spotlight. The problem: the fiber exists (hollow-core fiber), the amplifiers exist, the PDs exist — but a high-speed modulator that can span such a wide spectrum has been missing. This work from Huazhong University of Science and Technology, Fudan University and the Institute of Semiconductors, Chinese Academy of Sciences, published in Nature Communications, demonstrates a thin-film lithium niobate (TFLN) modulator in which a single device covers 1260–2060 nm, a continuous 800 nm operating bandwidth, taking in every telecom band — O, E, S, C, L, U — plus 2 μm, with EO bandwidth of about 100 GHz across the O–U bands and still 50 GHz at 2 μm. This isn't another baud-rate paper; it fills in the last missing transmitter piece of the "ultra-broadband optical network" puzzle.

1. Paper background: who did it, and why it's worth reading

The paper, titled "Ultra-broadband near- to mid-infrared electro-optic modulator on thin-film lithium niobate," was accepted by Nature Communications in December 2025 (2026, 17:1138). The authors are from the Wuhan National Laboratory for Optoelectronics at Huazhong University of Science and Technology, Fudan University, and the Institute of Semiconductors, Chinese Academy of Sciences.

What's worth reading isn't the "TFLN modulator" itself — TFLN has long been the hot platform for high-speed modulation. It's the system-level gap it closes: while the whole industry talks about "expanding spectrum toward 2 μm," three pieces are already in place — the fiber (hollow-core photonic bandgap fiber, HC-PBGF, has shown a low-loss window from 1240 to 1940 nm), the amplifiers (from rare-earth doping to integrated optical parametric amplifiers), and the receiver (2-μm PD bandwidth has passed 40 GHz) — yet the broadband modulator on the transmit side has been the bottleneck. Silicon and germanium modulators rely on the free-carrier effect, and at long wavelengths they are dragged down by waveguide dispersion and carrier-velocity mismatch, so they can only serve a single band. This paper fills that hole.

The fiber, amplifiers and PDs are all ready; the only thing missing is a modulator that can span such a wide spectrum.

***System concept of a single full-spectrum TFLN modulator bridging multi-wavelength lasers, multi-band amplifiers and broadband fiber, with micrograph and SEM images of the actual device***

2. The core question, in one sentence

How do you make a single modulator keep both "wide enough optical bandwidth" and "high enough EO bandwidth" across an 800 nm wavelength range?

The difficulty is that these two goals fight each other. For broadband optics, the passive components (splitters, mode converters) must work properly at every wavelength; for high EO bandwidth, the microwave and optical velocities and impedances must match across the whole band. Traditional designs gain one at the expense of the other, which is why — although TFLN had been demonstrated separately from the visible out to 2 μm — no one had "seamlessly" connected O–U plus 2 μm.

3. Key figures, one by one

3.1 The passive components are the real broadband gate

This figure shows the core design idea of the whole paper: broadband isn't achieved by forcing it through the modulation section, but by using adiabatic mode evolution in the passive components to "gently" carry light across the whole spectrum.

The edge coupler is a spot-size converter (SSC): a double-layer LN taper plus a silicon oxynitride (SiON) ridge waveguide, butt-coupled to ultra-high-NA fiber (UHNA4). The key is that SiON's refractive index (n≈1.54) sits right between LN (n≈2.2) and the buried-oxide SiO2 (n≈1.45), so light stays locked in the waveguide without leaking. The SiON ridge width was ultimately optimized to 4.4 µm, achieving <1 dB coupling loss from the O- to the U-band. The 3-dB splitter uses an adiabatic TFLN waveguide with three taper sections to split light evenly into two paths, again covering the entire broadband range.

3.2 The velocity-matching trade-off: aligned to the C-band, with 2 μm deliberately giving ground

The traveling-wave electrode uses a push-pull coplanar waveguide (CPW) design: 25 µm signal electrode, 150 µm ground, 1 µm-tall gold electrodes, and a characteristic impedance of about 42 Ω. A 38 Ω load resistor is deliberately used to suppress the low-frequency roll-off of S21 in exchange for bandwidth.

Here is a very "engineer" decision: the optical group index at 1310/1550/2000 nm is 2.18/2.13/2.04, and they set the RF effective index to 2.13, aligned to the C-band. The cost is that at 2 μm, residual index mismatch limits the theoretical 3-dB bandwidth to about 80 GHz. In other words, this device's velocity matching is centered on the C-band; 2 μm is taken care of, but it isn't the priority.

3.3 The measured numbers: insertion loss, coupling loss, Vπ·L

This figure shows the measured loss distribution of the device. TE-mode coupling loss is 0.78 dB/facet at 1310 nm, only 0.69 dB/facet at 1550 nm, rising to 2.21 dB/facet at 2000 nm. The loss peaks near 1510 nm and 2000 nm come from infrared absorption by N-H bonds in SiON — an intrinsic material issue that points to the limits of pushing the SiON platform to even longer wavelengths.

On-chip insertion loss (IL) is estimated at 1.2 dB at 1310 nm, 2.8 dB at 1550 nm and 5.8 dB at 1970 nm. The half-wave voltage-length product Vπ·L is 1.92 / 2.48 / 2.61 / 2.74 / 3.94 V·cm (at 1310 / 1485 / 1550 / 1590 / 2000 nm), staying throughout at a level CMOS can drive.

***Measured coupling loss of the spot-size converter from 1260 to 2060 nm: lowest in the C-band, with N-H absorption causing loss peaks near 1510/2000 nm***

3.4 Pushing the baud rate to the limit: >240 Gbps per lane

This figure shows the modulator's final results. Below the HD-FEC threshold (3.8×10⁻³), PAM-4 signals reached 260 / 260 / 260 / 280 / 280 / 240 / 170 Gbps in the seven bands O/E/S/C/L/U/2 μm respectively; OOK exceeded 170 Gbps across all of O–U and reached 150 Gbps at 2 μm. Among these, the 170 Gbps PAM-4 at 2 μm is the current single-lane record for that band.

***Eye diagrams and data rates of a single TFLN modulator across seven bands from O to 2 μm: single-lane PAM-4 above 240 Gbps across O–U and 170 Gbps at 2 μm***

4. Technical highlights: the two real differentiators

First, the design philosophy shifts from the modulation section to the passive components. Most broadband modulator work focuses on the electrodes and modulation section; this paper does the opposite, staking broadband success on the adiabatic design of the SSC and 3-dB splitter. It's a replicable methodology: if you want broadband, first make the passive components adiabatic across the full spectrum, then match the electrodes to them.

Second, a generational leap in 2-μm bandwidth. 50 GHz of EO bandwidth is currently the highest at 2 μm — a 2.3x bandwidth improvement over existing 2-μm modulators. This isn't progress after the decimal point; it moves 2 μm straight from "can communicate" to "can communicate at high speed."

5. Industry link: how far from volume production, and who benefits

Start with the sober part. This is a single, lab-grade device demonstration, measured on a benchtop with UHNA4 fiber and index-matching oil coupling, an external load resistor, and a tunable laser plus ASE source — still some distance from a "packageable, manufacturable commercial module." The SiON N-H absorption peaks and the higher coupling and insertion losses at 2 μm are tough problems to solve before volume production.

But the direction is right. Bandwidth demand across AI infrastructure and data centers is forcing optical communications to "find capacity in new spectrum," and 2 μm combined with hollow-core fiber is seen as the next potential window. This paper shows that TFLN has a shot at being that single transmitter platform that "covers every band" — for module makers and system vendors who want one production line and one material to cover many bands, this "universal transmitter" vision is very attractive. The beneficiaries would be the TFLN wafer and device supply chain (the paper uses the NanoLN platform) and players betting on the 2-μm / hollow-core fiber route.

That said, the real-world hurdles in taking TFLN from paper to production line are as tough as the device itself. We fully broke down TFLN's opportunity in optical communications and CPO, and the three bottlenecks standing before volume production, in TFLN Breaks Another Record — But Don't Misread the Signal; for the timeline of scaling up and how players are positioning, see OFC 2026: The TFLN Industry Inflection Point Arrives.

Worth noting: real commercialization of 2 μm still depends on the cost and maturity of hollow-core fiber and 2-μm amplifiers (TDFA, etc.). The modulator completing the puzzle doesn't mean the whole road is open.

6. Conclusion

This paper's historical place isn't "TFLN broke another baud-rate record," but rather the first time a single device has stitched the O–U telecom bands and 2 μm into one continuous 800 nm operating spectrum. It makes the "ultra-broadband optical network" — a concept that until now existed only on roadmaps — concrete on the transmitter side.

If you're tracking the next decade of optical communications, this device sends a clear signal: 2 μm is no longer an academic term but a candidate band with a complete transmit-receive chain starting to form; and TFLN is growing from "a high-speed C-band material" into "a full-spectrum general-purpose platform." What to watch isn't when this device reaches volume production, but how the TFLN supply chain and the hollow-core fiber camp respond next.

References

Qiyuan Li, Qiyuan Yi, Aolong Sun, et al. "Ultra-broadband near- to mid-infrared electro-optic modulator on thin-film lithium niobate." Nature Communications, 2026, 17:1138. DOI: https://doi.org/10.1038/s41467-025-67902-2
Affiliations: Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology; Key Laboratory for Information Science of Electromagnetic Waves (MoE), Fudan University; Institute of Semiconductors, Chinese Academy of Sciences
Platform: NanoLN 300-nm x-cut thin-film lithium niobate