Integrated Terahertz Photonic Receiving Frontend with Link Noise Outperforming Electronics

Technical Summary: Integrated Terahertz Photonic Receiving Frontend with Link Noise Outperforming Electronics

Problem Statement
The realization of sixth-generation (6G) wireless networks requires terahertz (THz) technology to support terabit data rates and millimeter-level sensing. However, the application of THz waves is severely constrained by free-space path loss and atmospheric absorption, resulting in ultra-weak signals at the receiver. Current electronic frontends face fundamental limitations in this regime: as operating frequencies approach the transistor cut-off frequency (300–600 GHz), transistor noise performance degrades significantly (noise figures can reach up to 40 dB), and on-chip antenna gains are limited (typically <5 dBi) due to parasitic effects and device quality factors. While photonic technology offers a pathway to circumvent transistor bandwidth limits, existing microwave photonic links often exhibit noise performance inferior to state-of-the-art electronics. Furthermore, previous attempts at integrated THz photonic frontends have been limited to frequencies below 100 GHz or suffer from low THz-to-optic conversion efficiencies due to lumped antenna designs and insufficient antenna gain.

Methodology
The authors propose a monolithic-integrated THz photonic frontend based on a thin-film lithium niobate (TFLN) platform. The core innovation is a distributed-antenna-driven architecture that co-designs a high-gain on-chip THz antenna array with a broadband electro-optic (EO) modulator.

• Device Architecture: The chip consists of an optical Mach–Zehnder interferometer (MZI) coupled with a coplanar waveguide (CPW) electrode and a high-gain dipole antenna array. The antenna array captures free-space THz signals and converts them into guided waves propagating along the CPW electrodes via periodic slots.
• Distributed Interaction: Unlike traditional lumped designs where antennas and modulators are separate, this architecture allows the received THz signals to interact continuously and coherently with the optical carrier across the entire device length. The dipole antennas are arranged to excite the differential mode of the CPW electrode, enabling push-pull modulation with minimal phase skew.
• Material Platform: The use of TFLN leverages the ultrafast Pockels effect and enhanced EO interaction strength. The design utilizes different substrates for different bands: silicon for the D-band (140 GHz) and quartz for the J-band (250 GHz) and Y-band (450 GHz) to minimize THz attenuation and maximize bandwidth.
• Detection Scheme: The modulated optical signals are heterodyned with a local oscillator (LO) laser in a balanced photodetector (BPD) to recover the intermediate frequency (IF) signals, operating in a shot-noise-limited regime.

Key Contributions

1. First Demonstration of Superior Link Noise: The work presents the first integrated THz photonic frontend that achieves link noise performance surpassing state-of-the-art electronic frontends in the J-band (250 GHz) and Y-band (450 GHz).
2. Frequency-Scalable High Efficiency: By employing the distributed driving architecture, the authors achieved THz-to-optic conversion figures of merit (FOM) of 0.72, 0.54, and 0.52 cm²/W for the D-, J-, and Y-bands, respectively. These values represent improvements of over two orders of magnitude compared to previous records in the upper two windows.
3. Universal Architecture: The design demonstrates a simplified, universal architecture that maintains undegraded performance across a 1.7-octave frequency span (140–450 GHz), overcoming the typical frequency-dependent degradation seen in electronic and traditional photonic schemes.

Experimental Results

• Conversion Efficiency: The chips achieved carrier-to-sideband conversion ratios (CSRs) of -37.6 dB, -38.5 dB, and -38.8 dB at 134, 250, and 450 GHz, respectively, under a normalized free-space power density of 10 W/m².
• Noise Performance: The system exhibited effective isotropic noise figures (EINF) of 13.6 dB at 250 GHz and 16.2 dB at 450 GHz. These values are lower than those of state-of-the-art integrated electronics in the respective bands. The authors note that the EINF actually improves (decreases) with frequency for a given optical insertion loss, contrasting with the degradation seen in electronic systems.
• Bandwidth: The frontends demonstrated broad 3-dB intermediate frequency bandwidths of approximately 40 GHz, limited primarily by the photodetector.
• Communication Demonstration: The authors validated the system's practical utility through a proof-of-concept 6G-oriented multi-link communication experiment.
  • Single-Link: Achieved wireless data transmission up to 10 Gbit/s with bit-error ratios (BER) below the soft-decision forward-error correction (SD-FEC) threshold.
  • Multi-Link: Successfully supported two concurrent, independent downlinks from different angular positions and carrier frequencies, achieving a combined data rate of 20 Gbit/s.
  • Carrier Aggregation: Demonstrated the ability to aggregate two QPSK-modulated streams to double the data rate while maintaining BER below the SD-FEC threshold.

Significance and Claims
The paper claims that these integrated frontends represent a significant step toward compact, cost-effective, and energy-efficient THz wireless systems for 6G and beyond. By eliminating the need for high-frequency electronic amplifiers and power-hungry optical amplifiers, the architecture offers a simplified and scalable solution. The authors assert that the demonstrated low link noise and high conversion efficiency enable a paradigm shift toward cell-free wireless networks and smart industrial applications, where dense distributed access points and centralized processing are essential.

The authors remain modest regarding current limitations, acknowledging that the achieved noise figures are currently constrained by optical insertion losses (fiber-chip coupling and on-chip losses). They project that with optimized fabrication (e.g., improved edge couplers and lower-loss processes), the noise figure could be reduced to below 5 dB at 450 GHz, potentially enabling negative EINFs for the full receiving frontend—a regime currently inaccessible to electronic solutions. The work positions the TFLN platform as a viable foundation for future wafer-scale, 2D phased-array frontends with dynamic beam-steering capabilities.