For decades, the terahertz (THz) region of the electromagnetic spectrum, roughly 300 GHz to 3 THz, has intrigued researchers with its potential to enable faster wireless communications, ultra-high-resolution spectroscopy and more precise sensing. Yet, despite a solid understanding of the underlying physics, real-world THz systems have remained elusive. A major challenge has been the lack of THz sources capable of delivering both the output power and spectral purity required for practical applications.

That may be starting to change. Researchers at IMRA America have demonstrated a new THz source architecture that overcomes several long-standing limitations. By combining a resonant tunneling diode (RTD), a photomixed dual-wavelength Brillouin laser (DWBL) and a low loss waveguide circulator, the team achieved more than 40 dB of gain at 260 GHz. Just as important, the work includes the first reported characterization of the residual phase noise of an injection-locked RTD at this frequency, providing new insight into how high-power, low noise THz oscillators can be built and scaled.

While the demonstration was performed at 260 GHz, the underlying architecture is not limited to that frequency. Instead, it establishes a system-level approach for combining power and spectral purity at THz frequencies: two parameters that have historically been difficult to achieve simultaneously.

More broadly, the results highlight a shift occurring across the THz field. For many years, advances in sources and detectors have been constrained not only by device physics but also by the lack of supporting components that operate efficiently at these frequencies. As THz components continue to mature, particularly in waveguide form, new system architectures are becoming feasible.

FROM DIODE TO THZ OSCILLATOR

At the core of the system shown in Figure 1 is the RTD, a semiconductor device that directly generates THz radiation when properly biased. RTDs operate via a quantum-mechanical tunneling mechanism enabled by a double-barrier heterostructure. When electrons tunnel through the quantum-well structure, the resulting negative differential resistance region enables the device to function as a high frequency oscillator.

Unlike many THz sources, RTDs are compact, efficient and compatible with established semiconductor fabrication processes. This makes them particularly attractive for scalable and integrated THz systems. The waveguide RTD used in the IMRA study, fabricated by ROHM Semiconductor, demonstrated a tuning range exceeding 20 GHz, spanning approximately 240 to 260 GHz. This frequency agility is achieved through bias tuning, allowing the oscillation frequency to shift as the device operating point changes.

At an operating point near 260 GHz, the device produced approximately 40 μW of output power. For a chip-scale THz oscillator, this power level is significant. However, like many electronic oscillators operating at extremely high frequencies, spectral purity remains a challenge. The free-running RTD exhibited a linewidth on the order of several megahertz. This linewidth reflects the oscillator’s phase noise characteristics, which are influenced by both intrinsic device noise and the relatively low quality factor of the resonant structure surrounding the device.

While this level of spectral purity may be sufficient for some sensing applications, it is insufficient for systems that require extremely stable carriers. High-resolution rotational spectroscopy, molecular clocks and next-generation wireless communications all demand signals with far narrower linewidths. This tension between power and spectral purity has long defined the design space for THz oscillators.

INJECTION-LOCKED AMPLIFIER IN WAVEGUIDE

To overcome this limitation, the researchers applied injection locking, a technique widely used in microwave engineering to stabilize oscillators. Injection locking allows a lower-quality oscillator to synchronize with a higher-purity external reference. In this system, the reference signal was generated using a DWBL. Two optical tones separated by the target THz frequency were combined in a photomixer, producing a THz signal with exceptionally low phase noise.

When this signal is injected into the RTD oscillator, the RTD synchronizes to the reference frequency. Once locked, the RTD re-radiates the signal at significantly higher power, effectively functioning as a narrowband amplifier. This allows the system to combine the spectral purity of the optical reference with the output power of the electronic oscillator.

While injection locking itself is well understood, implementing it successfully at THz frequencies has proven challenging. Many earlier demonstrations relied on quasi-optical free-space setups, in which signals propagate through lenses and mirrors rather than through waveguide structures. Although these systems can demonstrate fundamental concepts, they often introduce high insertion loss, alignment sensitivity and limited scalability. Small changes in alignment can significantly affect signal strength and stability.

The IMRA system takes a different approach by implementing the entire architecture in waveguide. This reduces insertion loss throughout the signal chain, avoids many of the alignment challenges associated with free-space THz propagation and provides more precise control over signal routing and isolation. These two parameters are critical for injection-locked systems. The result is a more stable platform capable of demonstrating substantial amplification while preserving spectral purity.

THE ROLE OF THE HYBRID CIRCULATOR

The Micro Harmonics HC-34 hybrid circulator, shown in Figure 2, operates in the WR-3.4 band and was used throughout the IMRA injection-locking experiment. The device combines low insertion loss with high isolation at 260 GHz, allowing the injected reference signal to reach the RTD efficiently while suppressing reflections that could destabilize the oscillator.

This isolation is essential in an injection-locked oscillator. The injected signal must reach the oscillator with enough power to force synchronization, while reflected energy and amplified output signals must be directed away from the source. Without adequate isolation, reflected energy can disrupt the injection lock or create unwanted external cavities within the waveguide system.

Traditional Y-junction circulators are difficult to implement effectively at sub-millimeter wavelengths. At these frequencies, device dimensions become extremely small, and bandwidth limitations can introduce excessive insertion loss. To overcome these challenges, the Micro Harmonics hybrid circulator uses a broadband architecture based on an orthomode transducer (OMT) and Faraday rotation.

The OMT separates orthogonal polarization modes within the waveguide, while the Faraday rotator introduces non-reciprocal polarization rotation under a magnetic field. When these elements are combined in the proper configuration, they produce the directional signal routing characteristic of a circulator. Because both the OMT and Faraday rotator are inherently broadband devices, the resulting circulator operates across the full rectangular waveguide bandwidth, which is valuable for injection-locking systems that may need to operate over a tunable frequency range.

Figure 3 summarizes the HC-34 hybrid circulator’s performance across the WR-3.4 band (258 to 330 GHz). Its combination of low insertion loss and high isolation enables efficient routing of the injected reference signal while minimizing destabilizing reflections within the RTD amplifier chain.

These characteristics are particularly important in injection-locked systems. At THz frequencies, even a few decibels of additional loss can significantly reduce the power available for injection locking, while reflected energy can degrade spectral purity or disrupt synchronization. By combining high isolation with low loss, the circulator enables a fully waveguide-based injection-locking architecture, which has been difficult to achieve in previous THz experiments.

EXPERIMENTAL ARCHITECTURE

The experimental system, as seen in Figure 4, integrates photonic and electronic components into a single waveguide platform. Key elements include DWBLs, erbium-doped fiber amplifiers (EDFAs), uni-traveling-carrier photodiodes (UTCs), the waveguide circulator, the RTD oscillator and a phase noise analyzer used for system characterization.

A DWBL produces two optical tones separated by the target THz frequency. These tones are directed into a high speed photomixer, which converts the optical beat signal into a THz electrical signal. This THz signal serves as the injection reference.

The signal enters the waveguide network through the circulator and is directed toward the RTD oscillator. Once the RTD locks to the injected signal, the amplified output is routed through another port of the circulator and delivered to the measurement system. The measurement chain includes a harmonic mixer that down-converts the signal to an intermediate frequency. This allows precise characterization of both spectral linewidth and phase noise.

Because the entire system operates in waveguide, signal routing remains stable throughout the experiment. This stability is critical for long-duration measurements of phase noise and locking behavior.

PHASE NOISE AND RTD CHARACTERIZATION

Before applying injection locking, the researchers characterized the free-running RTD oscillator. The measured phase noise spectrum showed the expected behavior for a THz oscillator dominated by flicker frequency noise, with phase noise following an approximately f^-3 slope over a broad frequency range. Figure 5 then shows how the RTD’s phase noise changes after injection locking. As injection power increases, phase noise is progressively suppressed, and the RTD more closely follows the spectral purity of the photomixed reference source. The measurements also show that residual phase noise decreases as the injection ratio increases.

Using a Leeson-type model, the researchers extracted an effective quality factor of approximately Q ≈ 165 for the oscillator. Although this relatively low Q contributes to the broader linewidth of the free-running oscillator, it also makes the device easier to injection lock. Lower-Q oscillators generally exhibit wider locking ranges and require less injection power. This property makes RTDs particularly well-suited for injection-locking architectures.

INJECTION-LOCKING PERFORMANCE

When the photomixed reference signal was injected into the RTD, the oscillator rapidly synchronized to the reference frequency. The free-running linewidth of approximately 5 MHz collapsed dramatically, approaching the measurement system’s resolution limit. This indicates that the oscillator effectively inherited the spectral purity of the reference source.

Phase noise measurements showed reductions of up to 90 dB at low offset frequencies depending on injection power. Remarkably, injection locking could be achieved with extremely low reference power levels. Even nanowatt-level injection signals were sufficient to establish stable locking.

Figure 6 shows the measured relationship between injection power ratio and locking range. Open circles represent locking-range values derived from phase noise measurements, while the fitted curve follows injection-locking theory. The inverse of the injection power ratio corresponds directly to amplifier gain, illustrating the fundamental tradeoff between amplification and locking bandwidth. This result highlights the efficiency of the waveguide-based architecture and the importance of minimizing signal loss between the reference source and oscillator. The system ultimately demonstrated more than 40 dB of amplification at 260 GHz.

COMPARISON WITH CONVENTIONAL THZ AMPLIFIERS

Typical waveguide amplifiers operating in the WR-3.4 band (220 to 330 GHz) provide gains of roughly 20 to 25 dB. These amplifiers offer relatively broad bandwidth and higher saturation power, but they do not improve spectral purity.

In contrast, the injection-locked RTD architecture demonstrated more than 40 dB of gain while simultaneously reducing phase noise. The tradeoff is bandwidth. Injection-locked amplifiers operate over a much narrower frequency range, determined by the oscillator’s locking range. However, for many THz applications, including spectroscopy and precision sensing, spectral purity is far more important than wide bandwidth.

PATH TOWARD 1 THZ

Although demonstrated at 260 GHz, the architecture is not fundamentally limited to this frequency. Both the RTD and photomixing source used in the experiment can operate at significantly higher frequencies. With appropriate device scaling, operation near 1 THz appears feasible.

The primary limitation today lies in the availability of components capable of operating efficiently at these frequencies. As THz components continue to mature, including waveguide circulators, isolators and mixers, the same architecture demonstrated here could be extended deeper into the THz spectrum.

WHY IT MATTERS

The ability to generate high-power, low phase noise signals at THz frequencies addresses one of the longest-standing challenges in the field. Applications span communications, spectroscopy, sensing and timekeeping.

In wireless communications, higher carrier frequencies provide access to significantly larger bandwidths. This could enable extremely high-capacity data links for satellite systems, inter-satellite communication or future terrestrial wireless networks.

In spectroscopy, narrow-linewidth THz sources allow precise measurement of molecular rotational transitions. These measurements are critical for applications ranging from atmospheric science to chemical analysis. The architecture may also support the development of compact molecular clocks that use rotational transitions in gases to provide highly stable time references.

CONCLUSION

This work demonstrates a practical approach to combining high output power and low phase noise in THz sources. By integrating photonic and electronic technologies within a waveguide architecture and equipping the system with low loss THz components such as hybrid circulators, the researchers have demonstrated a viable path toward practical THz systems.

As component technologies continue to mature, architectures like this could help move THz systems from laboratory demonstrations toward practical deployment in communications, spectroscopy, sensing and precision timing.