Plasmonic modulators rely on the concept of surface plasmon photonics, which comprises the generation, guiding and detection of surface plasmon polaritons (SPPs). SPPs are electromagnetic surface waves that exist at dielectric-metal interfaces (Maier, 2007; Leuthold et al., 2013; Melikyan et al., 2014). Since the 1980s many research groups have focused on the modulation of SPPs with electrical signals (Schildkraut, 1988; Nikolajsen et al., 2004; Cai et al., 2009; Dionne et al., 2009; Melikyan et al., 2010; Randhawa et al., 2012). The first high-speed demonstration was given by Melikyan et al. (2014) who reported a plasmonic phase modulator only 29 µm-long and capable of high-speed operation over a bandwidth of more than 65 GHz.

A typical plasmonic phase modulator (Haffner et al., 2016a), Figure 1A, consists in two gold electrodes forming a metal-insulator-metal (MIM) slot waveguide only a few tens of nanometers wide. Light is fed to the plasmonic slot by means of a silicon strip waveguide. A linear taper transforms the photonic mode in the silicon nanowire to a SPP that propagates along the slot. The slot is filled with a non-linear organic material (Heni et al., 2016b; Jin et al., 2016) whose refractive index change Δn is linear with the electric field in the slot (Melikyan et al., 2014), according to Pockels effect (Boyd, 2003). Therefore, the phase of the SPP can be modulated by the field applied to the electrodes, and light is then coupled back to the photonic waveguide mode by an identical output taper. By careful engineering of the architecture, high modulation indexes could be obtained thanks to the strong electrical and optical field confinement in the slot, and the excellent overlap of the optical and electrical field modes, as shown in Figures 1B, C (Melikyan et al., 2014).

For THz applications, the modulator bandwidth should be sufficiently large to operate over hundreds of GHz. This is possible thanks to the quasi-instantaneous nature of Pockels effect and the extremely small RC time constant of the structure. The small slot height (∼100–150 nm) and short length (in the order of 10 μm) of the plasmonic waveguides allow to keep capacitances within the femtofarad range, while resistances are also very low thanks to the high conductivity of the metal electrodes. This yields a theoretical cutoff frequency in the THz range. Another unique advantage is the device’s compactness, which has recently allowed monolithic integration on electronic integrated circuit platforms, while keeping a low footprint and cost (Koch et al., 2019; Koch et al., 2020; Moor et al., 2022).

POH phase modulators can be employed for intensity modulation by arranging them in a photonic Mach-Zehnder interferometer structure, as shown by Heni et al. (2016a), Figure 1D (Heni et al., 2015; Hoessbacher et al., 2016). The modulator is driven in a push-pull configuration by means of a ground-signal-ground (G-S-G) RF probe as indicated by the labels in Figure 1D. Plasmonic IQ modulators with 100 Gbaud capability were also reported (Heni et al., 2018).

In the structures described above, most of the chip footprint is occupied by the photonic routing to and from the active plasmonic phase modulator sections. In an effort to reduce footprint even further, all-plasmonic modulators have been realized where also the splitting and combining functions needed for Mach-Zehnder operation are embedded in the plasmonic structure. Haffner et al. (2015b) have shown a Mach-Zehnder modulator (MZM) with a footprint of only 10 µm with insertion losses of 8.1 dB tested at 72 Gbaud symbol rates. Suspended metallic bridges enabled electrical connections within the plasmonic structure, Figure 1E. The same concept was later extended to a full IQ modulator (Haffner et al., 2015a), Figure 1F. Ayata et al. (2017) demonstrated an all-metal structure without any photonic waveguides, where grating couplers, polarization rotators, and Mach-Zehnder region are realized on a single metal layer, Figure 1G.

The low RC constant of plasmonic structures promises cutoff frequencies in the THz range. Actual measurements of maximum speed of plasmonic modulator devices have been mostly limited by available measurement instrumentation. In an effort to experimentally assess the 3 dB bandwidth of those devices, laboratory tests were conducted by using electrical driving signals from 75 MHz up to 500 GHz (Burla et al., 2019). In these experiments, a continuous-wave (CW) laser at 1547.5 nm with 0 dBm optical power was fed into a plasmonic MZM. The frequency response to the electrical signals were then determined in two different setups that covered the frequency ranges 15–70 GHz and 200–500 GHz. Subsequently, the optical spectrum of the intensity-modulated carrier at the modulator output was measured using an optical spectrum analyzer (OSA). The amplitude ratio between the fundamental and first side lobe in the spectrum was then used to determine the relative response of the electrical signal onto the optical signal (Alloatti et al., 2011). The frequency response was then normalized with respect to the power of the electrical signal driving the modulator. Results show a flat frequency response up to 500 GHz with a ripple that does not cross the −3 dB line compared to the low-frequency response, Figure 2A.

Converting a THz wireless signal to the optical domain is generally performed in two steps: first, the sub-THz signal is down-converted to baseband, e.g., using a Schottky barrier diode, and then up-converted to the optical domain by electro-optically modulating an optical carrier (Harter et al., 2019; Ummethala et al., 2019). In principle, this double-conversion process could be avoided by directly converting the sub-THz signal to the optical domain. Unfortunately, typical modulators are limited to bandwidths in the order of 100 GHz, rendering the process very inefficient or simply impossible at sub-THz frequencies (Leuthold et al., 2022).

In 2015 a direct RF-to-optical conversion relying on plasmonic modulators was introduced (Salamin et al., 2015). This scheme used a modified plasmonic modulator whose electrodes were shaped into resonant antenna arms, designed to be resonant with the modulator structure and, at the same time, to efficiently collect the impinging electric field, Figure 2B. This provides a built-in field enhancement in the slot compared to the impinging field up to a factor 90,000. Experiments were initially performed at 60 GHz (Salamin et al., 2015); a direct, high-efficiency conversion of sub-THz to optical waves is possible thanks to the inherent resonant field enhancement occurring in the plasmonic slot, visible in Figure 2C. We successfully transmitted 20 and 10 Gbit/s over wireless distances of 1 and 5 m, respectively, at a carrier frequency of 60 GHz (Salamin et al., 2018). More recently the concept has been successfully extended to the 300 GHz band (Salamin et al., 2019), Figure 2D. Additional demonstrations of efficient THz links enabled by plasmonic modulators include a 16 m link carrying 50 Gbit/s at 288.5 GHz (Ummethala et al., 2019). A more recent record demonstration enabled by plasmonic modulators is the transmission of 150 Gbit/s (line-rate) data over a transparent, optical-to-THz-to-optical wireless link operating at a carrier frequency of 228 GHz and a length of 115 m, without using any electrical driver amplifier at the modulator input (Horst et al., 2021; Horst et al., 2022).

The workhorse of any microwave photonic system is the analog optical link (AOL) or microwave photonic link (MPL) (Cox, 2006; Marpaung et al., 2019). Recently, analog links started to play an increasingly important role in next-generation wireless communications, as they allow to remote mm-wave and sub-THz remote antenna units from central offices without compromising analog bandwidth, Figure 2E. To explore the potential of plasmonics in this field, an analog RoF link was then implemented using a plasmonic modulator similar to Figure 1D. The experimental setup used to emulate the described RoF link scenario is shown in Figure 2F.

Direct THz-to-optical-to-THz conversion was achieved with a flat response over the 220–325 GHz range (>100 GHz bandwidth), only limited by the available electrical test equipment, as visible in Figure 2G. Besides bandwidth and speed, a main figure of merit of analog link is the spurious-free dynamic range (SFDR). This is mainly influenced by the properties of the modulator, in particular losses and non-linearities. Intercept points (Pozar, 2009) of a plasmonic MZM was measured (third order input intercept point, IIP3 = 18.9 dBm) to be on par with those of commercial modulators of equal half-wave voltage optimized for analog applications (IIP3 = 19 dBm). The dynamic range of the link described above was also characterized and displayed an SFDR3 of 105.2 dB/Hz^2/3 (Burla et al., 2019).

Phase noise is a critical aspect in any communication link, especially at high frequencies, because it directly impacts the achievable data rate and range. Its value strongly depends on the purity of the optical and electrical signal sources. In this work we do not directly address this aspect, widely discussed in literature (Khayatzadeh et al., 2014), but rather take into account the noise added in the system by the introduction of a photonic link by means of the noise figure parameter (NF). While in our first attempt we had not optimized the sub-THz link, leading to a high, RIN-dominated noise figure above 45 dB (Burla et al., 2019), this value can be reduced by several orders of magnitude by reducing the fiber-to-chip coupling losses and using well-established approaches such as low biasing of the Mach-Zehnder modulator (Marpaung, 2009).

In a MZM, half-wave voltage (V _π) indicates the voltage required to induce a 180° phase shift between the arms and, in turn, the change from a maximum to a minimum of the modulator transmission function. A low value of this parameter is premium, as it indicates that the modulator can efficiently convert electrical voltage signals to corresponding variations of the optical intensity. Together with insertion loss L, V _π directly impacts energy consumption per bit in digital links (Heni et al., 2019) and gain in analog optical links, respectively (Ackerman et al., 2016).

Half-wave voltage is linked to the modulator length (and, in turn, to its loss) by the V _π L parameter. At first instance, the loss can be reduced by shortening the modulator, and the V _π can be made smaller mainly by designing a longer modulator. The width of the plasmonic gap also has an effect on V _π, loss and modulation efficiency, as described in detail in (Haffner et al., 2016b). Several techniques allow to reduce half-wave voltage and loss simultaneously. Besides using electro-optic polymers with higher non-linearity, two approaches have been proven to be particularly effective for the scope.

A strong progress is also seen in the stability of operation over time and wide temperature ranges (Heni et al., 2019). Eppenberger et al. have reported stable operation at 80°C for more than 70 min, while the non-linear material itself showed a stable performance at 85°C for over 2000 h with less than 5% variance (Eppenberger et al., 2022).

Other research direction include the use of ferroelectric materials, such as barium titanate (BaTiO₃, BTO), which are suitable for wafer-level processing and have proven long-term high-temperature stability (Abel et al., 2019; Messner et al., 2019). This, in turn, enables high optical power handling, which is of high importance in microwave photonics applications.

Plasmonic modulators have shown the capability to be integrated on a single metal layer (Ayata et al., 2017). This opens the possibility to embed them on many different technological platforms. For example, recent demonstrations showed successful integration on silicon nitride (Kohli et al., 2022) with operation up to 216 GBd symbol rates. This is a very versatile photonic technology, offering ultra-low optical losses together with negligible non-linear absorption in the telecom wavelengths. The addition of plasmonic modulators equips this attractive platform with a new key capability, thus opening it to many additional applications and functions that require extremely high speed modulation.