1. Introduction

High energy physics (HEP) particles are mainly comprised of charged particles, like alpha ray (He⁺), beta ray (e^-), and charge-neutral particles, such as gamma ray (photons) and neutrons. Those rays usually have MeV in energy and Mrad in dose. When incident on a material, they interact with the material’s atoms and transfer their large kinetic energy and momentum to the target atoms through both columbic collision and ionization, creating electronic configuration distortion and leaving behind material defects. These defects, in turn, alter the electric field distribution, carrier density, refractive index, and absorption of materials, altering device behavior.

The endurance of photonic devices under radiation harsh environment has long been a major concern for space and nuclear applications. As silicon photonics exhibit both speed and energy merits over their electronic counterparts, increasing number of photonic integrated circuits (PICs) are being deployed in satellites as well as in nuclear facilities [1–5]. For example, Starlink now takes advantage of space laser communication systems to interconnect satellite constellations [6], and the large hydron collider (LHC) at CERN is considering an upgrade to low power optical data links with silicon PICs to fulfill the ever-growing data traffic [7]. Both of them evidently indicate a growing trend of leveraging photonic devices under radiation situations. However, radiation induced damage to photonic devices still awaits systematic studies, and the damage mechanism also requires detailed investigations. In fact, numerous researches have reported radiation effects on silicon photonics over the past decades. They include both active components such as lasers [8–12], modulators [13–17], detectors [18–21] and passive ones such as waveguides [22–24], ring resonators [25–28] as well as MZIs [29]. It is worth highlighting that it is significant to thoroughly evaluate radiation induced effects and construct radiation-hard devices before launching photonic devices into those harsh environments. Understanding radiation damage requires correlating macroscopic device behavior to microscopic defect generations. This review article aims to provide a summary of the literature, from radiation damage theory to device performance, to present an overview of radiation induced effects and provide possible solutions for building radiation resistant photonic devices. We first introduce the microscopic particle interaction mechanism, then elaborate the corresponding defect generation and how they impact a materials’ property. Next, we move to device level, discussing macroscopic device performance modifications induced by radiation. Mitigation strategies are provided subsequently, and finally, we conclude this article with an outlook and summary.

2. Theory and modeling

High energy radiation effects are widely recognized and categorized into three types. Total ionizing dose (TID) effect that characterizes a cumulative ionization damage, displacement damage dose (DDD) effect that is associated with long-term structural damage, and a transient single event effect (SEE) [30]. The comparisons and schematic illustrations of them are summarized in Table 1

Table 1. A comparison between three types of damage effect

View Table | View all tables in this article

One of the primary ways where radiation particles transfer energies to the target is by the creation of charges. It is commonly observed when irradiated with charged rays. The incident ion either excite the electron of the target atom to the conduction band or direct absorb/provide the target atom with an electron, generating excessive charge inside the target substrate. These charges are easily trapped in dielectric materials, such as silicon nitride and silicon dioxide, forming bulk trapped defect states. Since electrons are more mobile than holes, the generated electrons move freely with the electric field, leaving behind net positive trapped charges. These free electrons modulate materials’ electrical conductivity, refractive index, and absorption coefficient via the plasma dispersion effect and free carrier absorption, respectively. Trapped charges, on the other hand, become sources of local electric field, modifying the target atom's polarizability and altering its refractive index [31]. Moreover, as this electric field accumulates, especially at interfaces, it is large enough to counteract externally applied voltage when operating, leading to device performance degradation. Another type of charge trap is surface states, which originate from surface dangling bonds. As the generated holes transport to the surface, the dangling bonds favor donating its unpaired electron to the hole, forming a positively charged surface defect. If the irradiation environment is ambient, these surface defects inevitably react with oxygen in air, developing into a thin oxide capping layer [25,29]. The surface oxidation no doubt further contributes to macroscopic device performance modification.

Apart from the ionization effect, when the incident atom is sufficiently energetic, whose kinetic energy is higher than the displacement threshold energy, it knocks out a target atom (primary knock-on atom, PKA) from its original lattice site, creating a vacancy and an interstitial atom (Frenkel pairs). If the PKA’s recoil energy is still above the displacement threshold energy of the substrate atom, it continues collision and produces a secondary knock-on atom (SKA), cascading the displacement process. These defects are true structural damage and are usually deep-level defect states [32]. It recombines charge carriers, hindering their transportation and reducing their lifetime, which is detrimental to most opto-electronic devices. Furthermore, the structural rearrangement also involves volume expansion and compaction, which in turn modify material’s refractive index [33]. It worth pointing out that, gamma rays as massless particles, bear no DDD effect at first glance. However, gamma rays, interact with the target substrate through the photoelectric effect and Compton scattering, generating highly energetic secondary electrons. These electrons are massed particles, capable of initiating the displacement defect cascading process. The DDD effect has been observed to be the dominant damage mechanism in 60Co gamma radiation. [26,34]

In addition to the aforementioned chronic effect, SEE also takes place and is of concern when the device is continuously operating under irradiation. When an extremely energetic single particle passes through the target substrate, it leaves behind a trail of charges generated by exciting valence electrons to their conduction band. Then those electron-hole pairs quickly recombine and recover. The sudden change and recovery in carrier concentration and local electric field yields a transient device behavior modulation. Macroscopically, an unusual spike in voltage/current or dip in the transmission spectrum is the typical consequence of SEE. SEE normally recovers in the pico-second range after the generated charges diffuse and recombine and are mostly non-destructive [35]. We want to emphasize that, though these three types of damages are independent of each other, they happen simultaneously under one radiation event. For example, an alpha ray is capable of generating electron-hole pairs while displacing the Si substrate atom, yielding both TID and DDD.

Modeling a single radiation particle’s trajectory and quantifying its energy loss is possible. However, it is a random process whose effects depend heavily on the local target atom arrangement, collision cross-section, and collision angle. A commercially available software called Stopping Range of Ions in Matter (SRIM) provides an ensemble distribution by adopting the Monte-Carlo method to simulate large groups of incident particles [36]. It calculates the trajectory, stopping range, and energy loss of the incident particle. Linear energy transfer (LET) is a function that characterizes the energy loss of an ionizing particle transferred to the target substrate per unit length. It’s proportional to the stopping power. On the other hand, displacement damage is quantified by another function, nonionizing energy loss (NIEL), of the impinging radiation atom, which measures the energy deposited on the target atom via collision. It is worth mentioning that both functions consider the energy losses of the incident particle instead of the substrate. As a consequence, secondary effects, such as SKA, etc., are neglected. Therefore, using NIEL generally underestimates substrate displacement damage when the incident particle is highly energetic. Figure 1 illustrate an exemplary plot of the stopping range, LET, and NIEL of a 100 keV alpha particle radiation on a 220 nm thick SOI photonic device from a SRIM simulation. It clearly indicates that under such circumstances, TID is orders of magnitude more pronounced than DDD in the Si device layer, and most displacement damage happens deep inside the BOX layer. A detailed analytical model describing the dynamics of TID and DDD can be found in [32]. The simulated results of LET and NIEL are a direct indication of the occurrence rate and intensity of ionization and displacement damage events, respectively. They scale with the population of generated carriers and lattice defects. By constructing appropriate physical models, microscopic damage events and macroscopic device behavior could be correlated. To date, unfortunately, such models are still challenging due to the random collision nature of the defect generation process.

 

Fig. 1. SRIM simulation result of alpha particle radiation damage on a 220 nm SOI device. The discontinuity of LET and NIEL at the interface is the indication of defect accumulation.

Download Full Size | PDF

3. Radiation damage effect in photonic devices

Compared to the microscopic damages discussed above, radiation induced device performance modification is much more predictable. Device-level operation parameters were monitored in-situ or evaluated before and after radiation. It is interesting to find that the robustness of devices against radiation varies depending on their specific operation type and principle. The detailed responses are elaborated below and summarized in Table 2.

Table 2. Summary of radiation effect on photonic devices

View Table | View all tables in this article

3.1 Lasers

Radiation study of lasers were launched merely several years after laser invention [37]. To date, a number of works have reported radiation effect on lasers. Among them, radiation type ranges from proton, neutron to gamma rays, covering both types of damage. A clearly observable damage is the increase in their threshold current, which is much more pronounced when irradiated with massed particles other than photons [38]. Under extreme conditions such as 2 × 10¹⁵ fluence of 20 MeV neutron radiation, catastrophic failure was reported [8]. In addition to the threshold current elevation, a slope efficiency reduction was also found [8]. It decreases in a sublinear manner with cumulative fluence or dose until completely stopped lasing. These effects primarily resulted from defect generation in laser materials. A laser starts to lase when the population inversion condition is met. However, radiation induced defects are carrier trap states and recombination centers that greatly enhance the nonradiative recombination process and decrease carrier concentration. Consequently, the domination of such process over radiative spontaneous and stimulated emission degrades laser performance.

3.2 Modulators

Modulators are mostly Mach-Zehnder interferometers, whose modulation arms are heavily doped silicon p-n diodes. As illustrated in Figs. 2(a) and 2(b), the leakage current increases after exposure to both neutron and photon irradiation owing to the mid-gap state and charge trap generation, which is consistent with radiation induced defect models [13,39]. The modulation efficiency degradation is quite pronounced, as is evidenced in Ref. [1]. The result shows that complete deterioration of the modulator is possible after receiving tremendous TID (X-rays up to 100 Mrad), while DDD effects (neutrons up to 10¹⁵ fluence) are much less significant. Studies suggest that the vulnerability to TID is primarily attributed to a “pinch-off” effect. Under radiation, the generated positive charges accumulate at the SiO₂/Si interface. It acts as a “gate” electric field, modulating the carrier transport. When this field is sufficiently large, it empties the minority carrier in the p doped slab region, which becomes non-conductive. Such separation between the p contact and the rib renders the modulator unfunctional [40]. On the contrary, since MZM are inherently heavily doped, the number of defects generated through DDD is much less significant compared to its doping level. Consequently, MZMs are generally robust against DDD.

 

Fig. 2. a) leakage current increase under neutron irradiation; b) leakage current increase under X-ray irradiation and c) relative phase shift under X-ray irradiation. (Reprinted with permission from [39])

Download Full Size | PDF

3.3 Detectors

The SiGe waveguide integrated photodetector is currently the unique type that silicon photonic foundries offer owing to its mature fabrication process [41]. It consists of a p-i-n structure where the intrinsic Ge serves as the light absorber. It has been demonstrated that both TID and DDD produce negligible degradation for SiGe detectors with respect to dark current, responsivity, and operation bandwidth [19,42,43]. The high radiation tolerance in SiGe detectors is a direct consequence of pre-existing bulk defects. The detector core material Ge is lattice mismatched to the waveguide core Si, making epitaxial growth of defect-free single-crystalline Ge challenging [44]. Though defects are considered detrimental in most scenarios, the high trap density in Ge guaranteed its dominance over radiation generated ones, making SiGe detectors resistant to radiation damages. Application-wise, however, it is worth pointing out that SEE needs special attentions in data communication. The sudden generation of a large number of carriers in a single event corrupts the data train and yields bit errors [20]. Unfortunately, such an event is difficult to be shield or mitigate using the methods discussed in section 4 and therefore requires manual data filtering.

3.4 Waveguides

In contrast to active devices, whose optoelectronic responses are the major concern, passive waveguides weigh more on their optical properties: material refractive index and absorption. Nevertheless, radiation induced optical constant change is usually in the order of 10⁻⁴ which is not accurately detectable with conventional spectroscopy. To overcome such a challenge, resonant cavity enhanced spectroscopy is introduced, where waveguides are made into resonators, the small change in their refractive index is reflected as a cavity resonance peak shift as is indicated in Fig. 3(a). It has been demonstrated that such technology is capable of detecting radiation induced refractive index changes down to 10⁻⁵ [25]. Modal effective index increases in most widely recognized silicon photonic waveguide core materials, such as Si, a-Si, SiN, and SiC, are identified with such a method [25,26,45]. It is worth pointing out that, though resonant cavity enhanced spectroscopy is sensitive to small phase shifts, detecting small loss changes is still beyond its scope. Instead, swept-wavelength interferometry on an arrayed waveguide structure (displayed in Fig. 3(b)) was built to measure a loss change with an accuracy of 0.1 dB/cm [23]. A propagation loss increase of about 0.5 dB/cm is found in the silicon waveguide after receiving 100 krad of gamma radiation. These refractive index and loss modifications are the overall consequences of surface oxidation, mid-gap state absorption, and displacement damage induced volume compaction or expansion.

 

Fig. 3. a) Radiation induced resonant peak shift in a micro-ring resonator and b) optical micrography of arrayed waveguide structure to measure radiation induce loss change. (reprinted with permission from [23])

Download Full Size | PDF

4. Mitigation strategy

Post-radiation annealing has been widely adopted to mitigate radiation induced effects since the damages are primarily created via defect generation. The defect kinetics follow the Arrhenius’s behavior [46]. Once the temperature is reached to overcome the activation energy barrier, they starts to recover [47,48]. Yet, most devices are not robust against high temperatures, and room temperature annealing is quite ineffective, taking even months to recover. Figure 2(c) clearly indicated that no slight trace of recovery was found after 50 hours. Fortunately, forward bias annealing was discovered as an effective way to address TID damage. It has been demonstrated that applying a forward bias can accelerate the annealing process down to tens of hours, and full recovery is observed in irradiated MZMs [48]. During forward biasing, the injected electrons tunnel to the interface and the oxide, neutralizing the trapped holes. This method has also been shown to be effective for compensating radiation effects in-situ [2].

Inspired by the robustness of heavily doped device to DDD. It is evidenced that shallow-etch or higher contact doping enhances the radiation tolerance of active devices to TID. Under such a configuration, it requires an even larger dose of radiation to deplete the carriers and “pinch-off” the device [13]. Simulation results have indicated that a 150 nm slab height MZM is still operational, while a 100 nm slab height MZM stopped functioning after receiving 100 Mrad of X-ray doses [40].

Lastly, in terms of a passive waveguide where the optical mode’s transmission property is the main focus, correlating the effective modal index to the waveguide’s material index turns out to be an effective approach. During an in-situ radiation analysis of the SiC waveguide, as illustrated in Fig. 4, it was found that the waveguide core and cladding refractive index responded to gamma radiation in the opposite direction. Therefore, a gamma radiation-hard waveguide design could be constructed by carefully engineering the waveguide modal confinement factor to compensate for radiation induce refractive index change [26]. With a proper modal field distribution inside both materials, the radiation damage could potentially be mitigated to minimal as shown in Fig. 4(c).

 

Fig. 4. a) gamma radiation induced material refractive index change; b) confinement factor engineered radiation hard waveguide geometry and c) projected radiation induce waveguide effective index change under gamma radiation. (Reprinted with permission from [26])

Download Full Size | PDF

5. Outlook and summary

Decades of research have shown that a radiation-rich environment can alter photonic device performance and be especially harmful to active components. We outlined microscopic mechanism of radiation induced damage and summarized macroscopic device behavior modification. It is found that, apart from high-temperature annealing, applying a forward bias to devices promotes device recovery. Waveguide modal confinement factor engineering also helps compensate for and minimize radiation induced optical constant modification. In general, shielding is by far the best way to prevent radiation damage. It is robust against almost all types of radiation and only requires centimeter-thick lead cages. Indeed, such methods have already been widely deployed in nuclear facilities. However, for space applications, launching weight is instead the primary concern, where bulky lead chunks are not an option. Therefore, it is essential to develop effective mitigation approaches to secure the functionality and fidelity of photonic devices. Currently, we still face several challenges in radiation damage studies that await further investigation:

We anticipate that with advances in radiation physics and material defect mitigation, we will be able to successfully deploy photonic devices on satellites in the near future, allowing for faster, more accurate and more reliable space communication and sensing.