High‐Precision Wavelength Tuning of GeSn Nanobeam Lasers via Dynamically Controlled Strain Engineering

Abstract The technology to develop a large number of identical coherent light sources on an integrated photonics platform holds the key to the realization of scalable optical and quantum photonic circuits. Herein, a scalable technique is presented to produce identical on‐chip lasers by dynamically controlled strain engineering. By using localized laser annealing that can control the strain in the laser gain medium, the emission wavelengths of several GeSn one‐dimensional photonic crystal nanobeam lasers are precisely matched whose initial emission wavelengths are significantly varied. The method changes the GeSn crystal structure in a region far away from the gain medium by inducing Sn segregation in a dynamically controllable manner, enabling the emission wavelength tuning of more than 10 nm without degrading the laser emission properties such as intensity and linewidth. The authors believe that the work presents a new possibility to scale up the number of identical light sources for the realization of large‐scale photonic‐integrated circuits.


Note 1. Device design, fabrication process and lasing characteristics
Design of a GeSn photonic crystal nanobeam cavity with two large pads: The width and length of the GeSn one-dimensional (1D) photonic crystal nanobeam cavity were designed to be 700 nm and 10µm, respectively. The photonic crystal was designed with the same approach as in our previous work. [1] The photonic crystal consists of periodic air holes with a lattice constant of 350 nm and three missing inner air holes at the center of the nanobeam. To reduce the scattering losses in the photonic crystal cavity by employing a gradual variation of the refractive index, the diameters of the first inner holes and the other holes are designed to be 112 and 217 nm, respectively. The two large pads connected to the cavity were designed with a width of 15 µm.
Fabrication of a GeSn nanobeam cavity with two large connected: Using GeSn-on-insulator (GeSnOI) substrate, [2,3] a GeSn photonic crystal nanobeam cavity was fabricated. Electronbeam lithography (EBL) was performed to pattern the nanobeam cavity and then Cl2 dry etching with reactive ion etching (RIE) was conducted to transfer the pattern to the GeSn layer. To suspend the nanobeam and two large pads, the Al2O3 layer under the GeSn layer was selectively wet etched with 30 wt.% potassium hydroxide (KOH) at 80 °C. After the undercut of Al2O3 layer, HfO2 thin film with a thickness of 90 nm was coated for the surface passivation through the atomic layer deposition (ALD) at 100 °C using tetrakis(ethylmethylamino)hafnium (TEMAH) and H2O as the metalorganic and oxidation precursors, respectively. Lasing characteristics: To investigate lasing characteristics from the fabricated devices, we conducted photoluminescence measurements on an unannealed GeSn nanobeam laser. Figure   S1a shows the emission spectra of the GeSn laser pumped with the power densities of 56. Note 2. XRD analysis to determine critical temperature Figure S2. Specular 2θ-ω scans of (004) diffraction for unannealed and annealed GeSnOI samples. Annealing was conducted at temperatures of 600, 700 and 800 K by RTA. The black and red arrows mark the positions of the GeSn diffraction peaks for all samples. The unannealed sample and the sample annealed at 600 K show the same GeSn diffraction peak at ~64.7°, indicating that the Sn segregation is not invoked at 600 K. The samples annealed at 700 and 800 K show the same GeSn diffraction peaks at ~65.9 °. The GeSn diffraction peaks of 700 and 800 K samples are closer to the Ge diffraction peak, indicating that Sn segregation is invoked beyond 700 K. Via XRD analysis, Tc is estimated to be 700 K.
To determine the critical temperature (Tc) that induces the Sn segregation in our GeSn with an Sn content of ~10 at%, X-ray diffraction (XRD) spectroscopy measurements were performed for unannealed and annealed GeSnOI samples. Rapid thermal annealing (RTA) was performed on GeSnOI, the same substrate used for the tuning experiments, at temperatures of 600, 700 and 800 K. Figure S2 shows specular 2θ-ω scans of (004) diffraction of unannealed and annealed samples. The GeSn diffraction peak positions for all samples are marked by black and red arrows. The unannealed sample and the sample annealed at 600 K show the same GeSn diffraction peaks at ~64.7° corresponding to the Sn content of ~10 at%, indicating that Sn is not segregated by the temperature of 600 K. For the samples annealed at 700 and 800 K, the GeSn diffraction peaks are observed at the same position of ~65.9° corresponding to the Sn content of ~1 at%. The GeSn peaks are very closer to the Ge diffraction peak, indicating a significant reduction of the Sn content after annealing at 700 and 800 K. According to these results, we estimated the Tc as 700 K. To determine an annealed area by localized laser annealing in our device, we performed a thermal simulation using finite-element method (FEM). Figure S3 shows a thermal distribution simulated with a heat source mimicking the pumping condition used in the annealing experiments. A pulsed heat source with a pump power density of 28 MW cm −2 was applied at one of the suspended pads. The repetition rate and pulse duration were set to 250 kHz and 5 ns, respectively. The base temperature was set to 4 K. From the calculated thermal distribution, an area exceeding Tc (700 K) can be seen at the heated pad. The area is calculated as ~28.2 µm 2 and we set this area as an annealed area in the strain simulation. It is noteworthy that the temperature rise occurs only in a very localized area, preventing temperature-induced changes in other parts of the devices.

Strain in the Sn segregated area based on the change in lattice constant
Strain is defined as a change in dimension divided by the original dimension. In our system, the lattice of Ge0.99Sn0.01 should be formed in the annealed area due to Sn segregation, but in reality the atoms are rearranged in the lattice of Ge0.90Sn0.10. Consequently, the strain based on the change in lattice constant can be written as: By employing lattice constants of Ge0.90Sn0.10 and Ge0.99Sn0.01 calculated above, we obtain inplane tensile strain of ~1.3%. Figure S4. Calculation of tensile strain as a function of emission wavelength. a) Simulated spectra (left panel) of our structure with varied refractive indices using FDTD simulation. The peak shift corresponding to each refractive index is shown in the right panel b) Theoretically calculated <100> uniaxial tensile strain as a function of refractive index.

Note 5. Calculation of tensile strain as a function of emission wavelength shift
To investigate how much the tensile strain is induced in the laser gain medium, we conducted finite-difference time-domain (FDTD) optical simulations and theoretical calculation. Figure   S4a shows simulated spectra (left panel) of our 1D photonic crystal nanobeam cavity structure with varied refractive indices using FDTD simulation. The peak shift corresponding to each refractive index is also shown in the right panel. The simulated spectra show a single mode for all refractive indices, indicating no mode hopping in our structure. From the peak shift with varied refractive indices, we confirmed that the peak shifts by ~12 nm for 0.02 index increase.
We then calculated the strain using refractive index change by referring to a recent study. [6]  Pulsed pumping with a pulse width of 5 ns and repetition rate of 250 kHz is applied. During the pulse duration of 5 ns, the temperature reaches a maximum of ~750 K. Until the next pulse, the temperature reaches a minimum of ~130 K due to the heat dissipation. It is evident that there is no heat accumulation over time.
To understand the fixed peak shift of ~12 nm for varied annealing times (Figure 3c), we analyzed the temperature of the annealing region over time in Fig. S3a. Figure S6 shows the simulated temperature of the annealing region as a function of time. Due to the pulsed pumping with a pulse width of 5 ns and repetition rate of 250 kHz, the temperature increases and decreases periodically. During pulse duration of 5 ns, the pumping is induced and the temperature reaches a maximum of ~750 K over the Tc (700 K). Until the next pulse, the heating dissipates and the temperature reaches a minimum of ~130 K, which is much lower than Tc.
This rising and falling of the temperature is repeated every pumping period without any heating accumulation over time. This indicates that the way our devices are annealed is not a thermal accumulation that takes a long time to complete the annealing process. It can further be deduced that the annealing process can be completed for the short pulse peak duration of 5 ns. According to the analysis, the annealing times performed in our experiments (Figure 3c) do not show any time-dependent effect, possibly because the annealing process may be completed before 10 s.

Note 8. Formula describing relation between annealing power and wavelength shift
We derived an empirical formula that correlates annealing power with the shift in emission wavelength. The formula is modeled as an exponential function because the relation between annealing power and strain shows an exponential correlation in the thermal and strain simulations (see Supporting Information Note 6 for more details on annealing power-dependent strain in our device). On the other hand, the other relations such as strain vs. refractive index and refractive index vs. wavelength shift show a linear correlation (see Supporting Information Note 5 for more details on strain-dependent refractive index and refractive index-dependent wavelength shift). Therefore, the emission wavelength changes exponentially as a function of the annealing power. We fitted an exponential equation to the data points acquired through experiments for annealing power-dependent wavelength shift in Figure 3b. As a result, the formula describing the relation between annealing power and wavelength shift can be written as follows: ∆ = 0.26293 5 ) )''"*+,'-*.,-.// − 17