1. Introduction

Quantum well intermixing (QWI) based on impurity-free vacancy diffusion (IFVD) has been widely used for bandgap engineering and photonic integrated circuit (PIC) applications because of its simplicity, high crystalline quality, and low optical loss. The scheme can be achieved through capping SiO_x, SiO_xN_y, or oxide on GaAs-related materials followed by rapid thermal annealing (RTA). The QWI process, i.e., atom interdiffusion inside the quantum well (QW), can generally be enhanced by the excess vacancies created at the contact interface between the capping layer and capped GaAs-related material. By controlling the chemical reaction of Ga₂O₃ from the oxide and Ga compound, the generated vacancies can promote the diffusivity of atoms in QWs. For example, with the InGaAsP multiple quantum well (MQW), QWI processing leads to a higher Ga content in the well and In content in the barrier, shifting the energy level of the QW toward higher energies. Therefore, by patterning the SiO_x capping, QWI can locally tune the bandgap for a better performance of each component in PICs, thus enhancing the overall performance. Therefore, the processing enhancement of local IFVD QWI will be quite important for future photonic integration applications.

However, IFVD intermixing also faces some challenges. Because of the poor thermal stability in the InGaAsP system, the control of the wavelength shift between QWI and suppressed QWI is a major challenge during the high-temperature stage, leading to small or unreliable differential shifts [1]. In addition, it has been found that the high diffusion properties of zinc, a typical p-dopant, at high temperatures will generate an active region with a high p-dopant concentration [2–4]. Thus, a low gain and high optical loss in the active region from the free carrier absorption can be formed, deteriorating the device performance. Furthermore, the diffusion of the p-dopant in the cladding layer will decrease the electrical conductance, leading to a connection problem between devices in the PIC. Such p-dopant diffusion into the active region can also inevitably cause a non-uniform electric field distribution, where each QW in the active region experiences different field-driven properties. In the viewpoint of PIC fabrication, all these factors make it difficult to control the individual device and limit the integration function. Based on the aforementioned discussion on IFVD QWI, the thermal problem of high-temperature processing is the main challenge for realizing PICs through local bandgap engineering.

In order to realize high-performance PICs through the simple IFVD QWI scheme, lowering the temperature during the intermixing process is thus essential to achieving a high wavelength contrast from a local SiO_x cap layer. By depositing the patterned SiO_x layer for performing bandgap engineering while maintaining the same properties in the other area, several groups have developed some methods for enhancing the QWI. The deposition of certain properties of SiO_x has been demonstrated by setting different conditions such as using plasma-enhanced chemical vapor deposition (PECVD). By controlling the chamber temperature, PECVD-deposited porous SiO_x can be obtained for enhancing the mobility of Ga atom outdiffusion, thereby increasing the amount of intermixing [5]. Oxygen-rich SiO_x was deposited through setting a precursor, N₂O and SiH₄, flow rate, which introduced an increase in the metallurgical reaction between the SiO_x layer and GaAs [6-7]. The enhanced QWI induced a monotonic increase in the blueshift. With another scheme, SiO_x tensile strain on the interface can cause the wafer substrate to exhibit a higher diffusion property, resulting in a higher intermixing mechanism [8]. Using Ge-doped sol-gel-derived SiO_x as an encapsulated layer, the thermal expansion coefficient increases to introduce stress relaxation while annealing, hence efficiently retaining more vacancies for improving the QWI [9]. Other than enhancing the QWI, p-doped silica with a denser SiO_x:P and fewer voids than the standard SiO_x can suppress the QWI in the cap area [10]. Controlling the atoms ratio of SiO_xN_y in PECVD-deposited films causes broad refractive index distribution in the capping dielectric layers, achieving both red and blue shift in the spectrum and also showing 300 nm range of laser diode for photonic integration [11-12]. The blue shift can be enhanced or inhibited by using different thickness of ZrO₂ cap layer, showing low extra loss from QWI [13-14].

Ion implantation techniques have also been applied to the substrate material properties, changing the properties to assist the QWI and lowering the annealing temperature. Phosphorous or arsenic atoms are used to create vacancies near the QWs, where high energies up to 8 MeV with a 7° tilt angle are used to achieve a 90-nm blueshift [15]. Regrowth after ion implantation [16-17], ion channeling [4], or double-charge ion techniques for overcoming the ion energy limitation have also been demonstrated [18-19]. Although large blueshifts can thus be obtained from the vacancies, such specific procedures require the accurate control of the parameters of ion implantation in the system, causing this process to be expensive, time-consuming, and complex.

Consequently, based on our previous work, a new approach called supercritical fluid (SCF) treatment on the SiO_x cap layer and also a detailed material analysis has been used to control the properties of the material for enhancing the IFVD QWI [20]. SCF processing has been widely known to possess advantages of both the gas and liquid phase, where high diffusivity, high density, low viscosity, and low surface tension can be found during the surface treatment of wafers. By using an SCF as a carrier to deliver chemical atoms or molecules, it can thus be applied easily for the deep penetration of the reaction into a porous structure. CO₂ is the common material used for forming SCF because of its low cost, low pollution, and low critical point [21]. By carrying H₂O₂, the CO₂ SCF treatment can introduce a high reaction and high oxygen content into the SiO_x cap layer for improving the chemical reaction of Ga₂O₃. With this treatment, a high tolerance in controlling the SiO_x properties can be expected because oxygen-rich surroundings can be applied to the inner part of the cap layer. The CO₂ SCF only affects the encapsulated SiO_x layer, thus being quite suitable for enhancing the QWI for local bandgap engineering and PIC fabrication.

In this study, the SCF treatment of a SiO_x cap layer has been proposed and demonstrated for enhancing the IFVD QWI of an InGaAsP material. The measurements of X-ray photoelectron spectroscopy (XPS) and Fourier-transform infrared spectroscopy (FTIR) on SCF SiO_x samples suggest that a high oxygen content can be produced after annealing. Lower-temperature QWI was then observed for obtaining the same photoluminescence (PL) shift of the MQW, suggesting that the diffusion of the p-type Zn dopant can be reduced. In addition, the PIC consisted of an electroabsorption modulator (EAM), and a semiconductor optical amplifier (SOA) was fabricated by SCF IFVD QWI, leading to a larger modulation depth and optical gain. A high-speed electro-optical (EO) response with an eye diagram of data modulation was found to verify such processing.

1. SCF treatment of SiO_x material

Figure 1 shows a schematic of how the SCF treatment on the patterned SiO₂ locally affects the QWI processing, in that local bandgap engineering can be realized. The concept is as follows: the CO₂ SCF with H₂O₂ first introduces an OH group into SiO₂ through the high diffusivity and high pressure of the SCF. After RTA at high temperatures, the dehydration process to remove hydrogen thus leaves more oxygen to contribute to more Ga₂O₃ reactions. The higher vacancy content will induce stronger QWI. In order to confirm this, two experiments were conducted: the material analysis of SiO_x related to the SCF processing (left of Fig. 1), and the fabrication and measurement of the EAM-integrated SOA for the demonstration of the PIC based on the SCF.

 

Fig. 1 Schematic depicting the SCF technique for generating OH groups for a higher Ga₂O₃ reaction, thus improving the local QWI.

Download Full Size | PDF

In the first experiment, to control the conditions of QWI during RTA, SiO_x with a fixed thickness of 200 nm for all the wafers was deposited by an electron gun beam evaporator system. Four samples (A, B, C, and D) with the same conditions as the SiO_x deposition were tested as follows: sample A was the reference without any further processing, B was set for RTA without the SCF, sample C was processed with SCF treatment directly after deposition, and sample D was processed with RTA after SCF treatment. SCF CO₂ with H₂O₂ was used for delivering OH groups into SiO_x, where a 3000-psi pressure and 60 °C were set for the SCF chamber. The samples then underwent RTA at 700 °C for 90 s under N₂ flow. In order to analyze the role of OH groups during the SCF treatment on SiO_x, spectroscopic ellipsometry, FTIR, and XPS were performed. The refractive index of the SiO_x layer was measured by spectroscopic ellipsometry in the visible range with an Hg–Xe source. Infrared measurements were performed on the oxide layer using a Nicolet FTIR spectrometer with a resolution of 1 cm^–1. The oxygen content of the SiO_x layer was determined by XPS.

FTIR was measured to obtain the related atom bond emission. As plotted in Fig. 2(a), the FTIR spectra in the region of 2600–3800 cm^–1 was measured to illustrate the OH-related band for different conditions. Clearly, sample C has a significantly higher OH bond content in comparison to the reference (sample A), suggesting that the CO₂ SCF can efficiently deliver OH into the SiO_x film by carrying H₂O₂. However, after RTA, samples B and D exhibit almost the same OH band levels. Further, the OH band clearly shows a lower level than the reference sample and SCF sample (sample C), implying a condensation reaction from dehydration. Furthermore, as shown in Fig. 2(b), the Si–O–Si in-phase stretching peak (S1) shifts from 1073 to 1068 cm^–1 after 700 °C annealing, further confirming the densification of the deposited layers. It can be concluded that most of the Si–O–Si band remains after RTA.

 

Fig. 2 (a) Measured FTIR spectra of the OH-related band for samples A, B, C, and D. (b) Spectrum for the Si–O–Si in-phase stretching band.

Download Full Size | PDF

The refractive index measurement of the sample revealed more information. Figure 3(a) presents the refractive index spectrum extracted from an ellipsometer, where the optical wavelength ranges from 500 to 700 nm. Sample C and the reference (sample A) give the highest and lowest values respectively, reconfirming that the defect filling process through the SCF leads to the highest FTIR measurement of sample C in Fig. 2(a). However, for the other samples, B and D, different trends are found. As shown in Fig. 3(a), the refractive index of sample B is higher than that of D for all wavelengths, while the FTIR results of samples B and D almost merge together (Fig. 2(a)). This implies that the condensation from the dehydration during RTA could increase the refractive index. Sample D experiences the same RTA processing as sample B for condensation, but a lower value is extracted, suggesting that the SCF CO₂ carrying H₂O₂ will lead to a higher content of oxygen in SiO_x, thus decreasing the refractive index.

 

Fig. 3 (a) Refractive index spectrum (500–700 nm) for samples A, B, C, and D. (b) XPS measurement against with sample depth.

Download Full Size | PDF

In order to further examine the SiO_x characteristics, XPS measurement was conducted to extract the stoichiometric ratio of the Si–O bond, where the measurement was performed by etching the layer. With the measured XPS, the effect of the Si–O bond for QWI could be examined. As can be seen in Fig. 3(b), the effect of SCF CO₂ (sample C) on the Si–O bond is limited to only a depth of 30 nm from the surface, which is due to the condensation from high pressure and temperature in the SCF processing chamber. Other than that, similar values between C and the reference (sample A) can be found, indicating no OH bond contribution. However, after RTA processing at 700 °C for 90 s, samples B and D show quite different results in comparison to the reference (sample A). The content of Si–O increases dramatically through the SCF with RTA processing (sample D); on the contrary, the content decreases by RTA (sample B). This suggests that the introduction of an OH group from the SCF with RTA can form condensation, enhancing the Si–O bond quantity. Therefore, by applying this technique, more oxygen bond-related Si compounds can be realized to improve the Ga reaction and thus QWI processing.

2. PL, EAM-integrated SOA fabrication, and characterization

In validating the enhancement of QWI through the SCF technique, a PL spectrum measurement on the MQW wafer was performed. As shown in Fig. 1, the wafer is defined by a 1550-nm InGaAsP-based MQW, sandwiched between a top 1700-nm-thick p-InP layer and bottom n-InP layer for the optical cladding layers. On top of the p-InP cladding layer, a 200-nm-thick InGaAs layer is grown for IFVD QWI. The SiO_x deposition, SCF, and RTA processing are kept the same. As the CO₂ SCF is the main carrier for introducing H₂O₂ into SiO_x, the subsequent RTA processing produces more oxygen in the Si–O groups through dehydration and condensation. Therefore, the primary mechanism for enhancing the QWI is mainly attributed to the increased oxygen content, creating more vacancies from the chemical reaction with Ga and thus improving QWI.

Figure 4(a) plots the PL measurement for three MQW samples: as-grown sample for reference, SiO_x deposition without SCF processing, and SiO_x deposition without SCF. Two RTA temperatures, 650 °C and 700 °C, for 90 s were set for comparison. As shown, the SCF treatment clearly exhibits a larger blueshift than that without SCF treatment for both temperatures. Further, a 30 and 70-nm additional blueshift at 650 °C and 700 °C, respectively, can be observed, suggesting that the reaction of the IFVD QWI can be improved through the SCF-carrying H₂O₂ to lower the RTA temperature. It should be noted that the reference samples were processed under the same RTA and SCF process, but no SiO_x cap layer was deposited. A blueshift difference within only 10 nm can be seen from the as-grown sample, suggesting that such SCF and RTA treatment can render IFVD QWI useful for PIC applications.

 

Fig. 4 (a) PL spectrum for three samples after 650 °C and 700 °C RTA processing. (b) Photos of surface morphology before and after RTA, showing no significant degradation on the whole area. (c) Top-view of EAM-integrated SOA device and two steps of QWI.

Download Full Size | PDF

In order to further verify the SCF-enhanced QWI on the PIC, a high-speed EAM-integrated SOA was fabricated, where a passive waveguide was fabricated between two devices for the optical signal connection. Before fabrication, a sample with the patterned SiO_x cap layer was tested by RTA. As shown in Fig. 4(b), no degradation in surface morphology of the whole area were observed before and after RTA, implying no significant effect of surface morphology would affect the device processing. Because of the quantum-confined Stark effect in the MQW of the EAM, a blueshift bandgap from the SOA is required to simultaneously obtain both efficient optical modulation and amplification. In addition, a maximum bandgap in the passive waveguide region between the EAM and SOA is necessary for reducing the propagation loss. Figure 4(c) shows the top view of the finished device, where the layer structure of the wafer is shown in the schematic in Fig. 1. In order to further investigate the effect of SCF treatment on the QWI and on device integration, there are three EAM-SOA devices fabricated for comparison: device I is the EAM-SOA as the reference without any other processing, device II is set with QWI processing without the SCF, and device III is controlled with the SCF-enhanced QWI. In the local bandgap engineering, as shown in Fig. 4(c), RTA and the two steps of the SiO_x cap layers defined by buffered oxide etching (BOE) are set in the QWI processing. In addition, the three sections are the EAM, passive waveguide, and SOA, respectively. Because of the SCF enhancement, the 650 °C RTA condition in sample C could obtain a near-identical blueshift at 700 °C in sample B. After defining the three bandgap sections, a straight ridge waveguide was processed for the EAM, passive waveguide, and SOA. Ti/Pt/Au and Ni/AuGe/Ni/Au were deposited for the p-type and n-type ohmic contacts. After planarization, the CPW line on Ti/Au was finally performed for the electrical feed and transmission lines.

Figures 5(a–c) present the transmission spectra of the I, II, and III devices against the bias of the EAM. The transmission was measured through the fiber insertion loss and two lens fibers were used for coupling light into and out of waveguide facets. A bias of 0–7 V on the EAM, and 10 and 50-mA of injection current on the SOA were applied for extracting the modulation depth and optical gain. The transparent current of the SOA is 10 mA. Device I exhibits a 22 and 14-dB modulation depth for 1580 and 1560 nm, respectively, while optical gains of 10 and 17 dB are found, respectively. A trade-off behavior between modulation and gain is observed, clearly attributed to the mismatch of bandgap between the EAM and SOA. In device II, although the QWI was performed to increase the bandgap of the EAM, high RTA temperature of 700 °C was necessary due to no SCF treatment, simultaneously leading to zinc atom diffusion and thus significantly deteriorating the modulation. Only a 5-dB modulation depth and 12-dB optical gain are found at all wavelengths. In device III, with the SCF-enhanced QWI, a low RTA temperature of 650 °C was used for obtaining a 40-nm blueshift. By operating the devices at 1630 nm wavelength for ensuring low optical absorption of EAM, an extra loss due to QWI was found to be around 4.71 dB/cm (device II) and 2.43 dB/cm (device III) by comparing with device I (reference), exhibiting the same order of magnitude as the work published in reference 14. A modulation depth of 17.5 dB and optical gain of 15 dB is obtained at 1565 nm, almost the largest modulation depth and optical gain at the same wavelength. In addition, an optical transmission as low as –5 dB (at 0 V) with a high >10 dB modulation depth is observed, suggesting that a low Zn diffusion and wavelength match can be attained from the SCF QWI. In Fig. 5(d), a high-speed EO response and a high bitrate of NRZ were also used for testing the EAM (device III). A bandwidth of –3 dB with a 40-GHz and 45-Gb/s open-eye response and V_pp of 2 V was found, so it can be concluded that such SCF-enhanced QWI can realize high-speed PIC fabrication.

 

Fig. 5 (a–c) Transmission spectra of devices I, II, and III, respectively, for testing the modulation depth and optical gain. (d) EO response and Eye diagram of device C at 40 GHz and 45 Gb/s.

Download Full Size | PDF

4. Conclusions

A new technique using SCF was used for enhancing the IFVD QWI. A patterned SiO_x capping layer was used for local bandgap engineering on an InGaAsP-based MQW. The temperature for the same blueshift (50 nm) on PL was reduced from 700 °C to 650 °C. FTIR accompanied with index spectroscopic ellipsometry and XPS proved the large blueshift, which was obtained from the oxygen-enhanced SiO_x film because of the high content of OH groups from SCF-CO₂ and H₂O₂. An improved photonic integration performance from an SOA-integrated EAM was also realized. In comparison to the devices with high-temperature QWI and without QWI, a lower transmission loss with a high modulation depth and optical gain was obtained. A modulation depth of 17.5 dB, SOA-integrated EAM optical gain of 15 dB, and modulation above 40 Gb/s was demonstrated in the integrated elements. It suggests that the bandgap engineering through SCF-enhanced QWI can be realized with reducing the effect of p-type zinc atom diffusion. The photonic devices under integration template could be processed at lower temperatures, thus allowing further optimization of the individual and integrated performance and enabling high thermal stability process.