Abstract
InGaAs integration on Si substrates is an important topic for next generation electronic devices. Rapid melt growth (RMG) has the potential to grow defect-free lattice mismatched materials on Si at low cost. Most previous publications have focused on growing binary III–V compounds by RMG, but none have discussed ternary compound materials. In this paper, we demonstrate the RMG of the single crystal ternary compound InGaAs on Si substrates. We discuss two main issues. The first is segregation along the stripe length. An analytical model is developed to describe the segregation of In/Ga in the grown stripe and the model is compared with experimental data. The second issue is the dissolution of the Si seed region during RMG, which leads to formation of Si islands inside the InGaAs stripe. The results of this study are applicable to any compound material in which Si is soluble at the elevated temperatures required for RMG.
1. Introduction
As the scaling of Si devices has become more challenging, researchers have been searching for alternative channel materials to enhance Si capabilities [1]. InGaAs has attracted a lot of attention [2] due to its high mobility and low leakage current. Integrating InGaAs on Si [3] could allow novel functionalities like optoelectronics on low cost Si substrates. There have been several approaches to integrate InGaAs on Si, but most of them are not easily applicable to large-scale manufacturing. For instance, direct epitaxial growth of large area InGaAs on Si [4] requires a thick buffer layer to attempt to accommodate lattice mismatches. This complicates integrating InGaAs devices with underlying Si devices. Wafer bonding uses expensive III–V substrates that are not compatible with large diameter Si wafers [5]. Local epitaxial growth in small seed holes [6, 7] can help to confine defects due to lattice mismatch but this method has not yet been applied to large-scale manufacturing. RMG, on the other hand, can avoid many of the disadvantages of the above techniques by terminating the defects generated by lattice mismatch to a confined neck region near a seed window [8]. To date, many publications have successfully demonstrated high quality materials [9–11] grown on Si by RMG. However, no publications have described the RMG of any ternary compound materials such as InGaAs.
In this study, we conduct RMG experiments on amorphous InGaAs samples deposited on an insulator with a small seed window to the Si substrate and study their crystal quality by Electron Backscatter Diffraction (EBSD). The segregation behavior of In/Ga atoms and Si dissolution from the seed in the RMG stripe were measured by Auger Electron Spectroscopy (AES). The Si dissolution from the seed is similar to that shown previously for Ge stripes [12] but is exacerbated by the low eutectic melting temperature between III–V materials and Si.
2. Materials and Methods
To prepare the sample substrates, we deposited approximately 50 nm thick silicon nitride films by Low Pressure Chemical Vapor Deposition (LPCVD) on clean Si (100) substrates. Positive photoresist was then coated, exposed, and developed to pattern 4 μm × 6 μm seed windows. Plasma etching was then used to remove the top 40 nm silicon nitride film in seed windows, followed by hot phosphoric acid etch to remove the remaining 10 nm of nitride to expose the underlying Si seed. We deposited 100 nm thick amorphous In0.27Ga0.2As0.53 films by molecular beam epitaxy (MBE) and the amorphous quality was verified by in situ reflection high-energy electron diffraction (RHEED) patterns. We patterned the InGaAs films into 2 μm × 10, 25, and 45 μm stripes using the same positive photoresist process used for the seed windows followed by plasma etching the stripes with chlorine chemistries. The whole structure was then capped with a 1 μm thick LPCVD SiO2 layer. Melt growth was performed in a standard rapid thermal process tool in a nitrogen ambient. The samples were heated at a rate of 20°C/s. Once the sample temperature reached the highest annealing temperature of 1145°C, it was held for 1 second before cooling down. Cooling rates of 15°C/s, 30°C/s, 50°C/s, and 100°C/s were used. AES was used to measure the composition of different elements near the sample surface. AES measures the concentration by detecting the kinetic energy and intensity of the Auger peaks. The resolution of the AES system was around 8 nm. EBSD inverse pole figure maps were generated to visualize the crystal orientation of the sample. EBSD measures the orientation by identifying the Kikuchi bands from the diffracted electron beam from the sample surface. The resolution of EBSD is about 500 nm.
3. Results and Discussion
The annealed InGaAs stripes were characterized by Electron Backscatter Diffraction (EBSD) [13] to check their crystal quality. Figure 1(a) shows the equivalent SEM picture of the measured stripe. The dark regions in the stripe are the Si islands. Figure 1(b) is the EBSD map of an InGaAs stripe after RMG. The uniform red color of the stripe suggests that the InGaAs is (100) orientated single crystal.
(a)
(b)
(c)
(d)
(e)
(f)
Figures 1(c), 1(d), and 1(e) are the energy-dispersive X-ray spectroscopy (EDS) maps of As, Ga, and In atoms. It can be seen that Ga is heavily concentrated at the front of the stripe near the seed region and In atoms segregate towards the end of the stripe. The nonuniform distribution of In and Ga atoms suggests a segregation effect similar to SiGe RMG [12]. The misorientation distribution of the same InGaAs stripe is shown in Figure 1(f). It can be seen that most misorientation angles appeared between 0° and 9°. The maximum misorientation angle is 9°.
The segregation effects can be understood by considering the InGaAs phase diagram. InGaAs has a pseudobinary phase diagram [14], where InAs and GaAs can form a solid solution at any concentration. Figure 2(a) shows the lens-shaped phase diagram of InGaAs, which suggests that the modeling approach previously used for Si/Ge segregation can also be applied to InGaAs. We develop an analytical solution similar to the SiGe RMG case [12] to calculate the segregation for any given initial concentration.
(a)
(b)
The differential form of the Scheil equation [15] is adopted here rather than the Scheil equation itself, since the segregation coefficient is not a constant. Equation (1) shows the differential form of the Scheil equation. represents the average liquid concentration in the melt and is the solid concentration at the liquid/solid interface during crystallization. is the fraction of the stripe.Assuming that equilibrium exists at the liquid/solid interface, the average liquid concentration at any time can be represented by the solid concentration at the growth front in the following equation:The variable segregation coefficient is represented by Ga concentration at the solidus line using linear regression as shown in Figure 2(b) and in the following equation: By inserting (2) and (3) back into formula (1), the differential equation can be solved as follows:
Equation (4) is the analytical solution of the complete mixing model, where is In concentration in the first solidified part of the stripe.
Figure 3 shows a comparison between the model and AES measured data. It can be seen that the complete mixing model (4) agrees with the experimental data. Figure 3(a) shows that the composition profiles are the same for different stripe lengths when the data is plotted against a normalized stripe length, as predicted in (4). Figure 3(b) shows the composition profiles of 25 μm long stripes at different cooling rates. All cooling rates result in the same profiles, which suggests that, for short stripe lengths, the profile cannot be modulated by cooling rates. The variation of the data points is likely caused by the Si islands inside the stripes.
(a)
(b)
Earlier in the paper, in Figure 1(a), we observed small dark regions embedded in the InGaAs stripes. A higher resolution SEM image of one such dark region and AES measurements are shown in Figure 4. Point 1 is measured in the brighter part of the stripe and is seen to be InxGa1−xAs. Point 2 is measured in the darker part of the stripe and is seen to be purely Si. This indicates that, for the experimental conditions used in this study, a large amount of Si dissolved from the seed has precipitated into Si grains. The Si signal detected at point 1 is from residual SiO2.
The reason for the significant Si dissolution can be seen in Figure 5, which is a typical phase diagram for III–V materials and Si. Per the phase diagram, Si and GaAs are completely miscible in the liquid phase. Thus, Si from the substrate will dissolve in the molten GaAs during RMG. During subsequent cooling, due to the low solubility of Si and GaAs in the solid phase below the eutectic temperature, the dissolved Si will phase-segregate and crystallize as a second phase. The Si islands can obviously adversely affect the electronic properties of the III–V material and the crystal quality. Our work on eliminating these Si islands from III–V stripes produced by RMG will be described in a subsequent paper.
4. Conclusion
In summary, we demonstrate the rapid melt growth of single crystal InGaAs. We studied the segregation effect of In and Ga atoms and developed a complete mixing model that agrees well with the experimental data. Si islands were found in the grown InGaAs materials as a result of eutectic melting and precipitation of Si which needs to be eliminated if RMG III–V materials are to be used to fabricate high performance devices.
Competing Interests
The authors declare that they have no competing interests.
Acknowledgments
This work was performed in part in the Stanford Nanofabrication Facility which is supported by the National Science Foundation through the NNIN under Grant ECS-9731293 and in the Stanford Nano Shared Facilities (SNSF) at Stanford University. K. Barton and the Department of Plant Biology, Carnegie Institution, are thanked for SEM access, and J. M. Warren, Stanford University, is thanked for EBSD access. This work was supported by Intel Corporation through the SRC MSR-Intel Liaison Program.