+44 (0)24 7671 8970
More publications     •     Advertise with us     •     Contact us
 
Technical Insight

SiC growth rockets with hydrogen chloride addition

The standard growth process for silicon carbide produces high-blocking-voltage devices, but suffers from long processing times. These delays are hindering commercialization, says Francesco La Via, who believes that the problem can be overcome by adding hydrogen chloride into the cell.

Rapid improvements in substrate production and advances in the growth of high-quality epitaxial films by CVD have driven the development of SiC material that is ideal for high-power microelectronic devices. This has led to the fabrication of several types of device with blocking voltages of at least 10 kV, including power depletion-type MOSFETs, implanted JFETs, PIN diodes and Schottky barrier diodes. However, to obtain these breakdown voltages, epitaxial layers 80"“100 μm thick are needed, which require processing times of 10 hours or more for conventional epitaxial growth rates of 6"“8 μm/h. These long growth times equate to high processing costs, and this is hampering the commercial development of high-voltage SiC devices.

CVD using "step-controlled epitaxy" is the standard approach for homoepitaxial growth of α-SiC, the most common form of the material. With this method the polytype is controlled through the surface steps on off-axis substrates. Mirror-like surfaces can be produced with carbon/silicon ratios of 1.4"“2.5 at a growth rate almost independent of the gas mixture. Within this regime the epitaxial growth rate is directly proportional to the silane flow rate, but at high silane flows silicon droplets are formed in the gas phase that are then deposited on the wafer.

Combining these observations with analysis of the gas-phase kinetics in the growth system has led us to conclude that epitaxial growth proceeds through silicon adsorption at atomic steps, followed by carbonization of hydrocarbon molecules. The main limitations of this process are the low growth rate that results from slow silicon species diffusion through the stagnant layer, and the limited silicon/hydrogen-gas ratio, which has to remain below 0.05% to prevent homogeneous nucleation of silicon droplets in the gas phase. Nucleation can also lead to a poor-quality surface through depletion of the gas-phase precursors that are needed for deposition.

An improved epitaxial process that overcomes the growth-rate problem has recently been developed independently by both our research team in Italy and scientists at the University of South Florida. The improvements include increasing the silane flow and introducing HCl gas into the deposition chamber, and have led to much higher growth rates combined with good surface morphology.

In our work, the 4H-SiC epitaxial films were grown using silane and ethene precursors in an LPE Epitaxial Technology ACSiM8 hot-wall reactor that has a high degree of temperature uniformity and can accommodate either six 2-inch or three 3-inch substrates.

HCl combines speed with quality

It turns out that HCl addition can significantly increase the silane concentration, while avoiding the homogeneous nucleation of silicon in the gas phase that usually occurs during the standard deposition process. With this limitation removed, high growth rates of up to 112 μm/h that are directly proportional to the silane flux are possible (figure 1). No homogeneous gas-phase nucleation has been observed, even at the very high silane concentration of 0.6% that was used for the highest growth rate. Even higher growth rates might therefore be possible.

We characterized 100 μm thick epitaxial layers grown with the HCl-based process using AFM to determine the film's surface roughness. Measurements from several regions of the wafers show that the average value is independent of growth rate and is typically 0.3 nm (figure 2). This roughness is similar to that of the standard HCl-free process.

Our potassium hydroxide etching experiments at 500 °C have also revealed that there are no major differences between the dislocation densities of the SiC epilayers produced with and without HCl addition.

We have also measured the minority-carrier lifetimes of both types of epiwafer to establish whether there are defects within the material that degrade device performance. These experiments were carried out with a modified microwave photoconductive decay instrument built by Semilab, which features a 350 nm laser excitation source. Figure 3 shows the minority-carrier lifetime distribution from a 100 μm thick epitaxial layer grown using HCl addition. The distribution peak shows that the average carrier lifetime is over 1 μs, which is a high value for SiC.

We confirmed the quality of the epitaxial layers grown with HCl addition using deep-level transient spectroscopy (DLTS) measurements. Although films produced with both types of epitaxial process each contain only one defect level "“ the EH7, which has an associated energy of 1.5 eV "“ the concentration of this defect is three times lower in the layer grown using HCl addition. The combination of the DLTS and the microwave photoconductivity decay results suggests that the faster process could also be used to produce high-quality bipolar devices and X-ray detectors.

We also know from capacitance-voltage measurements that the high growth rate process can produce low background doping levels of less than 1 × 1014 cm"“3. This means that the HCl-based process can yield intentionally doped layers with low doping concentrations and good uniformity. Figure 4 shows the doping distribution for several wafers grown with the same process. The average doping concentration is 5.6 × 1014 cm"“3, with a standard deviation of 6.7%.

The faster process also delivers extremely good thickness uniformities. Our Fourier transform infrared spectroscopy measurements of several epiwafers have shown that the standard deviation in the film thickness for layers with an average thickness of 56.8 μm was just 1.2%.

Very high wafer throughput

Although the ACSiM8 has a significantly lower capacity than the reactors used by Cree or produced by SiCED (see "Reactor comparisons"), its far higher growth rate capability means that it can deliver a higher wafer throughput for 2 and 3 inch material. In addition, the epilayer thickness uniformities produced with the LPE reactor compare very favorably with those reported elsewhere.

We have also started to investigate the performance of devices grown with the high growth rate process. We compared Schottky diodes grown on different substrates and produced by either the HCl-based process at a growth rate of around 20 μm/h or the standard process. Electrical measurements showed little difference between the leakage current distribution of both types of diodes "“ the average leakage current is 10"“5 A/cm2 at "“600 V in both devices. The forward voltage distribution at 20 A/cm2 is also very similar for both processes, indicating that both approaches give good electron mobility. In the near future we will characterize Schottky diodes fabricated with the much faster 112 μm/h growth rate.

Our new process, which can routinely produce 50"“100 μm thick epilayers with good morphology and low background doping, will provide the SiC industry with a lower cost production technique for high-blocking-voltage devices. This will aid the growing interest in these devices for a wide range of applications, including electrical power converters in ships and automobiles, and voltage transformers for electrical grids.

Further reading

D Crippa et al. 2005 Mat. Sci. Forum 483"“485 67.
M Das et al. 2005 Mat. Sci. Forum 483"“485 965.
A Elasser et al. 2002 Proc. IEEE 90 969.
J Zhao et al. 2004 IEEE Elec. Dev. Lett. 25 474.
J Zhao et al. 2003 IEEE Elec. Dev. Lett. 24 402.
Acknowledgements

Francesco La Via thanks Stefano Leone, Marco Mauceri, Giuseppe Pistone and Giuseppe Abbondanza (ETC) for SiC epitaxial growth, Giuseppa Galvagno (CNR-IMM) for Schottky diode characterization, Lucia Calcagno (Catania University) for DLTS measurements, Gaetano Foti (Catania University) for photoluminescence measurements, and Gian Luca Valenti (LPE), Danilo Crippa (LPE) and Maurizio Masi (Milan University) for helpful discussions regarding the interpretation of the results.

×
Search the news archive

To close this popup you can press escape or click the close icon.
×
Logo
×
Register - Step 1

You may choose to subscribe to the Compound Semiconductor Magazine, the Compound Semiconductor Newsletter, or both. You may also request additional information if required, before submitting your application.


Please subscribe me to:

 

You chose the industry type of "Other"

Please enter the industry that you work in:
Please enter the industry that you work in: