Plasma sources for high-rate etching of SiC
05/01/2005
SiC has become an attractive material for the semiconductor industry for both electronic devices and microelectromechanical systems (MEMS). Typical applications for SiC etching include through-wafer vias, shallow vias, and trenches for power devices; however, they suffer from inherently slow etch rates. In this work, SiC etch rates have been compared from two different inductively coupled SF6 plasma sources for 250µm-wide deep vias etched in 6H n-type SiC substrates, with a patterned nickel mask. SiC etch rates of 2.6-2.7µm/min were obtained for one type of plasma source, making it possible to use plasma processing for MEMS applications and deep-etch electronic device processing such as wafer thinning, through-wafer via etching, and deep trench isolation.
The properties of SiC - wide bandgap, high thermal conductivity, large breakdown fields, high saturated electron-drift velocity, and the ability to withstand extreme environmental changes - have made it a good choice for both electronic devices and MEMS. As the bond energy between silicon and carbon is high, it has high chemical resistance. A practical approach to patterning SiC is to use plasma-based dry etching. However, plasma etch applications (used for through-wafer vias, shallow vias, and trenches for power devices) suffer from inherently slow etch rates, typically ≤0.6µm/min [1-5] for inductively coupled plasmas (ICP). For through-wafer via etching, this means wafer throughput is very low. Kim et al. [6] have shown that it is possible to achieve etch rates up to 1.5µm/min using an SF6 ICP.
If greater etch rates could be achieved, then other possible plasma etch applications become more desirable such as SiC wafer thinning, which is usually accomplished through chemical mechanical polishing (CMP). However, CMP is a time-consuming process that leaves undesirable residues and damages the wafer surface, which could be reduced by using plasma etching. Electronic device processing such as deep trench isolation is used to reduce the distance between high-voltage devices, reducing the total chip area and thus increasing wafer yield. This is highly desirable because only 50mm and 75mm dia. SiC wafers are used for processing.
Many different gas chemistries have been considered for the high-rate etching of SiC, such as NF3, PF3, BF3, SF6, and Cl2, with the addition of Ar or O2, or both, to some of these chemistries. The most advantageous of these gases is SF6 [4-6]. When etching SiC using an SF6 plasma, the dominant etch-limiting factor is the breaking of the Si-C bond [1], leaving dangling bonds on the silicon and carbon atoms. Fluorine radicals generated in an SF6 plasma can then combine with silicon and carbon to form volatile SiFx and CFx. Therefore, there is a requirement for the plasma source to produce a high density of ions within the plasma.
Two types of ICP sources were investigated. The work demonstrates that it is possible to achieve greater etch rates than previously reported from an SF6 ICP. Also reported are the effects on selectivity to the Ni mask and the SiC etch rate experienced when O2 is added to an SF6 plasma.
Plasma sources
Two types of plasma source have been considered in this work [referred to as plasma source 1 (PS1) and plasma source 2 (PS2)].
PS1 is a decoupled, 13.56MHz ICP reactor capable of producing a very high-density plasma. The plasma is generated in a small chamber and is allowed to diffuse into a separate process chamber where the substrate being processed is clamped to a cooled 13.56MHz RF-driven electrode.
The chamber has been designed with multipolar magnetic confinement so that the plasma is restricted within the central region of the chamber to increase the density. A source enhancer can also be used with PS1. This can further increase the radical density present at the substrate, thus possibly increasing the etch rate.
PS2 is a 13.56MHz ICP planar-type coil arrangement with multipolar magnetic confinement at the chamber sidewalls. The plasma is generated in the main chamber, providing a higher-density plasma at lower pressures. This results in reduced scattering collisions, leading to increased directionality of the charged species. Compared to PS1, PS2 will produce a higher ion density at the substrate in which 13.56MHz RF-biasing of the substrate electrode can control the independent ion energy.
Experiment
The samples used in this study were 50mm, 6H n-type SiC (Grade R), face S, with a seed layer of TiNi (100Å Ti and 500Å Ni) and a plated Ni mask of 4µm and 8µm. Samples were cleaved into smaller sections and mounted with cool grease (a thermally conductive paste) onto a 4-in. Si carrier, with polyimide tape used to cover the exposed area of the Si carrier. The features on the samples were 250µm-dia. vias. Prior to etching, the samples were exposed to a 10-min Ar sputter to remove the seed layer from the features. The process etch time was 20 min, and the samples were etched in SF6, SF6/Ar, or SF6/O2 with varying amounts of O2 addition. Some samples were used for more than one etch depending on the remaining mask thickness. Samples were etched in PS1 with and without the source enhancer fitted. Etch rates were determined by stylus profilometry (Tencor TLA P10 profilometer) and scanning electron microscopy (SEM LOE440), while mask etch rates were determined by SEM.
Results and discussion
Plasma source 1. Initial pressure-regime experiments for PS1 showed that etching of SiC in SF6 achieved the highest etch rate of 1.13µm/min at low pressures (<10mtorr). This result was achieved using a very high coil power and bias power. These experiments were carried out with the source enhancer removed. An etch rate of 1.23µm/min was attained using the source enhancer with the same process parameters. Further experiments using PS1 were carried out with the inclusion of the source enhancer.
Figure 1 shows a plot of SiC etch rates for varying SF6 flow. A maximum etch rate of 1.68µm/min at an SF6 flow of 200sccm was achieved using the same process parameters as above. Figure 1 also shows that the SiC etch rate starts to saturate at ~130sccm. Therefore, there is very little benefit to increasing the SF6 flow, because there is either insufficient coil power to promote the dissociation of SF6 and produce fluorine radicals, or they are recombining before reaching the sample surface. A further reason for the saturation in SiC etch rate could be due to the number of insufficient broken Si-C bonds on the surface of the sample to enable formation of SiFx and CFx volatiles.
 Figure 1. SiC etch rate as a function of SF6 flow rate. |
For an SF6 flow of 130sccm, the bias power was increased further by 50%; however, this had no effect on the SiC etch rate. This suggests that the Si-C bond breaking is not the rate-limiting factor. Increasing the coil power by 25% also has very little effect on the etch rate, thus showing there is sufficient coil power to promote SF6 dissociation. It is unlikely that the generated fluorine radicals would recombine before reaching the wafer surface due to the low process pressure; thus, there must be a reaction rate-limiting factor at the surface of the sample. Chabert [7] proposes that the rate-limiting factor for SiC etching is the formation of a carbon-rich layer at the surface of the sample as fluorine preferentially combines with Si atoms. Therefore, the rate-limiting factor is the C-F reaction to produce CF2 volatiles. Chabert also shows that the ions in the process are only used to break the Si-C bond and play no other part in the process.
From previous in-house experiments - although on a different system setup on a different STS system - the use of pulsed-platen power increases the SiC etch rate, which indicates that ion bombardment of the sample during the surface chemical reaction may actually be detrimental to the formation of CFx and SiFx volatiles. Further work is required to clarify this.
Finally, 15% Ar was added to an SF6 flow of 130sccm to increase the ion bombardment of the sample. No increase in etch rate was observed, contrary to what Jiang [2] reports, although this may be due to a different process regime, indicating that there is sufficient ion bombardment from the SF6 ions. Instead, there is a small decrease in etch rate, which may be due to the dilution of the SF6 etch gas.
Figure 2 shows SEM images of SiC etched using PS1 at optimal process conditions to achieve a maximum etch rate of 1.53µm/min. Figure 2a shows a 250µm-wide via etch. It is possible to see that there is trenching around the base of the via as a result of the impact of directional ions onto the surface being etched [2, 4]. Although the via has not been cleaved to give a cross-sectional image, it is possible to see that the sidewalls are relatively straight and the etch is very anisotropic. The bottom of the via looks smooth; however, there is some pitting, which may be due to sputtering of the Ni mask into the via leading to micromasking or defects in the patterned area prior to etch, resulting in stalagmite formation. Ions striking the stalagmite are deflected toward its base, forming a pit around it. Eventually the stalagmite is etched, leaving only the pit at the bottom of the via. This can be seen more clearly in Fig. 2b, where the stalagmites still present have deeper pits around them, compared to the depth of the pits where stalagmites are not present. Another explanation for the pitting at the base of the via may be due to defects within the SiC crystal lattice that lead to areas that are more selective to etching than others. The pit formation requires further analysis. The selectivity of SiC to Ni mask for an etch rate of 1.53µm/min was ~90-100:1, which is adequate for etching through a 350-400µm-thick wafer with a 4µm Ni mask.
Plasma source 2. Figure 3 shows the pressure regime experiments for PS2. While varying process pressure and fixing all other parameters, a maximum etch rate of ~2.6µm/min was achieved at a higher process pressure than for PS1. Because the optimum process pressure was higher, the bias power was increased to keep the ion energy at an optimum level. If we consider the etch rate as a function of ICP coil power (Fig. 4), it can be seen that the etch rate saturates at ~2.6µm/min. The same SF6 flow was used to optimize the process pressure.
 Figure 3. SiC etch rate as a function of process pressure. |
For an SF6 plasma, PS2 shows an increase in etch rate of ~60% when compared with PS1. If we consider the difference between the two plasma sources, PS1 generates the plasma in a smaller secondary chamber, which then diffuses into the main chamber, while in PS2, the plasma is generated in the main chamber and the RF coil is closer to the substrate, thus providing the ability for greater ion and radical densities. The increased process-pressure regime in which the plasma is generated in PS2 will also be beneficial to the radical and ion density present at the wafer surface. Increasing the probability of Si-C bond breaking and SiFx and CFx volatile formation ultimately results in an increased SiC etch rate (Figs. 3 and 4).
 Figure 4. SiC etch rate as a function of coil power. |
Results from PS1 showed that the inclusion of Ar in SF6 to increase the Si-C bond breaking and etch rate had no significant effect. Thus, Ar was not included in the etch chemistry for PS2.
It has been reported that the addition of O2 to an SF6 plasma can increase the SiC etch rate [1, 2] due to the formation of new volatile products CO, CO2, and COF2. O2 may also prevent undesirable sulfur-based deposition on the chamber walls by forming volatile SOxFy, thus helping chamber cleanliness [7]. The addition of O2 to an SF6 plasma in PS2 showed no increase in SiC etch rate, contrary to the findings Cho [1] and Jiang [2] reported. Chabert [7] also reported no increase in SiC etch rate with the addition of O2, although this work used an SF6/O2 helicon plasma. Different process regimes may account for the contrasting data. The addition of 10% O2 dramatically decreased the etch rate, but O2 greatly reduced the Ni mask etch rate, thereby increasing the selectivity of SiC over Ni. A maximum selectivity of ~45:1 occurred at 5% O2 addition without degrading the etch rate - an increase of >100%. Although fluorine from the SF6 etch gas probably reacts with the Ni mask to form nonvolatile NiF, that probably helps to decrease the Ni etch rate. The O2 must react with the Ni to form NiOx or NiSOx, which acts as a barrier to the sputter effect of the Ni mask by energetic ions.
 Figure 5. SEM images of a 20-min SiC etch from PS2 with optimized process parameters to give a maximum etch rate: a) 250µm-wide via and b) via sidewall and base. |
Figure 5 shows the scanning electron micrographs from PS2 using the optimal setting to generate the highest SiC etch rate with the addition of 5% O2 to the SF6 plasma. The smooth via base shown in Fig. 5a shows no pitting compared with the via base presented in Fig. 2 from PS1. It is unclear whether this is due to the different process regime in which PS2 operates, or if the addition of O2 reduces the sputtering of Ni from the mask into the via, causing micromasking, as discussed for PS1. The via sidewalls (Fig. 5b) are considerably smoother than those presented in Fig. 2, and there is a reduction in base trenching. Further investigation is required to determine the reduction in base trenching, pitting, and increased base and sidewall smoothness.
Conclusion
Two ICP sources have been used to etch SiC in an SF6 discharge. PS1 achieved a maximum etch rate >1.5µm/min and a selectivity of 90-100:1 (Si:Ni) for 250µm-wide Ni-masked SiC vias. PS2 achieved a maximum etch rate of 2.6-2.7µm/min, for 250µm-wide Ni-masked SiC vias. The addition of 5% O2 increases the selectivity by >100% to ~45:1 (SiC:Ni).
The SiC etch rates and profiles achieved from PS1 are equivalent to the maximum SiC etch rates for an ICP that have been published to date. SiC etch rates less than those quoted for PS1 currently are used in the industry; however, they still limit the production throughput and restrict the plasma etch steps of the device process to shallow etches of a few tens of microns. Therefore, deep SiC etching must be realized by highly concentrated wet etching or physical and mechanical processes. SiC etch rates of 2.6-2.7µm/min quoted for PS2 now make it possible to use plasma processing for MEMS applications and deep-etch electronic device processing such as wafer thinning, through-wafer via etching, and deep trench isolation. Previous in-house experiments have proved that it may be possible to further increase the SiC etch rate using an ICP source.
Acknowledgments
The authors are grateful to NASA Glenn Research Center for supplying SiC nickel patterned substrates.
References
- H. Cho, K.P. Lee, P. Leerungnawarat, S.N.G. Chu, F. Ren, et al., J. Vac. Sci. Technol. A 19(4), p. 1878, 2001.
- L. Jiang, R. Cheung, Microelect. Eng. 73-74, pp. 306-311, 2004.
- D. Gao, M.J. Wijesundara, C. Carraro, R. Howe, R. Maboudian, J. Vac. Sci. Technol. B 22(2), p. 513, 2004.
- L. Jiang, N.O.V. Plank, M.A. Blauw, R. Cheung, E. v-d. Drift, J. Phys. D: Phys., Vol. 37, pp. 1809-1814, 2004.
- P. Leerungnawarat, K.P. Lee, S.J. Pearton, F. Ren, S.N.G. Chu, J. Electron. Mater., Vol. 30, No. 3, p. 202, 2001.
- D.W. Kim, H.Y. Lee, B.J. Park, H.S. Kim, Y.J. Sung, et al., Thin Solid Films, Vol. 447-448, pp. 100-104, 2004.
- P. Chabert, J. Vac. Sci. Technol. B 19(4), pp. 1339-1345, 2001.
Kenneth Robb received his BSc (Hons.) in physics from the U. of Paisley, his MSc in biomedical instrumentation engineering, and his PhD in physics from the U. of Dundee. He is a process engineer in the Department of R&D, Surface Technology Systems plc, Imperial Park, Newport, NP10 8UJ, Wales; ph 44/1633-652400, fax 44/1633-652405, e-mail [email protected].
J. Hopkins, G. Nicholls, L. Lea, Surface Technology Systems plc, Newport, Wales