The effect of pumping methods on wet etching processes
JUNG-SOO LIM, R. PRASANNA VENKATESH, and JIN-GOO PARK, Dept. of Materials Engineering, Hanyang University, Ansan, 426-791, Korea
The effect of pumping methods on the etching of PE-TEOS wafers was investigated. The experiments were conducted in conventional wet bath and single wafer process tools, measuring etch rate, etch rate uniformity, and wafer-to-wafer uniformity, for various pumping methods.
Etching is one of the most important processes in semiconductor manufacturing industries. One such example is B-doped poly etchback process where silicon dioxide, PE-TEOS on wafers are etched, with the target etch depth of 500?? using BOE (Buffered Oxide Etchant, mixture of HF and NH4F) as etchant. This process requires pumping to circulate and supply the etchant to the wafer's surface.
However, no studies have been reported in the literature regarding the influence of pumps on the performance of etching processes of wafers. In this present study, the performance of a magnetic levitation centrifugal pump (MLC-BPS 600, Levitronix) is evaluated and compared with the traditional diaphragm pumps (D1 and D2) for an etching process. The D1 has low pulsation intensity and high frequency of pressure oscillations compared to that of D2.
The etching tests were conducted using 8" PE-TEOS wafers and dilute hydrofluoric acid (DHF) as an etchant. In the conventional wet bath tool, the following conditions were employed: the concentration of DHF was 0.5 wt% and the process time was 10 min. The conditions were chosen to simulate semiconductor etching conditions. In general, etch rate normally changes with the position of the wafer in the bath. Thus, in the present study, five wafers including two dummy wafers were placed simultaneously in the etching bath for each test run and the wafer position in the etching bath was labeled as left, center and right. In the single wafer tool, the following conditions were employed: the concentration of DHF was 1 wt% and the process time was 3 min. The feed was injected at the wafer center for the three pumps. However, the spreading of chemicals on the wafer depends on the pumping method. As seen in Fig. 1, the chemicals are spreading over a fairly larger area of wafer in the case of diaphragm pumps, especially for D2 due to the higher pulsation intensity (30??6 psi) when compared to that of MLC pump (30 psi) and D1 pump (30??3 psi). Thus for the MLC pump, the experiments were conducted in an additional mode called swing mode where the chemicals are supplied using a movable arm in such a way that the chemicals are spread over the same area as with D2.
FIGURE 1. Schematics showing the spreading of chemicals over the wafer area for (a) MLC pump (b) diaphragm pumps and (c) MLC pump in swing mode. Note: The injection point is center for MLC pump, -10 to 50 mm (center is taken as zero) for D1 pump and -10 to 80 mm for D2 and MLC pump in swing mode.
The etch rate is calculated by measuring the thickness of PE-TEOS before and after the etching test. The thickness is measured by an inspection tool (K-Mac reflectometer, ST 4000, Korea) which is based on the principle of spectral reflectance. The thickness is measured at 33 specified locations on the wafer and the average etch rate and uniformity are calculated from these values.
Etching test in conventional wet bath tool
The etch rate of PE-TEOS oxide in 0.5 wt% DHF by the three different pumping methods is shown in Fig. 2. The average etch rate is 90 ?? 5 ??/min for the MLC pump, 85 ?? 5 ??/min for D1 and 95 ?? 6 ??/min for D2, respectively (i.e., the etch rate is higher for D2 pump followed by MLC and D1 pumps, which show the lowest values). The high etch rate in the D2 pump may be attributed to its higher pulsation intensity. However, in the D1 pump, although the pulsation intensity is higher than MLC pump, the frequency of pump pulsations is higher which makes the etch rate more non-uniform and hence the average etch rate observed is lower.
FIGURE 2. PE-TEOS etch rate in 0.5 wt% DHF for various pumps.
The within-wafer etch uniformity is lower for the MLC pump when compared to D1 and D2 pumps as shown in Fig. 3. In addition, the value of standard deviation is also relatively lower for the MLC pump when compared to the other two diaphragm pumps. This clearly says that flow behavior has a strong influence on etch uniformity. Since the flow is continuous and smooth in the MLC pump, the spreading of etchant on the wafers is more uniform which results in a lower within-wafer uniformity. This suggests that the MLC pump is more suitable for wet etching processes in semiconductor industries. Among the diaphragms, the etch rate uniformity is higher for D1 pump because of its high frequency of pressure pulsations as previously mentioned.
FIGURE 3. Etch rate uniformity for various pumps.
In general, the etch uniformity strongly depends on position of the wafers in the etching bath. Hence, the change in the etch rate as a function of position of wafers in the bath for the three pumping methods is observed as shown in Fig. 4. For both MLC and D2 pumps, the etch rate is higher for the wafer positioned in the left side of the bath when compared to that of wafers in centre and right side. Since the feed is injected at left side of the bath and the flow is from left to right side, the pressure would be higher at left side which may result in higher etch rate at left side of the bath. However in the D1 pump, the trend is opposite and the reason for this is not clear.
FIGURE 4. Etch rate for various pumps at bath position.
Wafer-to-wafer uniformity was also calculated and shown in Fig. 5. The value observed for MLC pump is 14% which is relatively low compared to that of 26% for D1 and 22.5% for D2 pumps. Thus, wafer-to-wafer uniformity is also strongly influenced by pressure variations in the pump. Since the pressure is constant with time due to the smooth flow, this does not affect the wafer-to-wafer uniformity. While in the case of other two pumps, the values are slightly higher as both the pumps exhibited larger pressure pulsations. Very low wafer to wafer uniformity values ( <3-5% in conventional wet bath system) could be achieved in the semiconductor fab tool by ensuring laminar flow on each point of every wafer and by eliminating any eddies or dead zones using uniform flow control system and with careful tool design. However, within the limitations of our tool model, lower uniformity values could not be achieved.
FIGURE 5. Wafer to wafer uniformity for various pumps.
Etching test in single wafer tool
The etch rate of PE-TEOS wafer in DHF solution for various pumping methods is shown in Fig. 6. The etch rate is 210 ??/min for the diaphragm pumps and, 215 ??/min and 212 ??/min for MLC pump when operated in normal mode and swing mode respectively. So the etch rate in single wafer tool is higher for the MLC pump unlike in conventional wet bath tool.
FIGURE 6. PE-TEOS etch rate in 1 wt% DHF for various pumps
The etch rate uniformity is lower for the D2 pump among the three pumps when they were operated in normal mode, as shown in Fig. 7. This might be because, in the D2 pump, the chemicals are spreading over a relatively large area of the wafer due to the high pulsation intensity (30??6 psi) compared to that of MLC (30 psi) and D1 pump (30??3 psi). In order to confirm this, we operated the MLC pump in swing mode such that the chemicals were spread on same area of the wafer as in the D2 pump. As expected, MLC pump operated in a swing mode shows a much lower value because of the additional impact from non-fluctuating (non-pulsative) flow.
FIGURE 7. Etch rate uniformity for various pumps.
Fig. 8 shows the wafer-to-wafer uniformity for all the three pumps. Like etch rate uniformity, the wafer- to-wafer uniformity is also higher for the D1 pump owing to the high frequency of pump pulsations. There is no significant difference between the D2 and MLC pumps. In short, the MLC pump operated in swing mode is preferred for single wafer etching process as it shows high etch rate with lower etch rate uniformity.
FIGURE 8. Wafer to wafer uniformity for various pumps.
Etching experiments using three different pumps (a MLC pump and two diaphragm pumps, D1 and D2) were conducted. In a batch type bath, the etch rate is higher for the D2 pump, and lower for D1. For MLC pumps, the value is in-between. However within-wafer uniformity is higher for D1 pump followed by D2 and MLC pumps owing to the high frequency of pump pulsations. Both within wafer uniformity and wafer-to-wafer to uniformity are lower for MLC pump when compared to the other two pumps. This is mainly because of continuous and smooth flow exhibited by the MLC pump. In a single wafer tool, the etch rate is higher for MLC pumps when compared to diaphragm pumps. Also, the etch rate uniformity is lower for MLC pumps when operated in a swing mode. Thus, the lower etch rate uniformity could be achieved with MLC pump both in conventional wet bath and single wafer tool without affecting the etch rate.
Suggested additional reading
1. F.C. Chang, S. Tanawade and Rajiv Singh, Effects of stress-induced particle agglomeration on defectivity during CMP of low-K dielectrics, Journal of The Electrochemical Society, 156 (1), H39-H42, 2009.
2. R. P. Venkatesh, J-S. Lim and J-G. Park, Random yield loss during wafer cleaning, Solid State Technology, 54(4), 16-18 &23, 2011.
Jung-Soo Lim received his BS in chemical Engineering and MS in metallurgy and materials engineering from Hanyang U., Korea and is a PhD candidate at the university. R. Prasanna Venkatesh received a bachelors' degree in chemical engineering from Bharathidasan U., India and a masters' degree in petroleum refining and petrochemicals from Anna u., India; he received his PhD in chemical engineering from the Indian institute of technology madras, India and is currently a post doctoral candidate at Hanyang U., Ansan 426-791, Korea. Jin-Goo Park received a BS in metallurgy and materials engineering from Hanyang U., Korea, and MS and PhD degrees in materials science and engineering from the U. of Arizona. He is a professor in the Department of Materials Engineering as well as director of the Micro Biochip Center and the Nano-bio Electronic Materials and Processing Lab. (NEMPL), at Hangyang U. ph.:82-31-400-5226; email email@example.com
Solid State Technology, Volume 55, Issue 3, April 2012