Reducing Time in Batch Cleaning

Cleaning is the only remaining class of process steps still dominated by batch product technologies, including immersion and spray. Use of single wafer techniques in other process areas has delivered improved process performance and productivity. Because of this, the industry is eager to use single wafer equipment to realise the same benefits in surface preparation. Analysis of advanced, 90-nm semiconductor device process flows show that out of an approximate total of 600 process steps, about 100 are cleaning steps.

Much effort has been placed on completely changing wellestablished and highly reliable batch cleaning processes with new and unproven single wafer chemistries and equipment [1- 4]. As an alternative approach, strategies have been developed to achieve a 50% cycle time reduction from the traditional batch immersion process without significant changes to process chemistries or sequences. This approach also results in a significant reduction of equipment footprint and in DI water usage. Also, a significant reduction in the cycle time of batch spray cleaning processes has been achieved.

Throughput must also be considered when deciding which process tools to use on the manufacturing floor. Shorter cycle time batch processes can continue to provide high throughput and superior performance while reducing the overall production cycle time. At first glance, one might assume a simple relationship between cycle time and throughput. However, the time required to batch and un-batch wafers in a batch system and the time to transfer wafers in a multi-chamber single wafer system can overlap the process time of other batches or wafers. Therefore, cycle time is calculated based on "FOUP-to-FOUP" operations, or the time required to process a given number of wafers completely through the tool, and throughput has been calculated based on "continuous" operations, assuming full loads.

"FOUP-to-FOUP" calculations are straight forward for batch systems. For single wafer systems the relationship between the number of chambers, number of wafers, and the process time needs to be represented properly to account for the waiting time after the last chamber has been loaded and before the first chamber is unloaded. Different considerations are necessary for calculating "continuous" operation throughput for single wafer, batch spray, and batch immersion. In batch spray, the wafers are processed in a single chamber, whereas in batch immersion the wafers will be processed in 2 to 7 stations.

Single wafer system
Single wafer systems which spin the wafer while dispensing chemicals are being considered in this work. The main parameters for cycle time and throughput calculations for the single wafer system include the number of chambers, number of wafers, initial preparation time, wafer transfer time and process time. Figure 1 is a schematic representation of the timing for processing 7 wafers in a 4-chamber single wafer cleaning system. The cycle time for processing n wafers is given by
i + f(2t + p) + rt + (p+t),
if n <= c then f = 0 and r = n
if n > c, then f = integer(n/c) and r = remainder(n/c)
if n = integral multiple of c, then f = integer(n/c)-1 and r=c where
i = initial preparation time
t = wafer transfer time
p = process time
f = number of "full chamber" cycles
r = remaining wafers after completion of "full chamber" cycles.
n = number of wafers being processed
c = number of single wafer chambers in the system.

For these calculations it is assumed that the process time is always long enough that there will be at least some waiting time after the last chamber is loaded and the first chamber is unloaded. It is also assumed that each chamber is identical, dispensing multiple chemical sequences. i.e. each wafer visits only one process chamber.

To calculate throughput for the single wafer system in continuous operation it is assumed there are sufficient load ports so wafers are always available for transfer and the initial preparation time will overlap processing time. In order to properly represent the throughput, calculations are carried out for n=kc, where k is an integer. Therefore, throughput for the single wafer system is given by
kc/[(k-1)(2t + p) + ct + p+t]
The ultimate throughput is calculated for k‡(, which gives c/(2t + p)

Batch system
In a batch spray system the wafers are batched and then transferred into the process chamber. A single chamber is used to dispense multiple chemical sequences in the batch spray system. The main parameters for cycle time and throughput calculations for the batch spray system include the number of wafers, initial preparation time, wafer transfer time and process time. In addition, batch loading and unloading time is also included The cycle time for processing n wafers in a batch spray system is given by
i + nt + b + p + b + nt
where b is the batch loading time and the other symbols are the same as for the single wafer system; however, the values for these parameters may be different for the batch spray system. The batch spray system under consideration here can process a 50-wafer batch.

To calculate throughput for the batch spray system in continuous operation it is assumed that there are sufficient load ports so wafers are always available for transfer and that process time is long enough that batching (nt) can be completed before the previous process is completed. Therefore the throughput for the batch spray system is given by 50/(2b + p).

In a batch immersion system the wafers are batched and then transferred through a sequence of process stations. The main parameters for cycle time and throughput calculations for the batch immersion system include the number of wafers, initial preparation time, wafer transfer time and batch loading time. Process time may be different for each station. In addition, bath immersion systems incorporate a FOUP buffer, so FOUP loading time needs to be included. The cycle time for processing n wafers in a batch immersion system is given by
F + i + nt + b + p1 + p2 + . . . + pj + b + nt + F
where F is the FOUP loading time, pj is the process time in the jth station and the other symbols are the same as before; however, the values may be different for the batch immersion system. The time to transfer the batch between process stations is included in each pj. The batch immersion system under consideration here can process a 50-wafer batch.

The sequential nature of the batch immersion system affects the calculation of throughput. Batches cannot be moved to the next process station until the previous batch is removed from that station. Therefore, the station with the longest process time will determine the overall throughput. This is sometimes referred to as the TACT time. It is assumed that there are sufficient load ports so batched wafers are always available for transfer and that the TACT time is long enough for batching (nt) to be completed. Therefore the throughput for the batch immersion system is given by 50/(pmax) where pmax is the longest process time in the sequence.

The first cleaning process to be analyzed is the critical cleaning process in the front end. This process is typically carried out before thermal processes and before the gate dielectric is grown or deposited. Its purpose is to remove particles, organic, metallic, and native oxide contaminants. Critical cleaning has been demonstrated on single wafer systems, but has not yet found widespread application in manufacturing [3,4]. This is because the chemical sequences used for critical cleaning in traditional batch immersion systems are not easily adapted to the short process times required in the single wafer system. For comparison, we will use the 2.5 minute single wafer system process time claimed by Funabashi et al. [4].

A key contributor to the long cycle times in traditional wet benches is the single function bath approach. If each chemical, rinse, and dry step is done in a dedicated tank, a 5-7 tank sequence will be required. As illustrated in figure 2, a traditional cleaning system using SC1 (a mixture of NH4OH, H2O2 and water) , SC2 (a mixture of HCl, H2O2 and water), and HF requires a 7 tank process sequence. Long process step times in dedicated process tanks and idle time waiting for serial process tanks to become available, The efficiency of the critical cleaning process on the batch immersion system was improved by first, combining multiple chemistry steps into a single tank and, then optimizing mixing and rinsing within the tank to achieve good uniformity, efficient chemical switch-over and efficient rinsing. This strategy has been used to reduce the 7- tank process to a 2-tank process with equivalent cleaning performance using the same chemicals as shown in Figure 2.

To optimize rinsing in the multi-function tanks, a new method was used to examine rinsing efficiency at the wafer surface. Figure 3 shows a rinsing comparison between a traditional tank and an optimized multi-function tank. Cycle time and throughput calculations for the traditional batch immersion, the improved batch immersion and the single wafer systems were carried out using the parameters shown in Table 1. Most of the single wafer systems in production are 4- chamber systems. Recently, 8-chamber systems have been introduced. Both configurations are being modeled here.

Figure 4 shows the results of the cycle time calculations for the critical cleaning process as a function of the number of wafers. The improved batch immersion system is able to process wafers in much shorter cycle time than traditional batch immersion, beating that of the 4-chamber single wafer system for larger lot sizes.

Throughput calculations for these four systems are shown in Table 2. It can be seen that while the 8-chamber single wafer systems have shown significant improvement in overall throughput, the batch immersion system still provides significantly higher throughput. The improved batch immersion system provides greatly reduced cycle time with only a small drop in throughput.

Post etch residue removal for Back End
In the second example, the single wafer system is compared to the batch spray system for interconnect post etch residue removal. Chemical manufacturers have made significant progress in developing new chemical systems that work quickly for single wafer systems. In the past, residue removal chemistries have required 10 to 20 minutes to achieve good cleaning performance. Newer formulations require only a few minutes. While these newer formulations are difficult to control in a batch immersion system, they have been successfully implemented in the batch spray system. For both single wafer and batch spray systems chemical exposure time is typically about 2 minutes. Rinsing and drying for the single wafer system usually requires about 1 minute. In the past, rinsing and drying in the batch spray system required almost 12 minutes. However, recent hardware and process improvements have reduced the rinsing and drying time to 7 minutes Cycle time and throughput calculations for the traditional batch spray, the improved batch spray and the single wafer systems were carried out using the parameters shown in Table 3.

Figure 5 shows the results of the cycle time calculations for the residue removal process as a function of the number of wafers for the batch spray system and for both a 4-chamber and the newer 8-chamber systems. The batch spray process can achieve shorter cycle time than a 4-chamber single wafer tool for lot sizes greater than 20 wafers. Cycle time is only slightly higher in the batch spray tool when compared to the 8- chamber single wafer tool.

Throughput calculations for these four systems are shown in Table 3. It can be seen that while the 8-chamber single wafer systems have shown significant improvement in overall throughput, the batch spray system still provides higher throughput. Advances in single wafer cleaning equipment and processes have made it more viable for semiconductor manufacturing. However, recent improvements in batch equipment and processes have provided significant reduction in cycle time while still providing high throughput. Short cycle time batch tools should be considered in situations where both rapid processing of small lots and efficient processing of high volume products is desired.

[1] H. Kruwinus and H. Oyrer, "Better Productivity and Faster cycle Times Through Single-Wafer Processing," Semiconductor Fabtech, 12th Edition, 299(2000).
[2] A. Hand, "Wafer Cleaning Confronts Increasing Demands," Semiconductor International, Vol v, No n, pp(2001).
[3] S. Verhaverbeke et al., "Single-Wafer, Short Cycle Time Wet Clean Technology," Semiconductor International, Vol v, No n, pp(2002).
[4] M. Funabashi et al., "Realization of single wafer wet cleaning in the 300mm mass production line," Proceedings of the ISSM 2002, pp(2002).

Reprinted with permission from ASMC Proceedings 2004 - IEEE/SEMI Advanced Semiconductor Manufacturing Conference. © 2004 IEEE.