Compact processing equipment sees rise in wet-process arena
The drive for increased integrated circuit functionality and performace places growing demand on the processes and the equipment involved.
By Steven Verhaverbeke, Steve Bay and Chris McConnell
In the wet processing domain, a worldwide trend toward more compact processing equipment as well as more compact processes is observed. The driving force for the move from 200 mm to 300 mm wafers is the increase in productivity. Reduction in cost per unit
Si surface area when going from 200 mm wafers to 300 mm wafers is achieved because the capital cost does not scale in a 1:1 ratio with the increase in surface area. Capital cost is expected to increase only marginally for 300 mm tools as compared to 200 mm tools, but the processed area is increased by a factor of 2.25. It is often assumed that consumables cost is directly proportional to area and therefore, consumables cost per Si surface area would remain unchanged by the transition from 200 mm to 300 mm.
Water and chemical consumption
Two basic designs of wet processing systems are commonly used in the semiconductor industry for precision cleaning: immersion benches with multiple baths and rinsing tanks, and single-bath processors such as Full-Flow, with a single processing vessel, as shown in Fig. 1.
Immersion tanks typically have relatively large tank volumes and 200 mm wafers are spaced at 6.25 mm in a full load. In contrast, single-bath processors are designed to have small vessel sizes and the wafer pitch is one-half that of an immersion tank.
Water and chemical consumption are roughly proportional to bath or vessel volume, whether it be the rinsing bath or the chemical process bath in the case of multiple bath systems or the processing vessel in the case of single bath systems.
Rinse tank volume is clearly linked to wafer diameter; but, it is also highly dependent on the wafer pitch.
For the coming conversion to 300 mm wafers, not only the effects of wafer diameter, but also the often-ignored impact of wafer spacing must be considered when evaluating increased material and water consumption.
The important number is the volume per Si surface area which is proportional to the consumables cost per Si surface area. The consumables cost per Si surface area is proportional to the pitch between the wafers.
The pitch is not going to decrease when going from 200 mm to 300 mm.
It is very important to make 300 mm equipment efficient in DI-water consumption, because when pitch increases, the DI-water consumption per Si surface area will increase dramatically.
This article compares 200 mm and 300 mm wafers between single-bath processors and conventional immersion systems for a full Sulfuric/Ozone + HF + SC1 + SC2 cleaning process. The comparison data is given for systems running at 80 percent utilization. Further we will show that although the differences in cost-of-ownership are not extremely outspoken for 200 mm equipment, the Cost-Of-Ownership for conventional immersion systems is going to be unacceptable for 300 mm processing and is going to nullify the productivity gains obtained by the increase in silicon area. Single-bath processors, on the other hand, are going to pass the productivity increase entirely to the semiconductor manufacturer due to increase in silicon area.
The data for single-bath processors is taken for actual Sulfuric/Ozone-HF-SC1-SC2 processes in an existing 200 mm system and from a 300 mm system in prototype development. The conventional immersion data for 200 mm is data taken from an actual commercial 200 mm immersion wet station presented by F. Tardiff [1]. The data as shown for a 300 mm immersion system is obtained by scaling the 200 mm system, using the same design for 300 mm as for 200 mm, i.e. the 200 mm system is scaled with wafer diameter and wafer pitch.
In our comparison, we use an analysis for the cost of DI water by Michael Lancaster of Rose Associates [2]. The total cost of DI-water per 1,000 (quantity 1) varies between $3.10 and $4.89. A major part of this cost is the capital cost of the DI-water installation.
Typical DI water installations for a new large wafer facility with the capability for an output of 5,000 wafers per month has been $12 million [2].
In the analysis of cost of water per wafer (and per cleaning sequence) of Sulfuric/Ozone + HF + SC1 + SC2, we will assume the lowest cost of $3.10/1,000 (quantity l).
The volume of the process bath or vessel scales with A*p (Area* pitch). For conventional immersion systems, this means that with the current design, the volume of the bath is going to scale with a factor 3.6.
In the case of single-bath processors, the wafer diameter will increase from 200 mm to 300 mm, but the pitch will actually decrease. The pitch in 200 mm single-bath processors is 3.125 mm and for 300 mm single-bath processors, the pitch will be 2.5 mm.
Therefore, in single-bath processors the volume will scale with 1.8.
This will lead to lower water consumption per Si surface area for 300 mm than for 200 mm in single-bath processors.
The result of the comparison is shown in Table 1. In Table 1, the following parameters are shown for a Sulfuric/ HF/SC1/SC2 cleaning sequence: the DI-water consumption per wafer, the DI-water cost per wafer, the DI-water cost per area of Si processed, the chemical usage per wafer, the chemical cost per wafer and the chemical cost per area of Si processed.
As the data in Table 1 indicates, the water consumption in standard wet benches scaled for 300 mm wafers approaches unacceptable levels.
Chemical consumption in the case of single-bath 200 mm system for a full Sulfuric/Ozone + HF + SC1 + SC2 cleaning cycle is about 20.4 ml/wafer (the sum of all chemicals). In the case of a conventional 200 mm immersion system, the typical chemical consumption is 17.6 ml/wafer for the same cleaning cycle [1]. This is surprising considering that for HF, SC1 and SC2 single bath processors use one-pass chemistry, whereas multiple bath processors use multiple pass chemistry. The fact that the numbers are comparable is due to the very small volumes of the Full-Flow single-bath processor.
Therefore, now, we can also estimate the effect on chemical consumption when transiting from 200 mm to 300 mm.
As the cost for the chemicals, $4.68/liter was used.This is a worst-case scenario, since most chemicals are less expensive.
From Table 1, it is clear that the DI-water cost is going to be the dominating consumables cost in wet cleaning of 300 mm wafers. Chemical cost is only a fraction of the DI-water cost.
Footprint
The average cost-per-square-foot of cleanroom for the typical fab is approximately $3,000 [3].
A Dual Vessel 8100 Full-Flow system has a throughput of about 250 wafers/hour for a HF + SC1 + SC2 cleaning cycle. For a reduced cycle such as SC1 + SC2 or HF, the throughput approaches 400 wafers/hour.
The footprint of a Full-Flow system with that throughput is 6.1 m2. This includes all the supporting modules which have to be placed in the cleanroom. In this 6.1 m2, the actual footprint of the process vessel itself is 0.1 m2 (2 process vessels of each 0.05 m2).
For single-bath processors, both the throughput and the footprint remain unchanged when switching from 200 mm to 300 mm wafers, since only the processing vessel is scaled, which is only a minor part in the total footprint of the system. Both 200 mm vessel and 300 mm vessels fit in the same cabinets, and therefore, the footprint does not increase. This represents a footprint cost both for 200 mm and 300 mm Full Flow systems of $196,800. This is shown in Table 2.
The typical footprint of a multiple bath immersion system for 200 mm wafers with all its supporting modules is of the order of 22 m2 for HF-SC1-SC2. This represents a cleanroom footprint cost of $721,600.
Each bath has a process footprint of about 0.1 m2. There are typically about 8 to 10 different process baths in a conventional immersion system, bringing the total processing footprint to about 1 m2.
For the transition to 300 mm, the process footprint area of multiple bath immersion systems scale roughly with diameter*pitch.
This means that multiple bath immersion systems` process footprint will scale roughly with a factor 2.4 when transiting from 200 mm to 300 mm wafers for the same process and the same throughput. It is expected that the overhead will not directly scale with process footprint and therefore we have assumed that the overhead area will only increase with 50 percent in the case of conventional immersion systems. This is a very optimistic assumption since the volumes to be handled are actually going to increase with a factor 3.6.
The impact of this on cleanroom footprint cost is shown in Table 2.
As was the case with DI water consumption, the footprint cost for 300 mm-scaled conventional wet bench equipment approaches unacceptable (that is, uncompetitive) levels. The footprint is that of a conventional 250 wafer/hour, 200 mm system for the HF-SC1-SC2 process. The 300 mm footprint is obtained by scaling the 200 mm system with diameter*pitch for the process footprint and a 50 percent increase in overhead, i.e. assuming a 300 mm immersion system using the current 200 mm system design.
Exhaust
In current 200 mm systems, already 50 percent of the whole fab exhaust is used for conventional wet immersion systems.
The exhaust/m2 wet bench is approximatively 40 m3/min. The area exhausted at this rate is the process area and the process surrounding area. The supporting modules do not have a large exhaust. Of a total footprint of 22 m2 in a current conventional immersion system, about 12 m2 is fully exhausted at laminar flow rates. This is shown in Fig. 3. The total exhaust for a current 12 m2 immersion system is about 480 m3/min.
This is cleanroom air which is exhausted and not reclaimed. Therefore, this exhaust must be matched by cleanroom air make-up. The cost of cleanroom air make-up for 1 m3/min is about $2.13/day [4]. The exhaust will scale with process footprint and therefore, the exhaust will scale with diameter*pitch.
This means that the exhaust volumes for 300 mm wet immersion stations will be on the order of 1,152 m3/min.
A single-bath system only requires exhaust of the cabinets, since the process itself is completely contained. The single-bath exhaust totals for the entire system about 7.6 m3/min.
The annual cost for wet station exhaust is shown in Table 3.
Like DI water consumption and footprint cost, the cost of process exhaust increases rapidly to impractical levels. For 300 mm wet processing, it becomes imperative to completely isolate the process fluids from the cleanroom and the cleanroom personnel.
Throughput
The throughput of Full-Flow processors for 200 mm production is already approaching 400 wafers/hour for certain processes and is exceeding the throughput of conventional multiple bath immersion systems for 200 mm. For single-bath processors, there is a continuous effort in shortening the recipe time, which will result in continuous increasing throughput. Therefore, it is expected that the throughput advantage for single-bath processors is going to be enlarged, certainly when taking into account the longer robot transit times for multiple bath immersion systems when going to 300 mm wafers.
Process control
For wet processing of 300 mm wafers, the control of dissolved gases becomes critically important.
The control of dissolved gases has two aspects — preserving the process consistency and preserving the chemical cleanliness of cleanroom air. At first, it becomes important to control the dissolved gases inside the process liquid in order to assure the process result necessary for 0.25/0.18 µm integration levels. Secondly, dissolved gases in the process liquid can evaporate in open tank systems and contaminate the cleanroom.
Ti or TiN stripping applications with SC1 is a typical example of a process which is strongly affected by dissolved gases, and is a typical example in which it is important to contain the SC1 solution.
Other examples where dissolved gases in an open tank change the process include oxygen from the cleanroom dissolving in HF or in rinse water and etching the Si surface. It is well known that Si is etched in water containing dissolved oxygen. Recently it was also shown that dissolved oxygen in HF solutions causes Si etching. Finally, several studies have reported that dissolved gases are an important parameter in the efficacy of megasonics in cleaning processes.
HF vapors from immersion wet benches can attack HEPA filters and cause boron contamination, which can cause counter doping on the wafers. NH3 evaporation from wet immersion systems can cause a multitude of problems in submicron manufacturing, such as degradation of deep UV-photoresist, degradation of stepper lenses and finally NH3 vapors are also a major source of time dependent haze on wafers. HCl vapors also cause corrosion of equipment in the cleanroom.
Drying
Drying wafers becomes more problematic with wafer size. Moreover, drying patterns on wafers become more problematic with decreased pattern size [5]. Spinning wafers to dry will become increasingly difficult as the water has to be transported over longer distances to remove the water from the center of the wafer. Moreover, by using spin-drying it will be extremely difficult to control the marginal capital cost increase which is necessary for realizing the increased productivity when switching from 200 mm to 300 mm. Conventional IPA vapor drying has similar problems in the sense that in this drying technique water has to be transported over longer distances. On the other hand, Direct-Displace Drying as used in single-bath processors strips the water from the wafer when lifting the wafers out of the rinse tank, and therefore, this technique is independent of wafer size. On top of that the capital cost is virtually independent of wafer size. Finally, the Direct-Displace Drying technique is the only known drying technique which can completely displace water out of fine patterns on wafers and completely prevent watermark formation [5].
Etch uniformity
Etch uniformity in conventional immersion systems is mainly caused by the FILO principle (first in last out) and by the transfer from the etching tank to the rinsing tank.
Both mechanisms have even more negative impact as wafer diameter increases. The wafers are getting larger and the transfers from etching tank to rinsing tank will take more time, due to increased diameter and heavier loads. This will make the uniformity on 300 mm wafers worse than on 200 mm wafers for conventional immersion systems.
Single-bath processors do not suffer from this wafer diameter effect, since they employ the FIFO principle (first in first out). In addition, there is no transfer from an etching tank to a rinsing tank, and there is no degradation in uniformity when going from 200 mm to 300 mm wafer sizes when using single-bath processors. Every point on the wafer sees exactly the same concentration profile during an etch. A typical concentration profile is shown in Fig. 2. The fact that every point of the wafer sees the same wafer profile makes this type of wet processing independent of wafer size. This means that the same etch uniformity can be obtained on 200 mm and 300 mm wafers. Typical etch uniformity on 200 mm wafers is within wafer 0.88 percent, wafer to wafer is 0.90 percent and the standard deviation batch to batch is 1.81 percent for a 200 Å thermal oxide etch.
Loading of chemical bath
For future 0.25/0.18 µm feature size processing, the metallic impurity level limit is going to be determined by the contribution from the wafer itself, and for these levels of integration one-pass chemistry will be mandatory, since we are currently working at impurity levels where the loading of the bath due to the wafers is becoming more important than the initial purity of the chemicals themselves.
Moreover, it has been known for some time that the source of most of the particles in cleaning tanks during wet processing is the wafers [6]. To guarantee low particulate level wet processing, the tank must be continuously filtered. However, the filtering of wet processing tanks is highly dependent on the flow rates used during recirculation, since the tank behaves as a continuously stirred tank reactor [5]. However, since the volume of the tank when going from 200 mm to 300 mm is going to increase with a factor 3.6 (Area*pitch), the recirculation flows have to increase by a factor of 3.6. This will require much larger pumps and much larger filters, seriously compromising the required marginal increase in capital cost required to maintain the productivity gain when going from 200 mm to 300 mm.
One-pass chemistry, on the other hand, can be achieved for 300 mm without any increase in capital cost, when performed in single-bath processors and can be accompanied by very low chemical consumption due to minimum bath or vessel size.
Conclusions
In this paper, we have reviewed some recent trends in wet processing equipment to address the move from 200 mm tools to 300 mm tools. Although the main driving force for the move from 200 mm to 300 mm wafers is the increase in productivity, wet processing is at risk of becoming the only process sequence that will actually lose productivity if current multiple bath equipment is scaled for 300 mm.
It was shown that DI water consumption, footprint and exhaust costs are unacceptable for existing multiple-bath immersion system designs when going to 300 mm processing. On the other hand, single-bath processors, which use immersion technology in a single-processing vessel, are a cost-effective alternative when going to 300 mm processing. A single, completely contained processing vessel also offers additional benefits for 300 mm processing, such as elimination of exposure of operators to chemical fumes, no contamination of the cleanroom with NH3, HF or HCl, increased process control, no problems with drying 300 mm wafers, and no uniformity degradation when going from 200 mm to 300 mm wafers. In addition, the use of one-pass chemistry is offered at a lower chemical consumption and a much lower capital cost.
References
1. F. Tardiff, T. Lardin, P. Boelen and R. Novak, “Diluted Dynamic Clean DDC,” presented at UCPSS, September 23-25, Antwerp, Belgium, 1996.
2. M. Lancaster, “The Real Cost of DI Water : To Companies and the Environment,” 1996 Semiconductor Pure Water and Chemicals Conference, Santa Clara, Ca, March 4-7, 1996, in `96 SPWCC Proceedings, Edt M.K. Balazs, (Semiconductor Pre Water and Chemicals Conference, Santa Clara, Ca, 1996), p. 25.
3. N. Wiegler, “Reducing the Cost of Building a Fab”, in Semiconductor International, October 1996, p. 276.
4. T. Ohmi, in The Ohmi Papers, (Cannon Communications, 1990).
5. S. Verhaverbeke, C. McConnell, J. W. Parker and S. Bay ,”Scientific Rinsing and Drying on Macro and Microscale,” 1996 Semiconductor Pure Water and Chemicals Conference, Santa Clara, Ca, March 4-7, 1996, in `96 SPWCC Proceedings, Edt M.K. Balazs, (Semiconductor Pre Water and Chemicals Conference, Santa Clara, Ca, 1996), p. 33.
6. D. Hess, S. Klem and J. Grobelny, MICRO (Cannon Communications, Jan 1996), p. 39.
 |
 |
Figure 1: A single-chamber Full-Flow processor (left) has small vessel sizes, and the wafer pitch is one-half that of multiple-bath immersion systems (right).
 |
 |
CFM`s 8100 Full-Flow wet-processing system
 |
 |
Figure 2: Every point on the wafer sees exactly the same concentration profile during an etch, making this type of wet processing independent of wafer size.
 |
Figure 3: In 200 mm systems, 50 percent of the whole fab exhaust is used for conventional wet immersion systems.