Using numerical simulation to optimize 300mm FOUP purging
10/01/2003
Overview
The front-opening unified pod is synonymous with 300mm wafer handling and contamination control. However, simple intuitive purging of these pods apparently does not provide wafer cleanliness compatible with high yields. Rigorous simulation with computational fluid dynamic software shows just how pod purging can be improved.
Despite the fact that 300mm wafers are transported in front-opening unified pods (FOUPs) to avoid contamination, these pods can still contain a significant amount of contamination with the potential to damage wafers. For example, the presence of humidity in an enclosed environment in contact with wafers can cause various phenomena, such as native oxide growth, corrosion, and film cracking. Furthermore, the presence of emitted organic compounds can lead to degradation of the electrical properties in circuits on wafers.
 |
null
Purging FOUPs is widely accepted as a means to eliminate any undesirable contamination; FOUP suppliers offer optional purging systems on their pods. For example, nitrogen is widely used to avoid native oxide growth and organic contaminants.
In our studies, we have looked at the potential inefficiency of purging a FOUP with an inlet and outlet located on the bottom of the pod — the "classical" purge configuration. We examined, using numerical simulation tools, the physical phenomena of gas flow in FOUPs to understand any drawbacks and inefficiencies of classical purging. Armed with a good comprehension of gas flow dynamics in FOUPs, we derived how to optimize purge efficiency and its quality.
We did our simulations with the computational fluid dynamic software FLUENT (Fluent Inc., Lebanon NH), applying the following scenario to simulate FOUP purging. We started with the FOUP filled with ambient air at 40% relative humidity, a typical value for a wafer fab cleanroom. We did not account for adsorption, permeation of water vapor, or any external leaks. We applied a constant nitrogen flow to the inlet of the pod and terminated purging when residual air in the pod reached 0.4% relative humidity (the "critical point" or a 1% air mass fraction). Simulations were done at 297°K and 1atm.
FOUP-purging efficiency
To decrease FOUP-purging time, a large volume of gas flow is required. However, rapid gas flow yields undesirable results, such as provoking further contamination by releasing particles from wafers and a FOUP's interior surfaces. Furthermore, it is desirable to reduce the consumption of purge gas to save cost.
We simulated purge gas flow rates from 0.5–20slm to find the time required to purge a FOUP to the critical point. We compare our simulation results with ideal purging time, expressed as
 |
null
where Tideal = the ideal purging time in seconds, VFOUP = the FOUP's interior volume in liters, and Q = the purge flow rate in liters/min.
From the formula, an ideal case using a 2slm flow will purge a FOUP volume of 25 liters in 12.5 min (see the table). We will use this "optimal" value in comparisons discussed below.
 Figure 1. Time plots for an "optimal" and simulated 2slm FOUP purge. Air mass fraction is the initial air in the FOUP divided by new gas introduced by purging. |
null
Optimal classical purging versus simulated data for 2slm shows (Fig. 1):
- To reach the critical point, purging time exceeds 40 min, three times longer than an optimal purge.
- At first, classical and optimal purges have the same speed, but the two curves diverge after a few minutes of purging.
We believe that the root understanding of successful FOUP purging rests in understanding these points.
Our calculated and simulated data show that as the purge flow rate increases, purge time decreases. However, comparing the ratios of the two shows how much injected gas is actually serving to purge the FOUP environment; there is less efficiency with larger flow rate (see the table). Nitrogen consumption increases for larger inlet flow to reach the critical point.
When we simulated 3-D flow velocity distribution of the FOUP with classic bottom inlet and outlet and loaded with 25 wafers, we observed a high gas flow rate on the side of the FOUP (Fig. 2a, b); incoming gas had difficulty going between wafers. This was further supported by simulated 3-D depictions that showed the remaining gas was not homogeneously distributed in the FOUP; some of the initial high-humidity gas was concentrated on the bottom of the FOUP and between wafers (Fig. 2c). We observed these same results with all of our simulated flow rates and concluded that classical purging is inefficient for production applications.
null
Inefficient classical purging
Trying to uncover the cause of this inefficient purging with classic FOUPS, we simulated gas density in a FOUP after ~7 min of purging at 2slm (Fig. 3). We observed a dense iso-surface (where the mass fraction of the initial gas is constant) that represented a stagnant zone of residual initial gas.
null
Further, these simulations showed that nitrogen inflow from the bottom of the FOUP, with initial velocity normal to the bottom, heads directly to the upper part of the FOUP, without filling spaces between wafers. In addition, nitrogen's weight of 28g/mol compared to 29g/mol for the humid air makes it tend to rise above the air.
During purging, our simulations showed that the inlet nitrogen concentrated in two zones: in the upper section of the FOUP, and at first as the nitrogen flows down along the wall of the FOUP, outside of the small (~1cm) spaces between wafers. This occurs after the gas density of the upper part of the FOUP becomes equal to the nitrogen density.
While it takes a certain time for the nitrogen to reach the bottom of the pod, it takes even longer for it to move horizontally between lower wafers, with the center of the bottom wafers the last to be purged because of slow horizontal diffusion of the nitrogen gradient.
Our conclusion is that classical FOUP purging has two drawbacks that must be solved:
- The ~1cm spacing between 300mm wafers in FOUPs has a certain conductance that must be overcome.
- Different residual and inflow gas densities cause stagnant zones that contribute to slow diffusion.
Through simulation, we looked at two possible solutions designed to provide better FOUP purging. One method uses a special injector with ports that project purge gas between wafers; the second uses gases with different densities.
Optimizing the concepts
Simulations revealed that the plenum injector did not result in as great an improvement in purging stagnant zones between wafers as might be expected. Using a 5slm inflow, the plenum injector takes 16.5 min to reach the critical point, compared to 24 min with classical purging and 5 min for an ideal purge; the plenum-to-classical purge ratio was 3.3.
Simulations showed that while the plenum injector improves purging speed, residual gas persisted between wafers. Initially, nitrogen from the injector meets air with higher density and it diverts upwards. The residual gas tends to stay close to the wafer surface where purging is needed most. Again, the root cause is due to different gas densities. (When we simulated a plenum with 25 ports aligned with each space between wafers, we did see a small improvement in purging time. However, we determined this approach was too complex an injector geometry and resulted in high gas velocity directly onto the wafer surfaces.)
Simulations using the plenum with two gases of equal density (i.e., air and air) show no particular stagnant zones between wafers, resulting in uniformly purged spaces (Fig. 4).
 Figure 4. Residual gas density change with the plenum injector purging air at 7slm — a) purge start, b) after 5 min, and c) after ~15 min reaching the critical point. |
null
Purge optimization
The result of all our simulations, those discussed in detail above and others, has revealed that with a good comprehension of the purge phenomena, we can adjust FOUP purge parameters to promote gas diffusion between wafers that approaches an ideal situation. The FOUP modifications include:
- adopting a plenum injector configuration where the number of injectors and nozzle direction purge the stagnant zones between wafers;
- controlling purge gas temperature to vary its density, thus inducing convection and minimizing gas stagnation;
- optimizing inlet and outlet locations; and
- selecting a purge gas other than nitrogen (we have simulation data with argon and dry air).
We have found through simulation that our optimized FOUP configuration purges to the critical point in only 7.5 min using a 5slm flow, a 1.5 comparison ratio with ideal purging (Fig. 5). This configuration achieves purging efficiency, avoids stagnant zones, and does not show any recirculation of gas that leads to stagnation of undesired contaminating particles in the pod.
 Figure 5. Comparison of different FOUP nitrogen purge times. |
null
Roland Bernard, Hisanori Kambara, Alcatel Vacuum Technology, Annecy, France
Arnaud Favre, INOPRO, Grenoble, France
Pascal Descamp, Agnès Roche, STMicroelectronics, Crolles, France
Acknowledgments
We thank Hervé Kubler, Eveline Rousseau, and others in the cleanroom group at STMicroelectronics in Crolles for providing us with experimental data and organization. We also acknowledge the applications department at Alcatel Vacuum Technology for simulation facilities and support.
Roland Bernard is a director of applications and business at Alcatel Vacuum Technology, 98 ave. de Brogny, 74009 Annecy, France; ph 33/4-50-65-74-94, fax 33/4-50 65-7582, e-mail arnaud.favre-ST @an.cit.alcatel.fr.
Hisanori Kambara manages the research-engineering department at Alcatel Vacuum Technology.
Arnaud Favre is head engineer at INOPRO.
Pascal Descamp is a methods and handling engineer at STMicroelectronics.
Agnès Roche is a microcontamination engineer at STMicroelectronics.