Reduce Vent-Up Times on Loadlock Chambers

Wafer throughput and particle counts are key metrics for any semiconductor manufacturer's throughput and yield enhancement programs. Recent advancements in diffuser technology have helped manufacturers enhance these key metrics while improving the attributes for most vacuum process tools. These processes include dry etch, chemical vapor deposition (CVD), physical vapor deposition (PVD), rapid thermal processing (RTP) and epitaxial deposition (epi).

An early implementation of this technology was used on 200 mm batch-style loadlocks, which had an inherently large internal volume. The loadlock was prone to long vent cycles to prevent particle contamination. Implementation of diffuser technology dramatically decreased required vent time and ultimately became a tool upgrade option.

As the industry transitioned to 300 mm wafer platforms, factories increased their single-wafer loadlocks (SWLLs) in an effort to boost tool throughput. Compared to the 200 mm batch-style loadlocks, the SWLLs had extremely low internal volumes and were designed to cycle from vacuum to atmosphere very quickly. With the low volumes inherent in the SWLL, the velocity of the incoming vent gas became critical, because any particles on the bottom of the loadlock chamber would easily sweep onto the wafer should they be hit with a high-velocity gas. Particles are typically present in the loadlock because of mechanical wafer-handling devices and environmental exposure. Gas diffusers with ultrafine filtration membranes solved these issues. They allowed a large, uniform volumetric flow rate of gas into the loadlock chamber at low downstream gas velocities.

The method most widely established to control particle disturbance on 200 mm semiconductor vacuum process tool platforms is a two-step venting process, which implements a “soft” vent followed by a standard vent. The soft vent is typically conducted using a second line equipped with a flow restrictor to minimize the flow rate and bleed gas into the chamber until a certain pressure is reached inside the loadlock. This helps reduce the disturbance of particles. Once a set pressure is reached in the loadlock, a second valve is actuated to complete the venting process and bring the pressure of the loadlock to atmosphere. Depending on the volume of the loadlock chamber, the soft vent stage alone can take anywhere from a few seconds to several minutes to complete.

This method is often acceptable. However, there are cases where 200 mm tool owners are required to increase wafer throughput because of capacity constraints or to enhance overall equipment effectiveness (OEE). While many of the critical variables that influence wafer throughput are fixed (such as the process times, robot speed and loadlock pump-down speed), the time to vent up the loadlock may become the rate-limiting step to wafer throughput. This is especially true with shorter process times or if dual batch loadlocks are not working in parallel. One approach is to provide a rapid pressure increase by boosting the flow rate of gas. However, with a standard screen or open porous material, the gas velocity at the chamber entrance will be high and non-uniform, resulting in the disturbance of unwanted particles that have settled in the chamber.

The situation also occurs where vent-up time is not a throughput-limiting step. In this case, the tool owner is faced with more stringent particle requirements or observes a spike in particles on the wafers in the loadlock. The focus then becomes yield enhancement, and the goal is to reduce the particle adders on the wafers. Common approaches to the particle problem on installed system loadlocks have included complete loader rebuilds, performing additional series of wet cleans, upstream filter replacements and screen diffuser replacements, which often do not yield the desired goal.

A 200 mm upgrade solution with membrane diffuser technology allows a rapid but controlled vent up of loadlocks, cool down, transfer and process chambers from vacuum to atmospheric pressure while protecting the wafer integrity. Membrane diffusers have successfully reduced vent times on a variety of 200 mm vacuum process tool platforms by an average of 65% and significantly reduced particle adders — all by maximizing the volumetric flow and minimizing the velocity of ultrapure gas. Figure 1 shows the clear benefits of retrofitting 200 mm process tools by enhancing both particle performance and reducing loadlock vent time. Development was qualified with Applied Materials (Santa Clara, Calif.) using its tools.

1. Particle performance is improved and loadlock vent time is reduced with a membrane gas diffuser.

Membrane diffusers are designed using fine porous media, which uniformly spread the gas flow across a large area, resulting in lower gas velocities at the chamber entrance. The patented porous media also serve as a particle filter, removing particles down to 0.003 µm from the incoming gas at high volumetric flows. The result is ultraclean, diffused gas delivered to the process chamber, which minimizes on-wafer defects.

Nickel diffuser membranes (Fig. 2 ) have been shown to be effective in all environments, including poly and oxide etch processes where highly corrosive gases are used.

2. Membrane diffusers use fine porous media to uniformly spread the gas flow across a large area, resulting in lower gas velocities at the chamber entrance. The media also serves as a particle filter, removing particles down to 0.003 µm.

The effects of the membrane diffuser on vent time can be seen in Figure 3 . The red trace clearly shows the two-stage process and how the slow vent stage delays the vent to atmosphere. Conversely, the black trace of the membrane diffuser depicts a rapid, single vent to atmosphere. The diffuser allows the duration of the soft vent to be significantly reduced or even eliminated, and increases the volumetric flow rate into the loadlock to dramatically reduce the overall vent cycle. It should be noted that the diffuser does not affect the pump-down cycle.

3. The effects of the membrane diffuser on vent time.

As previously discussed, the significant decrease in vent time does not come at the expense of higher velocities and particle disturbance, as typically seen with screens or coarse porous frits. With the membrane diffuser, high volumetric flows can be achieved with low gas velocities. The membrane media is designed to uniformly spread gas flows across a large area relative to a standard gas line, a series of drilled holes, or coarse screens. Figure 4 shows a comparison of different components used to create uniform flow. The measurements were taken using an ultrasonic probe in the fluid path exiting the component. The results show how the membrane diffuser is more effective compared with a frit or screen under the same volumetric flow conditions.

4. A comparison of different components used to create uniform flow.

In addition, Figure 5 shows a typical velocity profile of a disc-type filter membrane diffuser. For an inlet flow rate of 170 standard liters per minute (slpm), the downstream velocity is <100 ft/min at a distance of only 3 in. from the face, or entrance, to the loadlock chamber.

5. A typical velocity profile of a disc-type filter membrane diffuser.

The result of lower velocities is the decrease in particle counts (or adders) to the wafer while in the loadlock chamber. Figure 6 shows particle results taken on wafers prior to and after the installation of the membrane diffuser on a 200 mm loadlock. The combination of an ultrapure filter and fine-membrane gas diffuser allowed this dramatic reduction in particle levels.

6. Particle results taken on wafers prior to and after installation of the membrane diffuser on a 200 mm loadlock.

One of the most difficult questions to answer is precisely what velocities are acceptable with respect to particle re-entrainment. This is a problem compounded by the various mechanisms that adhere a particle to the surface, and the various sizes and shapes of these particles. This can make it extremely difficult to know the flow required to lift a particle. In addition, since the fluid flow conditions are dynamic, the boundary layer conditions are also active and contribute additional uncertainty in the fluid force available to lift a particle. Methods to resolve these issues are being investigated, but they are currently beyond the present scope of this cuticle.

To determine optimum venting is a relatively intractable analytical problem to fully solve. Physical geometries are fairly complicated, making computational fluid dynamics (CFD) modeling difficult. Additionally, fluid flow may be present in various flow regimes, including molecular flow, viscous flow (both laminar and turbulent) and even as shockwave fronts. Lastly, the size or adhesiveness of settled particles on the floor or walls of a loadlock or chamber make it difficult to determine the exact target for nearby fluid velocities to minimize re-entrainment.

Simplified CFD models can provide a general picture of the fluid flow in a loadlock or chamber leading to the best compromise between short vent time, minimal fluid velocity, physical placement of a diffuser, shape of the diffuser, chamber or loadlock geometry, and vent-up parameters (e.g., soft vent use).

Venting is, by nature, a dynamic process that considers the complexity of the fluid density in the loadlock or chamber, which is continually varying. It is a good practice to combine analytical results with actual tests to arrive at an understanding of the general processes and the particle reduction effects that a diffuser installation may offer.

The controlled permeability of the diffuser membrane also helps by making the fluid flow uniform across the membrane, and offers proper resistance to flow in this configuration.

Some of the variables controlled by the design engineer are diffuser location during installation, membrane shape and membrane permeability.

Our Chambergard diffusers, incorporated with the Applied Materials Fast Vent loadlock diffuser kit, have been installed onto a variety of 200 and 150 mm semiconductor tool platforms that cover a range of processes. These processes include dry etch, CVD, PVD and epi. In all cases, the vent time was decreased significantly and, where reported, particle adders to the wafer were reduced. The Table summarizes selected results of membrane diffusers on 200 mm tools, where reductions in both wafer contamination and vent-up times were observed. On a typical batch loadlock, vent times as high as seven minutes were reduced to less than three minutes with particle generation decreased by a factor of 10 with no negative effect on pump-down speed. The exact increase in wafer throughput depends on several factors. Typical results reported have ranged from a 6% to >10% increase in wph.

Discussion of results

When selected data points from etch, CVD, PVD and epi are plotted as particle adders (measured on the wafer) vs. loadlock vent-up time, the benefits of the diffuser product can easily be seen (Fig. 1). The Figure shows intuitively what can be expected. As the vent-up time of the loadlock is decreased, the number of particle adders increases — regardless of the venting method. After the installation of the Chambergard diffuser, a significant improvement in both vent-up time and particle adders is realized, causing the resultant shift of the curve.