Small Changes Boost Vacuum Productivity

		At a Glance
	Maintenance time and cost often can be reduced and OEE increased by modifying vacuum system hardware and operating procedures. Some improvements do not obviously reveal their value because they quietly prevent problems. Other seemingly insignificant changes may save only a few minutes per task but, when multiplied by the number of times the task must be performed, can easily increase tool uptime by several hours per month.

For the majority of semiconductor manufacturing processes, the vacuum system operation usually can be enhanced to improve reliability, throughput, uptime, contamination control, process control and process capability (C_p). For example, a vacuum pump-down or purge cycle often can be shortened without degrading the product, and the time saved can significantly increase throughput.

Some areas of improvement in hardware and operation that can increase OEE are:

• Periodic calibration and routine in situ verification of pressure sensors and mass flow controllers.
• Effective and time-efficient pumping, purging and venting.
• Control of unwanted effluent exhaust deposition.
• General improvement in
  operational efficiency.
• Personnel training.

Periodic calibration and routine verification

It is not enough for a particular tool to perform in a consistent manner, run-to-run and day-to-day; multiple tools running the same process must run it in the same manner and under the same conditions. From a vacuum and gas control standpoint, that means identical flows and pressures.

The most common instruments for process pressure and flow measurement and control are capacitance manometers and thermal mass flow controllers (MFCs). These instruments are individually calibrated at the factory and guaranteed to meet specific accuracy requirements. However, accuracy may degrade over time, requiring periodic recertification.

Calibration intervals for these instruments are usually on the order of one year. The calibration process is intrusive in that the device is removed from the tool and then tested on a dedicated calibration instrument, either at the user's facility or at a remote service center.

Fig 1 A portable transfer standard can be quickly and conveniently coupled to the process gauge and tool chamber. Application-specific diagnostics access valves can be installed with no modification to the process tool.

Calibrating pressure instruments involves attaching the device to a high vacuum system that has a controlled pressure manifold to which the gauges are attached; a set of transfer standard grade capacitance manometers (typically with accuracies of 0.05% to 0.08% of reading and traceable to a standards body such as NIST) with ranges that correspond to the ranges of the devices to be calibrated; and appropriate pressure readout and control devices. The device is compared to the transfer standard over its entire range. Adjustments to span and linearity are made to the device under test as needed.¹

There are various approaches to calibrating mass flow controllers, including the use of thermal or laminar flow-based transfer standards. One of the most versatile methods is the ROR, or pressure rate-of-rise, technique. Gas from the MFC flows into a known volume, and knowing the volume, pressure change, temperature and elapsed time, one can calculate the mass flow with great precision. Because this method is based on fundamental or primary quantities, the measurement is independent of the gas used.

As with the pressure instruments, the highest accuracies for flow calibration are best achieved with dedicated off-line instruments. These can achieve accuracies in the range of 0.2% to 0.5% of the reading. Also as with the pressure instruments, periodic formal calibrations are desirable, but they are done at long cycles to minimize tool downtime.

In situ verification

Long calibration cycles may not be adequate to catch accuracy deteriorations that may be due to factors external to the instrument. Slow contamination buildups and other process- related problems may compromise consistency of process parameters. However, frequent off-line calibration cycles will reduce OEE because of the added downtime. The calibration process takes at least several hours, and perhaps days if done off-site.

Substituting another device for the one that is removed minimizes downtime. Even with a substitute at hand, the swap-out can take several hours allowing for warm-up times, purge cycles and leak checking. But if nothing is done and accuracy deterioration goes undetected, product yields and reliability may suffer.

Maintaining consistency at the pressure and flow set points is the key. Simple techniques for in situ verification of these set points can do much to guarantee process repeatability while minimizing downtime. One approach involves coupling a pressure or flow standard to the system, which permits a comparison of the process instrument to the standard. Only if the process instrument were out of spec would removal or other corrective measures be taken.

Fig 2 Using a dedicated rate-of-rise measurement system incorporated in the gas box vent line, actual flow (dotted) vs. set point (solid) of four MFCs was monitored over a 2-week period at an actual fab. Note that the flow for gas 4 drifts while gas 2 is out of spec for the entire period.

Portable, high-accuracy capacitance manometer transfer standards are available that can be connected to an access port on the tool. The tool connection is easily implemented with a valve designed to isolate the capacitance manometer and provide a port for manometer calibration or for checking the process tool's pressure set points (Fig. 1). The consistent and frequent use of a calibration standard on all tools ensures process repeatability and tool-to-tool uniformity.

In situ pressure ROR verification can be incorporated into a process tool to determine the general health of its MFCs. In the simplest implementation, the MFC being tested flows gas into the pumped and isolated process chamber. The process capacitance manometer monitors the pressure rise, and the tool host computer performs the necessary calibration calculations. This is usually more of a qualitative consistency check, and results may vary significantly from one tool to another. In some tools, the process manometer may have too low a full-scale range for the given flow rate, or there may be imprecise chamber volume and uncertainties in chamber temperature. There also may be problems due to cross contamination, and, since the test has to be performed using the tool process chamber, some downtime is incurred.

A better in situ approach is to use a dedicated rate-of-rise subsystem incorporated in the gas box vent line. Verification can be performed independently of the process cycle, eliminating tool downtime. An integrated assembly consisting of a pressure transducer, a calibrated volume and the necessary isolation valves is capable of a traceable accur acy within 1% of reading over a flow range of 2 sccm to 20 slm. Fig. 2 shows the results of such a system monitoring flow vs. set point over a two-week period on an actual process tool.

Fig 3 Photo of heated exhaust line module with heated valves, high-temperature capacitance manometer and downstream condensation trap.

Pumping, purging and venting

A residual gas analyzer can track partial pressures during pumpdown and rate-of-rise tests. Identifying leaks immediately, rather than waiting in vain to achieve base pressure, can save time.

CVD and etch processes require repeated pumping and purging between process steps to rid the chamber of process gases. This pump/purge sequence is often 'over-engineered' to compensate for variations in equipment and chamber condition. Again, residual gas analysis can identify the dominant gases remaining after pumpdown or purging. Actual residual gas levels rather than time and total pressure can then determine the optimum cycle.

Particle contamination can also often be reduced if vacuum system pumpdown and venting are programmed to reduce turbulence during the initial stages of pump and vent cycles.² For example, when venting a loadlock chamber, a two-step process can be used: a slow vent to a crossover pressure followed by more rapid venting to atmospheric pressure. The crossover pressure is chosen to be high enough so that, during the subsequent rapid vent cycle, turbulence will not stir up particles, and condensables will not precipitate. An analogous situation occurs during pumpdown from atmospheric pressure; a two-stage pump cycle often is used to avoid condensation and turbulence.

The loadlock pressure usually is measured with a convection-enhanced Pirani gauge with a built-in crossover setpoint. If the setpoint is in a region where the gauge has low accuracy, the slow vent time can be longer than necessary. In one case, the crossover pressure was specified at 80 Torr, and a test with a capacitance manometer showed the loadlock actually was slow venting to 120 Torr, adding two minutes to the process time. The Pirani gauge was replaced with an absolute pressure switch based on a capacitance manometer sensor with a 100 Torr full-scale range. The crossover pressure was accurately sensed, and the two minutes per cycle were reclaimed, significantly increasing throughput. A capacitance sensor also is insensitive to gas species, a plus if vent gases other than air or nitrogen are used. Two-stage soft-start isolation valves are available that simplify and automate two-stage pumping and venting.

Process effluent management

CVD and plasma etching produce reaction products that condense in the vacuum line, clog valves, and damage or destroy mechanical vacuum pumps. Obviously, these occurrences cut into equipment uptime, and the resulting particulate formation reduces production yield. Effective control of effluent deposition can have a triple positive benefit: increased pump life, reduced maintenance time and a longer time between preventive maintenance, and improved yield.

Heating the vacuum line to prevent condensation and then capturing the material downstream in an efficient, high-conductance trap can control some effluents (Fig. 3). This is effective for processes that produce readily condensable byproducts such as aluminum plasma etching and silicon nitride low-pressure chemical vapor deposition (LPCVD).^3,4 This method will not work for effluents during LPCVD using tetraethoxysilane (TEOS).

In a typical TEOS LPCVD process, deposits form in the pump line, and frequent cleaning is required. Solid and powdery deposits form adjacent to the furnace pump port (Fig. 4). Back transmission of the powders increases the particulate level in the furnace and reduces product yield. Solid and viscous liquid deposits also can form further downstream, where they can clog pressure gauge ports and reduce the conductance to the vacuum pump. Powders can form inside the pump, where high temperature and pressure promote polymerization. Vacuum pumps can be destroyed, and pump life can be as short as eight weeks. In many cases, the pump line between the furnace and pump remains quite clean.

Fig 4 For TEOS LPCVD processes, powdery deposits can form at the furnace pump port (a). Unreacted and partially polymerized TEOS can react with water to form polymerized TEOS deposits in the effluent line (b).

The downstream deposits are primarily polymerized TEOS formed by surface chemical reactions between water and unreacted and partially polymerized TEOS. One way to inhibit these surface reactions is to create a heated flowing nitrogen boundary layer between the wall and the flowing effluent gas. This moves the polymerization processes downstream away from the furnace exit (Fig. 5). Adding an ambient-temperature, high-efficiency trap downstream and ahead of the pump can provide a greater than five-fold extension of the PM cycle and reduce particle levels by ~20%.⁵

Partial pressure sensors

Another often under-used vacuum sensor is the quadrupole mass spectrometer or residual gas analyzer (RGA). An in-situ RGA can provide a window into the 'process state' of semiconductor deposition and plasma etch tools. Many aspects of complex, multi-step processes can be tracked by monitoring, in real time, mass peaks that correspond to process gases, byproducts and the residual gas background.⁶ Data from an RGA-equipped production process tool also can be used for real-time fault detection and classification and can provide information for model-based advanced process control.

Robust pressure sensors

Simply changing to an alternative sensor can reduce costs and improve uptime. The cold cathode gauge is an often overlooked, cost-effective gauge for ion implanters and other high-vacuum applications. A cold cathode gauge has no filament, is not subject to burnout and has a significantly longer life and lower annual replacement cost than a hot cathode Bayard-Alpert gauge.⁷ A cold-cathode gauge can be replaced at predictable intervals, whereas a hot cathode gauge must be replaced whenever a burnout occurs. In addition, the annual cost of replacement is less. A modern inverted magnetron cold cathode gauge is stable and can operate over a wide pressure range from 10 mTorr down to ultra-high vacuum. Repeatability is better than 9%.⁸ Cold cathode gauges have a reputation as 'slow starters,' since they must rely on an ionization source such as a cosmic ray to initiate a discharge.⁹ The lower the pressure, the less likely an ionizing collision and the longer the start delay. At
1 mTorr or higher, there is no delay; but at 10^-8 Torr, starting can take up to 3 hours. The traditional solution is to start the gauge in the mTorr range, but often this is not practical. A relatively new technique uses ultraviolet induced photoemission to generate an ionizing electron to start the gauge.⁸ At 10^-6 Torr the start time is 4-5 seconds.

Fig 5 Nitrogen enters plenum near wall and is directed into the exhaust gas stream through overlapping chevron openings.

Adding the UV source is a simple, non-intrusive retrofit. A miniature UV discharge lamp is mounted in a double-sided 2 3/4 in. CF flange and isolated from the vacuum by a welded UV-transparent glass envelope. The flange is interposed between an inverted magnetron gauge and the vacuum chamber connection. The UV lamp is easily removable for bakeout or replacement without disturbing the vacuum seal.

Fail-safe operation

Fail-safe devices can save time and grief, but unless they prevent a hazardous condition, they usually are omitted as a 'cost-saving' measure. This is actually a 'save a little now -- pay more later' strategy. One vacuum fail-safe feature is a safety valve that protects a vacuum system in case of power interruption. Its function is to protect the vacuum system upon power failure by isolating and venting a mechanical vacuum pump, preferably to dry nitrogen or other inert gas. This prevents oil backup into the vacuum line, allows the motor to restart the pump easily, and prevents airborne contamination from entering the vacuum pump. This reduces the time required to bring a process tool back on line after a power interruption.

A fail-safe valve must close upon power loss and not require other external utilities such as compressed air. A properly designed valve operates with atmospheric pressure, activates upon loss of electrical power and requires no external pressurized gas source. A designed-in buffer volume minimizes the closing gas burst. The gas burst, upon re-opening, is less than the critical backing pressure for turbomolecular and diffusion pumps when appropriately sized mechanical pumps are used. Standard ISO-KF flange dimensions allow the valve to replace elbows, tees and crosses of the same flange size.

Little things

Little things can mean a lot. Opening and closing O-ring-sealed vacuum flanges historically required the removal and replacement of six to eight bolts and nuts. Today's standardized KF flanges are quickly sealed and unsealed with a single clamp and wing nut. A further improvement is a toggle clamp -- the seal is created with a pull of the lever on the clamp. The clamp seals itself with a spring mechanism that ensures it stays closed. Older technology required that a pin be inserted to maintain closure. These are small improvements that add up to shorter maintenance cycles.

Another small fix can reduce clogging of sensor ports by effluent condensation. It is difficult to efficiently heat the hex nut on a sensor VCR fitting since the heater does not conform well to the irregular shape. The resulting cold spot is a trap for condensation and a locus for particle generation. A simple aluminum clamshell adapter solves the problem by making the hex shape round. A dowel pin clamps the two halves together around the VCR connection. The aluminum efficiently transfers the heat directly to the fitting. An adapter for a standard 1/2-in. VCR fitting has an inside shape that matches the hex nut and a 1.5-in. outside diameter that conforms to a standard heating mantle.

Personnel training

Finally, training is critical. Training enhances skills and reduces the time wasted in troubleshooting and trial-and-error maintenance. A knowledgeable and skilled workforce is important for optimizing OEE and continuous improvement. Knowing how the various instruments work, their characteristics, how to set up, maintain and troubleshoot, and how they interact with the process environment can play an important role in reducing downtime and improving process uniformity.

Process tool and component vendors often can provide on-site or off-site training to promote an understanding of instrumentation basics, such as pressure and flow control devices, vacuum gauging, and mass spectrometer process monitors. Other training is offered by professional societies, such as the American Vacuum Society, and colleges and universities. The primary audience is equipment technicians and engineers.

Conclusion

Vacuum-related component and process tool vendors offer many solutions to the nagging problems and situations that reduce productivity. Some of these approaches are rather simple to implement, but can improve tool uptime significantly. Other upgrades require an investment of capital and time, but they often pay for themselves very quickly.

Acknowledgements

The authors thank Kevin Grout, Dick Jacobs, and John J. Sullivan of MKS; Kathryn Whitenack of Lytron Inc.; and John O'Hanlon of the University of Arizona for helpful suggestions and insights.

References

1. R.W. Hyland and R.L. Shaffer, 'AVS Recommended Practices for the Calibration and Use of Capacitance Diaphragm Gauges as Transfer Standards,' J. Vac. Sci. Technol. A 9 (6) Nov/Dec 1991.
2. S.P. Hansen and K. Whitenack, 'Pressure Mmeasurement and Control in Loadlocks,' Solid State Technology, Oct 1997, p.151.
3. F. Lee, E.M. Howard and D.A. DeMuynck, 'Detecting and Reducing Particles for LPCVD Silicon Nitride Deposition,' Microcontamination 12(3), March 1994, pp. 33-38.
4. Y. Gu and D. Hauschulz, 'Comprehensive Downstream Effluent Management,' Solid State Technology, May 1998, p.89-98.
5. Y. Gu, P. Dozoretz, J. Bhakta and F. Gologhlan, 'Management of TEOS LPCVD Process Effluents,' Solid State Technology, July 1999, pp. 151-156.
6. K.C. Lin, C. Marcadal, S. Ganguli, B. Zheng, J. Schmitt, and L. Chen, 'Characterization of Copper CVD Process by A Process Monitor,' IEEE/SEMI Advanced Semiconductor Manufacturing Conference and Workshop, 1999.
7. R.N. Peacock, N.T. Peacock, and D.S. Hauschulz, 'Comparison of Hot Cathode and Cold Cathode Ionization Gauges,' J. Vac. Sci. Technol. A 9(3), May/Jun 1991, pp. 1977-1985.
8. D.L. Hyatt and N.T. Peacock, 'Long Term Stability of an Inverted Magnetron Cold Cathode Gauge,' 37th Annual Proceedings, Society of Vacuum Coaters (1994), pp. 409-412.
9. B.R.F. Kendall and E. Drubetsky, 'Starting Delays in Cold-Cathode Gauges at Low Pressures,' J. Vac. Sci. Technol. A 14(3), May/Jun 1996, pp. 1292-1296.

Steve Hansen is MKS Instruments' product marketing manager for calibration products and training. He has 24 years of experience in process development in semiconductor-related areas and joined MKS in 1995 to develop and deploy customer education programs. Phone: 978-975-2350; Fax: 978-975-0093 Email: Steve_Hansen@mksinst.com

Steve Sexton is MKS Instruments' HPS Products Group director of sales and marketing for the U.S. and Asia. Phone 800-345-1967, 303-449-9861 Fax 303-449-6880 Email: Steve_Sexton@mksinst.com