A collaborative demonstration at G450C proactively trials and examines a solution to reduce cost for higher flows.
BY ADRIENNE PIERCE and CHRIS BAILEY, Edwards Ltd., Santa Clara, CA and BURGESS HILL, UK, and BILL CORBIN, G450C IBM Assignee.
The impending change in silicon wafer diameter from 300mm to 450mm will increase the surface area of each wafer by 2.25 times. A worst-case scenario suggests that process gas flow rates required to maintain wafer throughput at acceptable levels would increase by the same scaling factor. Since the main reason to go to a 450mm wafer size is to lower manufacturing costs, we need to explore how to best minimize the downstream impact of higher gas flow rates on capital and operating expenditures of tool-support equipment, such as vacuum and gas abatement systems. In the case of flammable process gases, some thoughtful consideration and innovative options are required. The combination of higher flammable process gas flows and their associated safety dilution guidelines could greatly increase sub-fab space, equipment and facilities requirements, especially in the event that abatement systems are necessary to handle the total exhausted gas. Safely minimizing or eliminating additional dilution volumes is a viable opportunity to reducing the need for additional abatement units in a 450mm high volume manufacturing environment, the implication of which is increased capital and operational costs, not only for the base equipment, but also for site infrastructure which must be scaled to handle any additional abatement consumables and waste.
This article considers the likely impact and trade-offs of such flammable gas flow increases on process vacuum and abatement systems, which under a “business as usual” model would scale up purge and equipment sizes based on safety multipliers. Therefore, we propose an alternative approach: implementing an integrated vacuum and abatement system with a common supervisory control and monitored joints, which allows purge nitrogen flows and equipment sizes to be significantly reduced while still maintaining operational safety and compliance with SEMI standards and NFPA codes. Both the technical and cost implications are explored and data is provided from laboratory trials. The results suggest that there is an opportunity to enable 450mm capabilities by leveraging novel integration of known technologies to reduce gas flow increases and related capital and operating expenses.
For higher gas flow rates in 300mm and 450mm, a particular challenge is safely handling and using flammable or pyrophoric process gases, such as hydrogen, silane, ammonia or phosphine. These gases are employed in a vacuum process chamber, pulled through a foreline using a pump which sits on the tool, in the sub-fab or both (FIGURE 1). These gases and their by-products are then exhausted at near atmospheric pressure to a point-of-use abatement device for treatment.
In a pure vacuum there is insufficient gas for combustion. A flammable gas can support combustion typically only above 50 mbar (0.725 psi or 5 KPa or 37.5 Torr). So the focus area is from the exit of
vacuum pump to the abatement unit in the subfab, which is at about atmospheric pressure.
In order for a reaction to occur, there are three requirements in any system: 1) an ignition source, 2) sufficient fuel concentration and 3) an oxidizer present within the flammable concentration range of the fuel. Combustion will not occur if the ignition source is not energetic enough to initiate the reaction, or either fuel or oxidant is not present within the flammable concentration range.
In considering the first requirement for combustion, process gases pass through a vacuum pump, which is a metallic, motor-driven mechanical device, therefore, an ignition source cannot be ruled out nor can its energy be predicted. For the second condition, flammable process gases are dictate by process recipe require- ments. Every “process fuel” gas has a lower flammable limit (LFL), which is the concentration in air below which it will not combust. TABLE 1 shows some common process gases with their LFLs noted as a percentage of total gas composition.
Best safety practice as per NFPA68 and NFPA318, is to add an inert diluent such as nitrogen (N2), to the process gas stream at or near the subfab vacuum pump. Flow rates are calculated to a fraction of LFL and based on maximum mass flow controller (MFC) settings and fab safety policy. For instance, a flow of 1 standard liter per minute (slm) of silane (SiH4) at 1/2 LFL (1.37%) requires a flow of 145 slm of N2. Many sub-fab vacuum systems include a N2 purge from 0 to 200 slm. So an MFC larger than 1 slm of SiH4 using this methodology, will drive a need for additional N2 added after the pump to retain a non-flammable diluted gas mixture. (TABLE 2).
At an MFC of greater than 1 slm of SiH4, the N2 requirement increases rapidly. Extra N2 not only increases the cost of the inert gas but requires that the downstream abatement and scrubbed exhaust system are able to handle the greater flow. This can double or triple the abatement capacity requirement, adding to the heat load in the sub-fab (when considering combustion type abatement), and increase facilities handling requirements. 150 slm of N2 costs (US average $0.05/ m3) about $5,000 annually and can occupy up to 25% of the abatement capacity, or more depending on the device. Using generic MFC flows for flammable process gases and surveying 300mm processes which could require 150 slm of N2 dilution (additive to the typical dry pump purge) yields the list shown in Table 3 of processes which may have flammable gas flows (TABLE 3). Fab-wide, these critical processes require a lot of chambers and additional N2 and this will only increase with 450mm flows. So, let us consider the case of the third condition needed to sustain a combustion reaction: oxidizers. Oxidizers can be present in the process gases, or oxygen can leak into the vacuum system from the environment. What if instead of diluting flammable gases, we prevented and monitored so that ambient oxidizers never enter the system?
An alternative: The monitored connection
In the case of process recipes that prescribe flammable gases but no oxidizers and where dilution flows have become very high, an option is to prevent the intro- duction of an oxidizer, ambient air. To this effect, G450C and Edwards are looking to actively monitor the connections on the downstream side of the pump (FIGURE 2).
Critical to safe operation and monitoring is a fault tolerant, safety rated control designed to be compliant with NFPA 79, Section 9.4.3, “Control Systems Incorporating Software-and Firmware-Based Controls”. A safety rated PLC (programmable logic controller) monitors an array of hardware based sensors, and will alert operators whenever a system fault is detected. Further, this control system will be integrated with the connected processing equipment and the factory safety system (often referred to as Toxic Gas Monitoring System or TGMS) to shut down gases when an out of specification condition exists.
Each connection has a secondary seal encircling it to create a space, which is then pressurized with N2. A pressure change in that pressurized space indicates a breach either through the inner connection, where N2 will be added to the process gas stream, or through the outer joint, with N2 flowing to ambient (FIGURE 3). This arrangement of monitored connection provides the additional benefit of not allowing process gases to leak to ambient in the event of a connection failure. Monitoring looks for a change in pressure and is managed by the safety rated PLC (noted as system controller in FIGURE 4). The system also incorporates the active monitoring of an existing flame arrester just up-stream of the abatement to ensure that there is no flame propagation up the exhaust line. This monitor is interfaced to automatically shut off the process gases if a flame is detected at the monitored location. The monitoring of the exhaust joints depends on the presence on nitrogen pressure. The uptime of the vacuum system would depend on nitrogen pressure to maintain purge flows and, for safety reasons, verification of the N2 supply would be a fail-safe requirement. In the event of nitrogen pressure loss due to facilities failures or other reasons, this monitoring system is fail-safe, so that loss of nitrogen pressure will stop the process. It is not believed that this monitoring connection system is any more likely to fail than a high flow nitrogen purge system.
Prior to testing the system at G450C on the integrated vacuum and abatement system, a hardware test rig was set up in the laboratory to verify software, performance during leaks, fluctuations and signaling protocol.
The two main test objectives were:
1. to determine the pressure response when an o-ring fails and
2. to confirm independent, non-interfering monitoring of each coupling.
Test 1 was set up to measure the response to a hairline failure of an o-ring and compare that to a major failure by measuring the inter-seal pressure response in each case. The major failure responds with a noticeably different (larger) inter-seal pressure change than for a hairline failure. These test outputs established the set point levels for indicating an alarm or warning status (FIGURE 5).
Test 2 was used to confirm that individual couplings can be monitored independently from each other by observing that the measured pressure response from a failed o-ring in one coupling does not cause inter- ference with the monitoring of another coupling. The same component parts that are to be installed at G450C were used in these verification tests monitored by the safety PLC and IO (input/ output) unit (FIGURE 6). TABLE 4 summarizes the test results. In addition, a risk assessment and SEMI S2 third party review will be conducted to ensure thorough consideration of the equipment, implementation and safety. Once installed at G450C, the monitoring will be exercised and regular reviews will verify performance and other necessary procedures. Further evaluation will be given to ensure effective abatement performance with less dilute gases and potential options for additional utilities savings.
Demonstrating at G450C
A CVD (chemical vapor deposition) tool with its silane MFC set to greater than 2 slm was chosen for the demonstration. With a single chamber, the current integrated pump, dilution and abatement system (Zenith Flex – FIGURE 7), can provide the required performance and capacity. However, if a second chamber is to be added, a second abatement device would be required just to accommodate the extra nitrogen required for dilution. In this case, the opportunity presents itself to set up the installation with the traditional dilution and have the monitored connection option for cost savings and demonstration purposes without risking wafer test runs. If there is an unforeseen issue with the monitored connections, the additional N2 dilution and the required abatement is available.
Potential savings: Rough scenarios
In the case of a process tool using a 2 slm MFC for SiH4 and requiring 1⁄2 LFL to meet facility safety requirements and assuming a 96 slm pump purge, an additional 194 slm of N2 dilution is required. Using a conservative cost of $200,000 capital cost per unit for 600 slm of abatement capacity without accounting for additional footprint, maintenance, or operating costs, this represents $83,333 in apportioned abatement capital expenditure and instal- lation just for the additional N2 purge. Cost avoided in the first year is $88,000 and would likely be higher as abatement devices are supplied in discrete units and not in fractions. The additional cost of this monitoring system would be lower than the cost of the additional abatement capacity required with a N2 purge system and would easily accommodate flammable process gas flow changes.
Additional considerations include the protocol for what to do in the event of a drop in the monitored connection pressure and to where the information is sent:
- safety management system
- tool process gas panel to initiate immediate shut down
- advisory warning system
- or a combination of the above
A process may contain multiple flammable gases and/or an oxidizer. In this case, a monitored connection strategy might be used in conjunction with a reduced dilution targeting the oxidant.
In the photovoltaic industry, for example, operating systems already exist where dilution was not employed and exhaust pipeline pressure monitoring and bolted joints were used. Edwards has seen no adverse effects on a properly set up vacuum and abatement system at these facilities under standard operation.
Long exhaust lines can affect the gas velocity in the pipe and can cause by-products to solidify or precipitate out. In general, it is best to keep the exhaust line short and if necessary, heated to ensure that all process gases and by-products reach the abatement device for treatment. On some applications (TABLE 3) reduced dilution may lead to increased deposition in exhaust lines due to reduced gas velocity or increased chemical reaction rate. Where this is the case, exhaust dilution may be beneficial. However, the dilution factor is unlikely to be as high as is required to achieve 1/2 or 1/4 of the LFL.
As we build information from these case studies, lessons learned are codified in best known methods for process specific, integrated subsystem designs that provide the highest reliability at the lowest cost of ownership.
Based on increasing flow rates for flammable, pyrophoric and energetic gases, using the traditional N2 dilution to keep gases below their LFL may no longer be economically feasible for some processes and could pose as a non-starter from a facilities perspective. The collaborative demonstration at G450C proactively trials and examines a solution to reduce cost for higher flows which could also be used in 300 mm process applications. An integrated vacuum and abatement system provides the communications platform and optimized piping to ensure the best design. Following a successful implementation on the CVD application, we will seek to expand this option for savings to other processes.
This work was originally presented at the 2014 Advanced Semiconductor Manufacturing Conference.
Special Thanks to Frank Robertson and Ken Neff at G450C; Timothy Stoner at CSNE; and Julian Huang, Jason Holt, Al Brightman, David Hunt, Anthony Keen and Joey Pausic at Edwards.
1. B Corbin, A Pierce, C Bailey. “Rethinking the Approach to Higher 450mm Process Gas Flows: A Case Study”, ASMC 2014
2. PumpingFlammableGases.ApplicationsNoteP411-00- 090. Edwards 2009.
3. BLewisandGVanElbe.“Combustion,FlamesandExplo- sions of Gases.” (New York: Harcourt Brace Jovanovich,
4. Material Safety Data Sheet, TEOS. Air Products and Chemi-
Its Chlorides.” Plant/Operations Progress Vol 9 (1). P16-38.
NFPA 79, Section 9.4.3, “Control Systems Incorporating Software-and Firmware-Based Controls”. http://www.nfpa. org/codes-and-standards
7. 2013 VLSI Research Doc : 490112, v14.01. Doc : 490113,v14.01
ADRIENNE PIERCE is Director, Product Development, Edwards, Santa Clara, CA. CHRIS BAILEY is Global Technical Manager, Systems Engineering, Edwards, Burgess Hill, UK. WILLIAM CORBIN is the Infrastructure Supplier Manager, Tool Hookup PM, IBM Assignee, G450C Consortium, Albany, NY.