February 1, 2011 — Sanjay Yedur et al from Pivotal Systems and J.H. Lee et al from Samsung’s R&D Equipment Engineering Team, discuss the use of a real-time gas flow monitoring system that allows for in-situ flow measurements, based on a highly accurate rate of pressure drop over a known volume and temperature. Using this system, insights are gained on the run-to-run repeatability of mass flow controller (MFC) transient and steady-state flow during wafer processing.
MFC transient and steady-state analysis
As process geometries within the semiconductor industry continue to shrink, the need for accurate and repeatable gas flow control during wafer processing is essential. To date, most wafer fabs run periodic offline calibrations to ensure average flow meets the typical mass flow controller vendor specification of ±1% of set point. This independent verification is performed using industry-standard techniques, such as a rate of rise pressure measurement into a known volume and temperature, and typically takes 30-60 seconds for an accurate measurement.
With the emergence of low gas flow rates (as low as 2sccm), short processing times (as short as 2s), and continuous plasma processing (which removes gas flow stability steps), run-to-run repeatability of MFC transients and steady-state performance during wafer processing is quickly gaining importance. MFC transient performance is defined as the actual flow pattern (settling time and overshoot/undershoot) that occurs in the first several seconds after a set point change command while steady-state flow is the average flow rate achieved once the transient regime is completed. As chip fabs push beyond 40nm and drive semiconductor device dimensions to atomic levels, run-to-run variations in these two MFC characteristics can adversely affect critical process parameters like CD and film thickness. Variability in MFC performance from chamber to chamber can also exacerbate chamber matching issues.
As it pertains to transients, MFC vendors primarily focus on maximum settling time – the maximum time required in milliseconds from set point command to achieve the specified flow with accuracy of ±1%. Typical vendor specifications for MFC settling time range from 500 to 1000 milliseconds. MFC vendors however do not provide any specifications on the maximum or average flow overshoot/undershoot during the transient or any understanding about the repeatability of these transient parameters (e.g., standard deviation).
Such transient behaviors are difficult to measure independently, given the time response required to accurately see how the MFC performs during a set point change. For example, even some of the most accurate and sensitive gas flow verification techniques such as fixed volume rate of pressure rise and fixed orifice pressure differential meters are not equipped to provide the time resolution in flow rate measurement required to analyze MFC transient behavior.
As it pertains to steady-state performance, all known methods, including the industry standard pressure rise measurement of gas flowing into a known volume, are “off-line” approaches, thereby precluding any possibility of understanding, monitoring and controlling against wafer level run-to-run gas flow variations.
As a result, fabs are effectively blind to this important set of flow transient and steady-state parameters, relying only on the MFC’s own feedback signal, which can often be erroneous and prone to drift. Within advanced etch or atomic layer deposition (ALD) processes for example, plenty of time and resources can be spent troubleshooting variability issues that are due to transient variations.
Real-time gas flow monitoring system with high sampling rate
One approach to address in-situ MFC transient and steady-state analysis is use of a real-time gas flow monitoring (GFM) system with a high data and measurement sampling rate. This system measures an induced rate of pressure drop within the gas stick upstream of the MFC over a known volume and temperature during wafer processing. The method is based upon first principles, specifically:
PV = nZRT
Where:
- P is the absolute pressure of the gas
- V is the volume of the gas
- n is the number of moles of the gas
- R is the universal gas constant
- T is the absolute temperature
- Z is the compressibility factor
Therefore: n = P V/ZRT
And finally: flow rate = Δn = ΔP V/ZRT
As such, this system offers a primary standard, NIST traceable measurement of gas flow. In accuracy tests (conducted at Samsung’s R&D center and its Gas Calibration and Standards Lab) comparing this system’s steady-state flow measurement to other gas flow verification systems, a tight correlation was demonstrated with a Molbloc system, with its accuracy of ±0.2% of actual flow (Fig. 1). Furthermore, a 50ms sampling rate allows the real-time gas flow measurement system to easily characterize MFC transients that occur within the first several seconds of set point change.
 |
| Figure 1. The GFM system (steady-state flow measurments) demonstrates a tight correlation to a Molbloc system (±0.2% accuracy). |
Because the system measures a rate of pressure drop upstream of the MFC (Fig. 2), it is measuring the gas flow in (FI) to the MFC at any given point in time vs. the gas flow out (FO) of the MFC, which is the important parameter concerned. For thermal-based MFCs, during both transient and steady-state regimes of flow:
GFM flow measurement = FI = FO
For pressure-based MFCs however, due to the internal MFC volume that is regulated at a different pressure, the above condition is only applicable during the steady-state regime of flow. Therefore, while the gas flow measurement system can accurately report overshoot/undershoot and settling time behavior for a thermal-based MFC, it can only accurately report settling time behavior for a pressure-based MFC.
 |
| Figure 2. The GFM system measures a rate of pressure drop over a small isolated volume upstream of the MFC. As such it is a measurement of the flow in (FI) to the MFC. |
The system (Fig. 3) consists of a typical gas stick — where a wetted part replaces the existing regulator and pressure transducer (combines the function of regulator, pressure reading, temperature reading, and NIST-traceable volume measurement) — and a control/data collection system.
 |
| Figure 3. The gas flow monitor (GFM) system includes a replacement wetted part on a typical gas stick connected to a control/data collection system |
Every time an MFC changes its set point, specifically turn-on from zero flow to a higher flow; step-up from lower flow to a higher flow; and step-down from a higher flow to a lower flow, the system immediately measures the induced pressure drop during this flow transient regime and displays the flow pattern. The measurement approach is transparent to and does not affect MFC performance. Both pressure insensitive (PI) and non-PI MFCs have been tested to show them well-equipped to handle this small induced pressure drop from the system. Armed with this flow transient information, an MFC’s settling time, overshoot/undershoot, and steady-state flow behavior can be independently verified.
Sample MFC transient findings
The real-time gas flow measurement system was recently used in Samsung’s R&D center and Gas Calibration and Standards lab to better understand MFC transient behavior. Two leading MFCs were selected: a thermal-based MFC with full scale of 100sccm and a pressure-based MFC with full scale of 400sccm. The MFCs were installed on two gas sticks and run through two tests: Staircase step-up flow from 0% to 40% to 60% to 80% to 100%, and Turn-on cycling from 0% to 50% cycled 150 times to mimic a production environment.
Both MFC feedback and the gas flow measurement data were captured for these experiments. For the thermal-based MFC, statistics on flow overshoot (maximum flow rate) and settling time (time to reach ±1% of target set point flow) were computed for the cycling data; for the pressure-based MFC, statistics only on settling time were computed. Some sample findings of MFC transient performance as measured by the gas flow measurement system are described below.
MFC feedback signals mask transient activity. This is especially true for thermal-based MFCs where the initial overshoot spike observed by the gas flow measurement system is not observed in the MFC feedback signal (Fig. 4)
 |
| Figure 4. MFC feedback signals mask transient activity measured by the GFM system. |
Step-up overshoot from zero flow is larger than from a non-zero flow. Moving from a zero flow to a non-zero flow vs. moving from a lower flow to a high flow places greater demands on the flow measuring and valve control systems within the MFC. As indicated in the thermal-based MFC staircase testing (Fig. 5), the gas flow measurement system observes a larger transient overshoot from 0% to 40% than for example, 40% to 60%.
 |
| Figure 5. Staircase flow pattern measured by MFC feedback and GFM. GFM observes significant transient behavior when moving from a zero flow position. |
MFC transients exhibit a wide range of variability from cycle to cycle. When cycled 150 times between 0% and 50% flow, both the thermal- and pressure-based MFCs demonstrated a wide range of variability on overshoot as well as settling time (Fig. 6). The thermal-based MFC had overshoots that ranged from 7% to 17% of the 50sccm target set point with settling times that ranged from 0.7 to 2.85 seconds. Similarly the pressure-based MFC had settling times that ranged from 0.5 to 2.90 seconds.
 |
| Figure 6. Transient repeatability is studied in the above figure for both thermal and pressure-based MFCs. In both cases for example, GFM observes settling times greater in average and standard deviation than the MFC feedback. |
Sample MFC steady-state findings
Steady-state gas flow measurements were also taken during in situ wafer processing on a leading oxide etcher platform. O2 and CF4 flow rates were of special interest given their sensitivity to critical wafer parameters such as CD.
In-situ steady-state flow rates are often out of specification. Out of specification steady-state flow was observed in many of the various flow set points for O2 and CF4. In one particular case the system indicated an O2 MFC was flowing about ~5% lower than the expected set point of 30sccm (Fig 7.) The MFC’s flow feedback however did not report such a large deviation in flow rate. An off-line verification with Molbloc confirmed that the MFC was performing out of specification.
 |
| Figure 7. Steady-state flow measurements by GFM on an O2 MFC at 30sccm set point during wafer processing. GFM observes an out of specification flow for each run that is about 5% less than targeted flow set point. |
|
FDC SetPoint-30.0-Total run#15 |
Conclusion
The GFM system sheds valuable insight on a critical regime of the MFC’s behavior that has been little understood in an in-situ production environment. For example, the system is being used to better understand transient topics such as MFC step-downs, transient behavior as a leading indicator for MFC failure, and head-to-head MFC vendor comparisons on transient repeatability. Additionally, deploying this real-time gas flow measurement system in a production environment is being pursued due to increased accuracy and frequency of measurement this system offers over traditional off-line methods such as chamber rate-of-rise. In particular, given that this system leverages existing gas stick components like regulators, isolation valves, and pressure transducers, implementing this measurement scheme as an upgrade to tool controllers with minimal additional hardware is a natural step.
Acknowledgements
GFM is a trademark of Pivotal Systems.
Molbloc is a trademark of DH Instruments.
Sanjay Yedur is Director of Marketing & Applications at Pivotal Systems, 4683 Chabot Drive, Pleasanton, CA 94588; ph.:+1-925-924-1480 ext. 234; [email protected]
Arvind Sankaran, Ray Malone and Renny Reed are Field Application Engineers at Pivotal Systems
Mukund Venkatesh is VP of Marketing at Pivotal Systems
Jae Hun Lee is Senior Engineer in R&D Equipment Engineering Group at Samsung’s R&D Center
Yeong Hun Han is Senior Engineer in the R&D Equipment Engineering Group at Samsung’s R&D Center
Ki Young Kim is Field Support Engineer in the R&D Equipment Engineering Group at Samsung’s R&D Center
Sung Ho Han is Manager in R&D Equipment Engineering Group at Samsung’s R&D Center
Subscribe to Solid State Technology/Advanced Packaging. Follow Solid State Technology on Twitter.com via editors Pete Singer, twitter.com/PetesTweetsPW and Debra Vogler, twitter.com/dvogler_PV_semi. Or join our Facebook group