The control of gases to a process chamber seems like a straight forward task to accomplish. However, the problems encountered can have serious results. Improper flow control can cause rejected product, longer than expected processing time, or even catastrophic system failure due to improper system pressure or leaks. In general, mass flow control problems can be separated into three types, (1) Intrinsic to the flow control device, (2) Systematic problems, and (3) Environmental effects.

Environmental effects are the easiest to identify and correct. Examples of environmental effects are changing thermal environment, vibration, exposure to high levels of RF Interference, etc. The identification and removal of such existing conditions will make it easier to identify and correct problems in the other two categories.

Systematic problems are the hardest problems to separate from others. Often they do not exist except under certain process conditions. Systematic problems are often mistaken for problems intrinsic to the flow controller. Some common systematic problems are changing or unstable system pressures, changing process gas temperatures, valve sequencing and timing, system pumping speed, process gas quality, etc. All of these factors have an effect on the gas delivery devices. Understanding the effects can allow the process engineer to eliminate or compensate for those that have detrimental effects on gas delivery.

The flow controllers available from all manufacturers share some critical intrinsic characteristics and therefore problems. These are due in large part to the sensing technique and bypass arrangement used to allow different flow rates of instruments to be produced. The most significant intrinsic performance problem areas:

• Linearity and performance problems associated with Gas Correction Factors used in calibration for gases other than the calibration gas.
• Zero drift
• Free convection in the sensor (thermal siphoning)
• Controller oscillation
• Seal degradation and leaks
• Overshoot

Other problems reported with repeatability or linearity are usually manifestations of one or more of the problems specified above.

With the understanding that gas delivery problems come from several areas, the approach to their solutions must encompass the areas identified and be rigorous in nature. In the pages that follow, a logical approach to identifying and solving many common mass flow control problems is presented. Following this approach will at least ensure the process engineer that he has looked beyond the flow control device to address the desired gas delivery.

Avoiding Problems

As with most problems, those occurring in flow control can be diminished greatly through understanding the operation of the device. This will allow the process engineer to adjust for the limitations of the flow controller and the manufacturer to understand the most critical functions required of the device.

Sensor Operation: Set Point Repeatability - Accuracy - Zero Drift

These performance functions are lumped together because they are interrelated in the flow controller design. A brief and generalized explanation of the operation of the thermal mass flow sensor element will aid in illustrating the relationships and inherent shortcomings of the devices.

Theory of Operation - Sensor

Two physical components make up the sensing element of a flow controller, (1) a thin-walled small diameter stainless steel tube with multiple high temperature coefficient windings applied to the external diameter, and (2) a secondary flow channel or bypass through which the bulk of the gas flows. Since the inlet and outlet of both components are coincident the pressure drop across each is equivalent. In theory, flow split ratio remains constant through the entire flow range of the instrument. The total flow through the instrument is inferred by measuring the flow through the sensor tube and then adding the amount (determined by the ratio) flowing through the bypass. The amount of gas flowing through the tube is measured by excitation of the sensor tube windings with electrical current and noting voltage differences that occur as flow is varied. These voltage differences are then referenced to the aero-flow voltage difference (if any) and used as a direct indication of flow. Since a gas property dependent upon mass rate of flow is inferred (heat capacity) this voltage output is proportional to mass flow of the gas.

Inherent Limitations

Because the flow sensing technique is thermal in nature, all heat transfer modes affect the flow indication. The heat transfer modes that exist in the sensor are:

• Free convection, both to the sensor environment and to the process gas at zero flow
• Conduction along the tube, and to insulating media
• Radiation heat transfer

Changes in any of the above quantities can have an effect on the performance of the MFC. The largest detrimental effect is from a change in free convection to or from the environment. This can be caused in several ways, one of which is an internal (to the MFC sensor cover) change in temperature or temperature gradient caused by the heating of the valve or an electrical component. This temperature change can cause the natural zero to shift resulting in corresponding shift in calibration.

Another limitation of this sensing technique, inherent in flow controller design is aging of the sensor. Fine resistance wire is wound on the tube in a specific pattern and with a specific tension. Through expansion and contraction caused by heating and cooling of the sensor tube and wire elements these patterns and tensions change. These changes can cause a zero drift or even a change in sensitivity of the sensor, due to the change in heat transfer quantities from the moving or sifting of the heating element/temperature sensing wire.

Free convection to the process gas at zero flow (thermal siphoning) has been recognized recently as a significant problem. This problem is reported virtually exclusively in flow controllers that are mounted vertically in a system. A zero shift is observed due to heated gas rising in the sensor tube and cooling upon exit. This sets up a real flow through the sensor and the bypass and is not an indication that the natural zero has changed. The danger here is that if this non-natural zero output is adjusted, a calibration error will result.

Minimizing Inherent Problems

Both the manufacturers and users of mass flow controllers must recognize the factors that minimize sensor drift as stated above. Most major MFC manufacturers have now realized the value of extended or accelerated burn- in of their sensor prior to assembly into the controller. Some manufacturers are burning-in sensors for as long as two weeks prior to assembly. Many manufacturers have introduced circuity features such as temperature compensation and auto-zero circuits that attenuate or mask internal thermal effects. These work well for the most part with one outstanding limitation: auto-zero circuits will create a calibration error in virtually all cases when used to zero the drift associated with thermal siphoning.

MFC users can enhance the performance of the flow controller by doing several things. First, always allow a flow controller to warm-up thoroughly. Second, where possible do not adjust the natural zero of the flow controller but instead record it and adjust as necessary in external control electronics. Last, if space allows, do not mount the flow controller in a vertical position. If this practice is avoided, thermal siphoning should not be a problem. If vertical mounting is unavoidable, do not adjust the no-flow, zero, output.

Environmental and Systematic Effects

Several environmental and systematic conditions will have an effect on the performance of the sensing element. Many of these conditions can be eliminated through system design and others reduced to the point where they are manageable.

Again, because the sensing technique is thermal, the operation of the MFC is influenced by changing temperatures of its environment. If the temperature in which the MFC operates changes significantly, the zero output of the device may seem to track these changes. This will result in calibration shifts and may show up as a repeatability problem.

Because of the construction of the sensor element (fine wires wrapped around a capillary tube), it acts as a very effective antenna. When exposed to high levels of Radio Frequency Radiation, the output of the sensor can be effected. The sensor has an inherently low raw signal to noise ratio, so when this interference is received it is amplified along with the flow signal. This can create a control band around the desired set point that is not as precise as desired and can create controller oscillation that renders the device useless.

Valve sequencing can cause many of the systematic problems encountered with flow control. A common scenario is as follows: during a manufacturing process, a pneumatic on/off valve external to the MFC is closed while a non-zero set point is applied to the MFC. The MFC internal control valve will open to its maximum in an effort to maintain the desired flow through the MFC. The system is then pumped-out until a specified vacuum is reached and the process is then restarted by opening the pneumatic on/off valve. Extreme overshoot with possible oscillation and long settling times can result. This is due to the change in system pressure and gas composition within the flow controller itself. Further, with the set point left on , the control valve will respond by driving full open, creating more internal heat. All these factors work to change the thermal equilibrium and therefore change the natural zero of the MFC. As the process is restarted, the thermal equilibrium of the sensor is upset once again and the oscillation and overshoot mentioned above may occur.

Another effect that has been erroneously discounted is changing process gas (supply gas) temperature. These effects are seen more on a day-to-day rather than a run-to-run basis and are often mistaken for long-term drift of the MFC. To understand the mechanism that accounts for this calibration change, we must remember that the sensor element is actually two components, the sensor tube and the MFC bypass. The ratio of the flow passing through this combination is dependent upon their hydraulic diameters providing that the Reynolds’ Numbers in each are compatible and maintain a consistent ratio. The Reynolds’ Number is a function of velocity, diameter (or hydraulic diameter), density, and kinematic viscosity. Density is extremely temperature dependent for most gases used. This means that if the density of the gas passing through the tube varies disproportionately to the density of gas passing through the bypass, the bypass ratio and therefore the calibration will change. This will happen if the supply gas temperature changes. With up to 10 degree swings in temperature, the bulk average temperature of the gas passing through the sensor tube remains constant (the capillary tube is a very efficient heat exchanger), but the gas passing through the bypass retains the variation in temperature and thus density, creating the density change that effects the bypass ratio and therefore calibration.

Process gas composition and quality, especially the presence of moisture, has an effect on both calibration and performance of an MFC. With varying compositions, the heat capacity of the gas changes and the calibration will be effected. Moisture can have the same effect, or worse, it can cause acids such as HCl to form in the flow controller. These acids can degrade seals and create leaks or have a more insidious effect of combining with other compounds to form clogs inside the sensor tube or bypass. This can cause very large changes in the MFC calibration (sensitivity) or the MFC performance (speed of response).

Eliminating Environmental and Systematic Effects

Keeping the MFCs in an environment maintained at a relatively constant operating temperature and one that acts as a shield against RF Interference is the first and easiest step to take toward ensuring proper and consistent flow controller operation. Further, properly grounding the cable shield and isolating the electrical common from RF interference will achieve notable results.

Sequencing valves to keep the flow controller exposed to the process gas at relatively constant pressure will help eliminate zero-shifts , overshoot and oscillation.

Conditioning of the process gas will greatly enhance the performance of the gas delivery system. Gas purity should be monitored, with filters and dryers installed to eliminate moisture and other unwanted trace elements. Also, heat exchangers in the supply line used to control the temperature of the supply gas to within reasonable limits of +/-2 degrees will improve day-to-day repeatability of the gas delivery system.

MFC Operation: Valve-Oscillation-Inadequate Flow-Speed

A thermal mass flow controller is a complete closed-loop feedback (or feed-forward) system. This is a very complex system dependent upon the operation of discreet components with varying performance characteristics and time constants. Signal amplifiers, linearizers, PID circuits, valve drive circuits, and constant current generators are among the components found in these small packages. Though the manufacturers specify large operating ranges of temperature, pressure (and differential pressure), gases, and flow rates, they do not guarantee that all performance specifications are valid under all conditions. Because of the dependence of operating specifications on operating conditions, users should communicate as much information as possible about the actual operating conditions of the MFC to the manufacturers. If this is done, optimum performance can be tuned around these conditions. The use of generic flow controllers will not assure optimum performance. (A generic mass flow controller is one that is provided by the supplier with a nitrogen calibration, or another non- actual system gas, and no operating pressures, temperatures, or orientation specified.)

Intrinsic, Systematic, and Environmental Problems

The performance of the MFC, as a system, makes it impossible to decouple the three types of problem causing effects. Fortunately most are easily understood and corrected.

The mass flow controller valve is, in general, made up of (1) a motive force (electromagnetic, thermal, piezo- electric, motor-driven, etc.), (2) a resistive force (spring, diaphragm, etc.), and (3) a variable flow restriction (orifice-poppet, orifice-ball, etc.). The motive force is activated by the control circuitry to act against the resistive force to vary the flow restriction. Virtually all of the pressure drop across the flow controller takes place across the control valve.

Proper performance of the valve is determined by all of these components being sized properly for the intended actual gas used, range and differential pressure of the system. Problems can occur in several ways. First, the valve restriction or orifice must be sized correctly. An orifice that will pass 100 sccm of Nitrogen when fully opened at 30 PSID may not pass 100 sccm of Nitrogen at 5 PSID. Further, this orifice may not pass 100 sccm of WF6 at either condition. This is because the orifice is too restrictive for the heavier gas. Oscillation can also occur if the resistive force or valve spring is too heavy or too light for the application. A spring with an appropriate spring rate for nitrogen may work fine if used for BCl3 or Argon but can cause oscillation in Hydrogen or Helium.

The differential pressure across the flow controller can be created by elevated upstream pressure or by the creation of a vacuum downstream. This can have a substantial effect upon system performance. As discussed, the exposure of the sensor to vacuum can have performance effects. Further, system time constants such as sensor speed of response change radically under these conditions. The PID control circuitry may not have the margin to provide optimum response under both conditions and possibly must be tuned to the correct application or oscillation or long settling times may result.

Avoiding the Problems

The above problems can virtually all be eliminated by characterizing the gas delivery operating conditions and communicating them to the manufacturer or service facility when ordering a flow controller or having one serviced. Further, never apply a generic flow controller to a critical requirement; always have it re-tuned and properly labeled for the specific application.

Minimizing the changes seen by the flow controller will also help eliminate systematic problems. Keeping the controller pressurized with the process gas can eliminate oscillation and over-the-hill conditions that can occur.

Other Applications Problems

Long Cable Lengths

Because many flow controllers do not draw the same current from the positive and negative supply voltages, current is present on the reference common. For short (low resistance) cable lengths this is not a problem. In applications with longer cable lengths having significant wire resistances, the effect may be that the displayed output signal and the set point signal will not match.

Gas Correction Factor

Due to internal geometries and operating regimes (temperature, Reynolds’ Number, etc.) gas correction factors vary from MFC manufacturer to manufacturer. Further, the accuracy of some gas correction factors is no better than +/-10% and is flow dependent. The derivation of gas correction factors is the subject of another complete paper but, the recommendation is to use the manufacturer specified factor and take steps to ensure absolute repeatability, not accuracy.

Summary

The performance of thermal mass flow controllers is affected by many factors. These have been divided into intrinsic, systematic, and environmental effects. The best way to ensure the performance of the gas delivery system is to minimize the causes for error. The major actions that can be taken to eliminate flow control problems are listed below.

• Choose MFCs from manufacturers that burn-in the sensors prior to assembly.
• Always allow MFCs to warm-up completely under approximate operating conditions before recording the natural zero or flow values.
• Where possible, do not adjust MFC calibration pots in-situ; rather record output changes and compensate externally, by changing the set point to the MFC.
• Where possible, prevent thermal siphoning by not mounting the MFC in a vertical position. When vertical mounting is unavoidable, do not activate an auto zero circuit or manually adjust flow controller zero output in-situ.
• Keep flow controllers in a thermally stable or controlled environment.
• Properly shield flow controllers and cabling from RF interference, and isolate circuit common from case ground.
• Sequence shut-off valves to maintain a stable pressure condition and gas composition if possible in the flow controller.
• Supply clean, pure, and dry process gas to the flow controller.
• Control the temperature of the process gas to the flow controller from run-to-run and day-to-day.
• For all critical applications, use an MFC calibrated and tuned for the actual gas and range to be used; do not use a generic MFC.
• Do not expose MFCs to sudden pressure surges or leave them with no process gas pressure with a non-zero set-point commanded.
• Avoid long cable lengths with flow controllers drawing power unevenly from the power supply rails.
• When a gas correction factor is necessary, always use the test gas and CF value specified by the flow controller manufacturer.
• Always communicate as close to the actual operating conditions as possible to the MFC manufacturer or service facility when ordering an MFC or having an MFC serviced.

Most flow controller problems can be eliminated or diminished significantly through understanding the effects on these complicated devices. Communication of the most critical process parameters to the vendor is the best way to ensure that the flow controller will perform as required. As always, with any equipment you operate, read the operating manuals before using the equipment!