Detection and measurement of fluorocarbons is key to both process control and safety.
BY STEPHEN D. ANDERSON, Sensor Electronics Corp., Savage, MN
Fluorocarbons (FCs) are widely used in the semicon- ductor industry in dry processing applications such as film etching, chemical vapor deposition (CVD), chamber cleaning, and as coolants for semiconductor manufacturing tools. Although toxicity levels are not well established, many FC compounds are considered somewhat toxic. Many FCs are also significant green- house gases, while others are flammable. Detection and measurement of FCs is key to both process control and safety. Examples of commonly used FCs in semiconductor manufacturing are given in Table 1.
1. Although the table may list an FC a snon-toxic, many are heavy gases that can cause asphyxiation. Others can cause severe frostbite. Some produce toxic byproducts, such as CO or HF if heated or burned.
2. The naming of organic compounds is often confusing, especially as to whether something is a fluorocarbon (FC) versus a perfluorocarbon (PFC). The latter usually refers to a compound in which all of the hydrogens have been replaced with fluorines. Then, trifluoromethane, for example, is an FC but not a PFC. Note also that the last two table entries don’t follow this rule with respect to their common names.
3. The last two table entries have the same formula, C5HF7 , but much different structures and properties. In fact the first listed is a chain (aliphatic) compound while the second is a ring (aromatic). Always buy using the CAS Number and not the formula.
Among the available gas detection methods are:
- Catalytic bead – Generates heat when exposed to combustible gas.
- Electrochemical cell – Generates electrical current in response to specific gas.
- Photoionization detector (PID) – Ionizes gas using UV light, measures ion current.
- Pyrolyzer – Decomposes gas using heat, measures decomposition products.
- Infrared absorption (IR) – Gas blocks infrared path from source to detector.
- Metal oxide semiconductor (MOS) – Increases resistance in presence of gas.
To detect FCs, electrochemical cells are eliminated, since none are designed to sense FCs. The PID can also be eliminated because UV light used is not energetic enough to ionize FCs. Catalytic beads are poisoned by halogen compounds and shouldn’t be used.
The Pyrolyzer can measure FCs. But, since it destroys the gas being measured, it cannot distinguish one FC from another. It is also difficult to make the Pyrolyzer explosion-proof or intrinsically safe.
MOS sensors require ambient air to operate, are easily contaminated, and are not specific.
IR alone has the ability to sense a specific FC gas.
The F-C bond
The common feature of FCs is the carbon-fluorine bond. The stretching vibrations of this bond result in infrared (IR) absorption at wavelengths ranging approximately from 7 to 10 micron . The precise wavelength of absorption varies with the overall molecular structure, and is given in TABLE 2.
For example, measuring the absorption at 10.4 micron can tell us how much C4F6 is present.
An entire branch of chemical study deals with deter- mining structure from absorption bands and vice-versa.
Infrared spectrometers, commonly used in these studies, have the ability to sweep through many wavelengths, looking at absorption versus wavelength.
The technique of gas detection by measuring absorption at one wavelength is termed NDIR (non-dispersive infrared). Non-dispersive means that a particular wavelength is selected using a fixed optical filter, in contrast to the variable mechanical filter used in an IR spectrometer. An NDIR is less flexible than an IR spectrometer, but has the advantage of no moving parts or complex optics, making it ideal for industrial environments.
The principle of the NDIR is illustrated in FIGURE 1. The IR spectrum at specific points in the NDIR device is included (spectrum plots 1 – 4). The plots assume that target gas is present and absorbing at 7 micron.
The IR source (spectrum 1) is a Graybody source (the term applied to a bsource with an emissivity less than 1), which provides IR light across a range of wavelengths from about 1 to 15 micron. A range of wavelengths is desired so that one source can be used to sense a variety of gases.
The light from the source passes through the target gas in a “waveguide” – a reflective chamber open to the atmosphere. The spectrum at the saveguide output (spectrum 2) shows a notch (attenuation) at 7 micron, due to absorption by the target gas.
The light is next applied to optical filters – the wavelength-selective parts of the NDIR. Each optical filter is a narrow bandpass filter, made from a window with various coatings that “create” optical interference except at wavelengths of interest. A typical filter bandwidth is 0.25 micron.
The target filter passes light in the range of about 6.9 micron to 7.1 micron, resulting in spectrum 3, which shows the effects of both the gas and filter. The NDIR usually includes a reference optical filter – a filter that passes light where the target gas is transparent – as a means of maintaining a fixed gain or sensitivity in the presence of varying Source light levels. In the example, the reference filter at 3 micron passes light in the range of about 2.9 micron to 3.1 micron, resulting in spectrum 4 (Note: The filter bandwidths in Fig. 1 are wider than actual for purposes of illustration).
The outputs of the two filters are applied to separate detectors (bolometer or thermopile) and converted to electrical signals.
The IR source is usually driven by a square wave to create a modulation in the IR output. The resulting AC signal allows for easier removal of offset and drift. For the highest possible modulation frequencies, newer MEMS- based sources with extremely small thermal mass, have recently become available.
Note that an NDIR can be created by other means. For example, the source might be an IR laser diode or IR light- emitting-diode. These alternate sources are generally at a disadvantage to the thermal source because of their limited bandwidth. Tunable laser diodes exist but are very expensive at present. Also, at the receiving end of the light path, an NDIR might use one or more photo- diodes rather than thermal detectors.
The light absorption by the target gas is exponentially related to gas concentration. This non-linearity is removed using a microcontroller algorithm that is generally different for each FC and each concentration range.
The NDIR also measures temperature and pressure, allowing for ideal gas law correction, which is used when the measurand is concentration rather than density. Temper- ature measurement also allows for temperature compensation of zero and span.
An NDIR device, as described here, typically measures concentration in the range of 100 ppm to 1000 ppm (by volume) or density in the range of 200 to 2000 milligram / liter. Typical accuracy is 5% of range. Special designs with long optical path length are now available for smaller concentrations.
NDIR maintenance is generally limited to periodic calibration of zero and span, and keeping the optical surfaces free of dust and obstruction. Because the NDIR
measures transmittance, it is inherently fail-safe:
No light = Lots of gas (or obstruction or Source fail) = Alarm
The simple and rugged optical system keeps unit cost and maintenance low, while increasing reliability.
Several FCs are liquids at room temperature or have boiling points close to room temperature. NDIRs are ideal at sensing the liquid vapor, since the NDIR is easily made a part of the calibration setup. All that’s needed to calibrate is a controlled temperature and a table of vapor pressure versus temperature at equilibrium.
The NDIR is proving to be a the instrument of choice in detecting fluorocarbons. Its main features of selectivity, mechanical ruggedness, and operational simplicity are pushing aside other detection methods. Future NDIRs are expected to further this trend with improved ability to pinpoint a given FC gas in the presence of industrial cleaners and other interfering products.
STEPHEN ANDERSON, is an engineer at Sensor Electronics Corp., Savage, MN, phone 952-938-9486. He has a B. Chem and MSEE, both from the University of Minnesota and has been active in the process control industry for 35+ years.