Gas Filter Correlation Techniques

header

[Missing Image]

Peter Middleton
Servomex Ltd.,
Crowborough,
East Sussex,
TN6 3DU, UK

Abstract

Infrared gas filter correlation techniques have been developed to allow highly selective measurement of a range of gases (NO, CO, HCl, CO2, SO2 , N2O and CH4). The methodology is described in detail together with the various ways of optimising the design parameters to meet the performance requirements of the application. The advantages over other techniques are described and the resulting performance figures presented.

Introduction

The improved performance that gas filter correlation (GFC) conveys has allowed infra red (IR) photometry to displace more complex measurement technologies over the past decade, at lower cost. The earliest description of the technique was by Goody [1] and its early practical use for pollutant analysis with folded path cells was typified by the work of Chaney and McClenny [2]. A review of the various established techniques used in absorption photometer design has been presented by Hanst [3]. This paper reports the development of a practical GFC photometer without the need for a folded path White cell [4] as shown below.

Principle of operation

Broad band IR radiation from a hot element source passes alternately through one of two gas filters mounted on a chopper wheel (Figure 1). One contains nitrogen and the other contains the gas of interest, for example nitric oxide (NO) for a NO analyser. When the nitrogen gas filter is in position, no absorption takes place. When the NO gas filter is in position, absorption takes place reducing the intensity in the beam at the characteristic wavelengths for NO.

Figure 1. Schematic of the Servomex Gfx1210 series gas filter correlation transducer, showing the detector output before and
after electronic zero adjustment with nitrogen in the sample cell.

The temporally spaced beams then pass through a narrow band pass filter which limits the IR region to a specific part of the absorption spectrum and then into the sample cell (Figure 2).

When radiation that has passed through the nitrogen gas filter passes through the sample gas containing NO, absorption occurs according to Beer’s law and produces a reduction in the detector signal. The transmitted energy of the measure beam can be described:

Em = M ò Is (l ) Tf (l ) exp(-ccell lcell aNO(l )) dl (1)

where Is(l ) = Source intensity
Tf (l ) = IR filter transmission
aNO(l ) = absorption coefficient for NO
ccell = sample cell gas concentration
lcell = sample cell path length

M describes the optical path relating transmitted intensity to energy throughput. If we assume it is independent of wavelength and equal for both beams, it becomes a constant.

Radiation that has passed through the NO gas filter has had the intensity at the characteristic wavelengths significantly reduced and so further absorption in the sample cell is small. However, the energy which is transmitted makes an excellent reference signal as it has the same spectral distribution as the measure signal.

Figure 2. The mid IR spectrum of NO (10 vpm.m) at 0.2 cm-1 resolution
and a typical narrow band pass filter used for
defining the region of interest for gas filter correlation.
Water vapour is also shown (top spectrum).

Assuming Beer’s law is obeyed in the gas filter, the energy falling on the detector can be described:

Er = M ò Is (l ) Tf (l ) exp(-cgf lgf aNO(l )) exp(-ccell lcell aNO(l )) dl (2)

where cgf = gas concentration in gas filter
lgf = gas filter path length

Figure 3. The spectra of the reference (top) and
measure (bottom) beams after passing through
a 500mm sample cell containing 300vpmNO.
Much of the characteristic energy has been
absorbed from the reference beam by the
gas filter so that little further absorption takes
place in the sample cell.

The spectra of the two beams at the detector with 300vpm NO in the sample cell are shown in Figure 3. The resulting detector signals can be represented by the total energy from each beam and an output proportional to NO concentration can be determined from the difference between these energies. Dividing by the reference energy provides the necessary normalization:

ccell µ (Er – a Em ) / Er (3)

where a = Er / Em when ccell = 0

The factor a is effectively a coarse instrument zero which may be effected optically or electronically. However an optical method, such as masking the nitrogen gas filter, may detract from the spectral similarity of the two beams.

Calibration and linearization provide the necessary mapping to experimentally measured values to give the NO sample cell concentration from the GFC photometer.

Advantages of gas filter correlation

One of the features of GFC that conveys many advantages over other photometric techniques is that only a single IR region is required for the measurement. That is, a single narrow band IR filter defines the region of interest. Any changes in spectral content due to source ageing, window contamination and any other broad band interference occur equally for both measure and reference beams and thus cancel out in the differential measurement.

The greater the spectral similarity of the two beams, the greater the immunity from interferences, as both would be affected equally by any absorption. As one beam must pass through the gas filter, this is best achieved by gases which exhibit widely spaced rotational fine structure of which NO is a prime example (Figure 2). This feature is usually referred to as the common mode rejection ratio and figures approaching 1:104 have been measured for variations in source energy.

Figure 4. The spectra of HCl and water vapour showing
a typical position of the IR filter to avoid cross
sensitivity to water vapour.

Heterogenous diatomic gases exhibit rotational fine structure from the splitting of their vibrational transitions giving the desired optimum spectra for GFC. Fortuitously these gases (CO, NO, HCl, HF) are also the gases we are interested in for CEM applications in particular, where insensitivity to large varying background concentrations of water vapour and carbon dioxide is required. This gives GFC its suitability for CEM applications where high selectivity and inherent immunity to interferences are major advantages of this technique over more complex compensating spectroscopic techniques.

A background gas with an absorption in the same region of interest may interfere only if there is direct overlap with the narrow gas filter absorption lines of the target gas. This immediately reduces the likelihood of interferences. Then by careful selection of the IR interference filter parameters to exclude such regions of overlap where possible, very low cross sensitivity can be achieved. This is demonstrated for the case of HCl (Figure 4) where the spectrum allows a choice of P or R branches of the absorption. By choosing the lower energy P branch, avoiding the water vapour absorption, a suitable IR filter can be chosen. For maximum sensitivity the gas in the gas filter should be in a similar dynamic molecular state (temperature, pressure) as the gas in the sample cell when maximum correlation between the gas spectra occurs. This is because both line intensities and line widths are affected by the molecular dynamics of the absorbing gas.

Figure 5. The theoretical cross sensitivity of a NO gas filter
correlation measurement to water vapour.
By comparing these results with the NO relative
sensitivity (dashed), the optimum value of CWL
can be found for a given IR narrow band pass
filter and NO gas filter. These results will depend
on IR filter parameters including bandwidth and
asymmetry and other gas filter and instrumental parameters.

A further method of reducing cross sensitivity is to utilise a zero crossing point for cross sensitivity with respect to centre wavelength (CWL) of the IR filter. For example, for NO there are a large number of water vapour lines across the 5.3µm region. Some of these lines overlap the NO lines exactly, giving a positive cross sensitivity and some fall exactly between the NO lines giving a negative sensitivity to water vapour. As the CWL of the IR filter is changed from one such region to another, the cross sensitivity falls to zero where the contributions from overlapping and non-overlapping lines cancel. By including the absorption term product for the cross interfering gas in both integrals (1) and (2),

exp(-ccell lcell aH2O(l ))

where aH2O(l ) = absorption coefficient for H2O

the relationship between cross sensitivity and IR filter CWL can be found which allows the estimation of the value of CWL for zero cross sensitivity. Figure 5 shows the results of a theoretical determination of NO sensitivity and H2O cross sensitivity as the CWL of the filter is varied across the region. This helps determine the optimum filter parameters for maximum sensitivity and selectivity. It can be seen that for a particular set of filter parameters a maximum in sensitivity occurs near a CWL at which positive and negative contributions of sensitivity to water vapour cancel. By finely specifying the filter parameters it is possible to manufacture 0-100vpmNO transducers with sensitivity to water vapour less than 2 vpmNO / 0.5%v/v H2O.

Figure 6. The spectra of SO2 (centre), water vapour (left and centre)
and a typical narrow band pass IR filter (centre),
showing the difficulty in avoiding cross sensitivity to water vapour.

A similar approach is necessary for SO2 where again water lines cross the 7.5µm region (Figure 6). Moreover, the triatomic SO2 molecule gives many closely spaced absorption lines which is not as ideal as the spectrum from a heterogenous diatomic molecule. This results in less inherent immunity to interference requiring more precise optimisation of the filter parameters.

Figure 7. Cross sensitivity simulations for 0.5, 1, 1.5
and 2%v/v H2O together with experimentally measured
results from 5 IR filters.
These results will depend on IR filter parameters
including bandwidth, asymmetry and other gas filter
and instrumental parameters.

Figure 7 shows a simulated cross sensitivity over an expanded region near the zero crossing point together with experimentally measured values for several different IR filters. It can be seen that agreement is good. Slight deviations are expected due to different filter line widths and symmetry etc. From this work we can reduce the sensitivity to water vapour to values which become limited by the manufacturing tolerance of the IR filters.

 

Figure 8.The spectra of CO, CO2 (top) and a typical IR filter.

For CO analysis in CEM applications the largest interfering component is CO2(typically 15 vol.% in a flue gas) as can be seen in Figure 8. Similar careful selection of filter parameters allows a transducer to detect 0-50vpmCO with an excellent cross sensitivity to CO2 of less than 1 vpm/20%CO2.

Figure 9.The spectra of CO2 (left), CO (top and right),
N2O (centre) and a typical IR filter.

The measurement of nitrous oxide (N2O) without cross interference from CO or CO2 which both overlap the N2O absorption presents additional problems (Figure 9). Optimising the filter parameters for zero CO cross sensitivity do not give the same values as for CO2. A different approach is thus adopted and that is to remove the CO2 characteristic energy from both measure and reference beams by placing a static gas filter in the beam containing a high concentration of CO2. Alternatively, the measure gas filter can be filled with CO2 and the balance of the reference gas filter can be CO2. This is the approach we take and results in a cross sensitivity of <0.5vpmN2O for 500vpmCO2, 10vpmCO or 2%H2O for an instrument measuring 0-50vpmN2O minimum range. For CEMs applications 20%CO2 gives ~ +3.5vpmN2O and 100vpmCO gives ~ -3.5vpmN2O.

The measurement requirement for monitoring low concentrations of methane demonstrates how we can take advantage of the low absorption of CO2 and H2O in this region and so trade off immunity from interfering gases for maximum sensitivity. By filling the reference gas filter to a pressure and optical density equal to or greater than that expected in the sample cell we can increase the line widths and intensities of the absorption lines and so maximise the degree of correlation, or overlap, between the spectra. The filter parameters were optimised for maximum sensitivity whilst maintaining a reasonably low figure for cross sensitivity. This results in a minimum range of 0-50 vpm CH4 and cross sensitivities of +2.5 vpm CH4 / 2%H2O and +1 vpm CH4 / 10%CO2. The cross sensitivity to other hydrocarbons is expectedly larger due to the common CH stretch absorption characteristic at this wavelength. (Eg 0.2% propane gives a reading ~50 vpm CH4).

Figure 10.The spectra of CH4 and a typical IR filter.
Water vapour (1%v/v.m) is shown at the top.

Good mechanical and optical design should also ensure that the common mode rejection ratio is as high as possible for a variety of other influences, including ambient temperature, source output, window obscuration, etc. This is most easily brought about by making the optical paths of both measure and reference beams identical, for example by alternately interrupting a single beam with the gas filters. This is the design adopted by Servomex where a single chopper wheel serves three functions; to produce temporally spaced measure and reference beams, to separate the beams with periods of darkness to provide a zero energy reference and to allow the use of a low cost pyroelectric LiTaO4 detector. Any design which uses spatially separated measure and reference beams will compromise performance, as changes in one path will not affect both beams equally and so cause the output to change.

As most sources of instability and drift are associated with instrumental effects, removing these with a design with high common mode rejection will mean that the major limit on sensitivity is instrument noise. With increased instrument stability, the absorption path length can be reduced and the gain increased until the instrument becomes noise limited. For example a 0.5m single path design can achieve a 0-50 vpm CO measurement range with <1%fsd peak to peak noise.

One source of drift which is not common mode is the stability of the gas inside the gas filters. Any loss of gas due to leakage, permeation, chemical reaction or adsorption will cause a positive zero drift and a reduction in sensitivity. The gas filter design must be able to transmit IR radiation in the region of interest and be physically and chemically stable. Servomex currently use various gas filter designs; one used to contain very reactive gases such as HCl uses a simple one piece IR quartz envelope which has proved extremely stable. Another design for less reactive gases uses a metal to glass sealed construction which is designed to be leak-proof over a wide storage temperature range (-20 to 70C). Earlier attempts to overcome the difficulties in making gas filters by using adhesive window bonding frequently resulted in leakage causing drift and instability, and these are no longer favoured.

Although heterogenous diatomic gases such as CO, NO and HCl exhibit ideal absorption spectra for gas filter correlation, many other IR absorbing gases exhibit a suitable region of adequate fine structure making the GFC method a far more widely applicable technique. Absorption line intensities and widths can vary significantly with ambient conditions for polyatomic gases and so care must be taken to account for these changes (e.g. with NH3, CO2, SO2, CH4, etc).

The high stability and high sensitivity of this technique also conveys a further advantage of a wide measurement, or dynamic, range. This allows for example, without electrical range changing, the monitoring of CO from 0 to 1000vpm with 1% of reading resolution or 0.5vpm, whichever is larger. This is again particularly useful for CEM applications where occasional high excursions from a normally low value can be accurately monitored.

Conclusions

The optimization of the various design parameters of an infrared gas filter correlation photometer allows this technique to be used for many difficult applications involving measurement of low level gases with high levels of interfering background gases. Over the past decade it has often been the technology of choice for gas analysis of the infrared absorbing gases providing the most cost effective performance for many applications.

References

  1. Goody R., ‘Cross-correlation spectrometer’, Opt.Soc.Am., 1968, 58(7), 900
  2. Chaney L. W., McClenny W. A., ‘Unique Ambient Carbon Monoxide Monitor Based on Gas Filter Correlation: Performance and Application’, Environmental Science and Technology, 1977, 11(13), 1186

  3. Hanst P. L., ‘Spectroscopic Methods for Air Pollution Measurement’, in Advances in Environmental Science and Technology, 1971, Vol.2, 91, Ed. Pitts, J. N., Metcalf R. L., John Wiley & Sons (New York)

  4. White J. U., ‘Long Optical Paths of Large Aperture’, Journal of the Optical Society of America, 1942, 32, 285

REF: P. Middleton. Int. J. Vib. Spect., [www.irdg.org/ijvs] 5, 3, 3 (2001)