- Hamish Adam

PRESENTED AT: Annual ISA Analysis Division Symposium, Framingham, MA

Keywords

Ultraviolet, Photometers, Absorption, Process Control

Abstract

Many different analytical techniques have made the transition from the laboratory to online process monitoring. UV photometry has made this transition more easily than most because it possesses some inherent advantages. Simple, robust optics and absence of interference from water vapor and carbon dioxide are two of the main benefits of UV photometry. In addition sensitive, high quantum efficiency detectors and intense light sources are available for this region of the spectrum. This allows high signal-to-noise ratios when measuring the light intensity, which results in low noise absorbance measurements. However, the noise specification for the analyzer is rarely the factor which determines the suitability of the device for a process or emissions monitoring application. For accurate analysis of gas samples, it is often the baseline stability and span drift of the analyzer which determines the absolute accuracy of the analytical result. The noise and drift specifications for an analyzer are largely determined by the design of the . Photometer and the choice of components. A review of different spectrometer designs, as well as the properties of the components used in the photometer, has been performed. The strengths, weaknesses and typical specifications for the different spectrometer designs is presented.

Introduction

The capabilities of an ultraviolet absorption (UV) photometer for use in process or emissions monitoring are determined largely by the choice of components used and the design or physical layout of the photometer. The specifications for an analyzer, such as zero drift, span drift, reproducibility etc. are thus strongly dependent on the photometric design. Increasingly stringent environmental regulations, as well as the demand for more efficient process operation, have placed additional requirements on the capabilities of the photometers used in these applications. To meet these requirements, a number of factors must be considered in selecting the appropriate spectrometer. These factors include wavelength accuracy and precision, span drift, resolution, linearity of response, the number of analytical wavelengths available, noise and baseline stability.

The wavelength accuracy and precision of a photometer determine how well the measuring wavelengths are known, and how reproducibly the measurement can be performed at those wavelengths. These factors determine, to a large extent, the degree to which span drift will occur on the analyzer. Span drift is the change in response· or sensitivity to the analyte species of interest. The resolution of a process photometer is an important factor in determining its linearity of response, and the degree to which interferent species can be measured and compensated for. This is especially true in gas phase applications, where many species exhibit highly structured spectra. In liquid phase applications, the spectral structure is broadened through solvent interaction effects, and few sharp absorption bands appear. The number of analytical wavelengths available determines the number of species which can be measured simultaneously. This is important both for providing analysis of multiple species, if required and also for correcting the concentration of the analyte of interest for potential interferents. Baseline stability, both in terms of zero drift and temperature drift, is often an important factor in determining the accuracy of the measurement. On many photometers, the noise specification far exceeds the drift specifications, and thus it is drift which determines whether the measurement will be sufficiently accurate, while noise determines the precision. The impact of the choice of optical components, such as light sources and detectors, as well as the impact of the spectrometer design, on these parameters, is assessed.

Geometry of Spectometer Designs

In broad terms, there are two common geometries for the layout of the components of an optical absorption spectrometer. These are typically referred to as single beam and dual beam photometers. Each of these geometries has its own merits and demerits.

SINGLE BEAM NON-DISPERSIVE SPECTROMETERS

In a single beam spectrometer (see Fig 1 ), the radiation from the light source is directed through the sample cell, and at a measuring detector. A mechanism for selecting the wavelength( s) of interest is situated either before or after the sample cell. In practice, it is beneficial to place the wavelength selection mechanism before the sample cell, as this minimizes the photon flux through the sample. Exposing the sample to excessive light can change the concentration of the analytes through photolysis-induced chemical reactions.

This is particularly important if there is oxygen present in the sample stream, and the light source emits wavelength in the deep UV (200 nm or less). At these short wavelengths, oxygen absorbs the ultraviolet radiation and decomposes to produce O³P and O^lD [12]. The oxygen atoms can then react with species in the sample gas, for example converting NO to NO₂ or forming ozone by reaction with the remaining molecular oxygen. Since both ozone and NO₂ possess high molar absorptivities in the longer wavelength regions of the spectrum, the formation of these species will bias or interfere with the analysis of others.

Figure 1 – SINGLE BEAM NON-DISPERSIVE SCHEMATIC

Single beam spectrometers appear in a number of different wavelength configurations. These configurations are used to increase the baseline stability of the analyzer in some cases. or to increase the number of species which the analyzer can measure simultaneously. · The single beam, single wavelength analyzer is the simplest design.

In the single beam single wavelength design, one analytical wavelength is used, typically chosen to coincide with the maximum of the absorption band for the species of interest. The analytical wavelength is isolated through the use of a wavelength selective filter. The only signal that is measured is the variation in light flux to the detector as a function of time. Under the assumptions that there has been no drift in either the light source or the detector electronics, and that there are no interfering species present, the variation in light intensity with time may be correlated with the change in concentration of the species of interest. These assumptions are rarely true and thus drift and interference are common problems on single beam, single wavelength analyzers (see Table I).

One of the most common UV spectrometer designs in use today for process and emissions monitoring is the single beam dual-wavelength spectrometer. Single beam spectrometers usually employ two or more wavelengths of light. In a dual-wavelength analyzer, one of the wavelengths is used to measure the analyte species of interest, while the other wavelength is used for drift or interference correction. Single beam dual-wavelength spectrometers are very similar in design to single beam single wavelength spectrometers, except that two (or more) measuring wavelengths are included. This is often done through the use of a chopper wheel, which moves different wavelength-selective filters into the light path as a function of time. In the two filter case, one of the filters (measure) is chosen at the peak of the absorption band of interest, while the other is chosen in a region where the species of interest does not absorb at all (reference). Under the assumption that the light source will drift the same at both wavelengths (ie the source gets brighter or dimmer uniformly across the entire spectrum), normalization of the "measuring" wavelength to the "reference" wavelength will correct the analyzer for drift. If no interference species are present, the variation in normalized signal strength will be proportional to the concentration of the species of interest.

Table 1 – TYPICAL PERFORMANCE AND SPECIFICATIONS FOR A SINGLE BEAM, NON-DISPERSIVE SPECTROMETERS

Some analyzers employ more than two wavelengths, to allow for the simultaneous measurement of more species. The baseline stability of single-beam dual-wavelength devices (Table I) is dramatically better than that of single wavelength devices. This improvement results primarily from the use of the reference wavelength, which helps to account for both drift in the light source, detector drift, and fouling of the windows in the optical path.

SINGLE BEAM DISPERSIVE SPECTROMETERS

Single beam dispersive spectrometers (Fig 2) employ a wavelength dispersive medium, such as a grating or a prism, to separate the radiation from a broadband light source into its component wavelengths. Prisms are rarely used today, due to the advent of low-cost holographic gratings. After the light has been dispersed, the intensity at any number of the component wavelengths is detected, either by scanning a single photodetector across the spectrum or by placing a multi-element detector array at the appropriate position to sense the wavelengths of interest. Scanning monochromators have seen little attention in process control applications, due to ?e lack of wavelength stability and thus the associated span drifts. Diode array-based spectrometers, having no moving parts in the spectrometer, exhibit what is referred to as Connes advantage. This advantage deals with wavelength precision and implies that once the wavelength axis is calibrated, there should not be any drift in this axis assuming that none of the parts of the spectrometer have been moved. This leads to high wavelength precision (in comparison to other dispersive devices) but not necessarily high wavelength accuracy due both to difficulties in providing an accurate wavelength calibration and due to the slight non-linearity of the dispersion system. Furthermore, disassembly and reassembly of the sample cell or the spectrometer components can lead to further errors in the wavelength accuracy.

The resolution of these devices is strongly dependent on both the design of the spectrometer and on the choice of diffraction grating used. Many manufacturers report the pixel dispersion in place of resolution. Unfortunately, this often does not reflect the actual resolving capabilities (bandpass) nor the wavelength accuracy of the device. The actual resolution is determined from the size of the entrance slits, the degree of aberrations in imaging the slit on the photodetectors, and on the precision to which the components are mounted at the appropriate focal points. The Rayleigh criterion is of primary importance to the separation of spectral lines. The images of two spectral lines are said to be resolved when the combined diffraction pattern can be identified as being due to two separate objects rather than one[1].

Figure 2 – PHOTODIODE ARRAY SPECTROMETER SCHEMATIC

The spectral resolving power of a diode array spectrometer is at best three times the pixel dispersion. "The term 6A/pixel has nothing to do with spectral resolution; it is merely the linear dispersion of the diode array spectrometer. The pixel dispersion and the spectral resolution are related to each other by the width of the entrance slit and the imaging properties of the spectrometer." [11] Consider a monochromatic light source, imaged on the entrance slit to the grating spectrometer. This will produce an image at the detector which is, at a very minimum, of the same dimensions as the entrance slit. Any errors in alignment of the grating or positioning of the array detector, an,d any aberrations in the optical system, will result in a larger image. Assume that the pixel dispersion of the array is 1 nanometer {ie. the spacing between adjacent pixels covers one nanometer in the dispersed spectrum). The image of a single monochromatic line, centered between any two given pixels, (see the left-hand diagram in Figure 3), would be observed by two pixels. Thus a single monochromatic line may be imaged as being two nanometers wide, and could not be differentiated from two monochromatic lines imaged at the center of the two pixels. Two monochromatic· lines, centered two pixels apart (the right-hand diagram in Figure 3), would appear as a broad continuum 4 nanometers wide. To resolve two nearby features reliably, they must be at least three pixels apart, even if aberrations, image broadening, and focal point issues are ignored. In practice, the bandwidth or resolution of a diode array spectrometer should be considered to be on the order of 3 to 6 pixels [2]. Most photodiode array spectrometers used for process or emissions monitoring are configured to disperse either 200 to 400 nm radiation or 200 to 800 nm radiation across a 1024 element photodiode array. This results in pixel resolutions of approximately 0.2 to 0.6 nanometers, and thus achieved resolutions are 0.8 to 3 nanometers. For special applications, much smaller wavelength regions may be chosen, with the advantage that the resolution is increased. For example, a diode array spectrometer developed for the measurement of ammonia operates in the 200 to 230 nm region of the spectrum [3], and appears to have a Rayleigh resolution of 0.1 to 0.15 nanometer.

Figure 3 – THE IMAGE OF MONOCHROMATIC LINE ONTO A PDA (left) AND TWO SEPERATED LINE (right)

Photodiode arrays do not possess high photometric linearity ( or photometric accuracy). Typical uniformity of response (pixel to pixel) is on the order of +/- 3 % [4], and the baseline drift specifications[5] of these devices contribute to the photometric inaccuracies. Single beam dispersive instruments are capable of measuring the concentration of a number of species simultaneously since there are a multitude of measurements occurring at a wide range of discrete wavelengths. One may view this as a large number of possible dual-wavelength pairs which may be used in the analysis? In practice, chemometric techniques such as Principal Components Regression or Partial Least Squares· are often used. in the data analysis. These techniques are required to deal with the non-linear behavior of the devices and of the measured absorption due to low spectral resolution.[6]. Additional non-linearities are caused by stray light [2], and may be induced by the intense UV source (typically 25 to 100 watts) used in these devices. Dispersive instruments employing diode array detection systems require that the spectrometer be placed after the sample cell. This requires that the complete photon flux from the source pass through the sample first, increasing the possibility of photochemical reactions.

While the spectrometer itself may have no moving parts, a mechanical chopper is often used in the light path to provide dark current correction. The mechanical chopper is not required if a pulsed flash lamp is used as the light source, since the dark current spectrum can be acquired during periods of time in which the lamp is not being pulsed. However, the noise specifications of pulsed xenon flash lamps indicate that analyzers which employ these devices will suffer from poor signal-to-noise characteristics.

Th? noise, linearity, and drift specifications for a photodiode array-based spectrometer are similar to those of other single beam spectrometers (Table II).

Table 2 – TYPICAL PERFORMANCE AND SPECIFICATIONS FOR A SINGLE BEAM DISPERSIVE (BASED ON REFERENCE 5) AND DUAL DEAM MULTIWAVELENGTH SPECTROMETERS

DUAL BEAM, MULTIWAVELENGTH

A schematic representation of a dual-beam multiwavelength spectrophotometer is given in Figure 4. "Accurate measurement of the absorption coefficient at different wavelengths are best made using double-­beam spectrometer"[9]. A separate reference and measure path are maintained for each of the measurement wavelengths. The light from the lamp is split into two separate beams through the use of a beam splitter. One beam passes through "clean" air and impinges on a reference photomultiplier tube. ·The other beam passes through the sample and impinges on the measure photomultiplier tube. Data is acquired from the two PMT's simultaneously. By taking the ratio of the signal strength from the measure and reference PMT, any drift in the strength of the light source will cancel out. Electronic drift and variations in detector gain are minimized through the use of a reference wavelength in the measurement procedure. In addition, most of the photometric noise in the light source will cancel out due to the common-mode nature of the measurement. This provides exceptional baseline stability and signal-to-noise ratios. Photomultiplier tubes possess excellent photometric accuracy and linearity, typically on the range from 0.1 to 1 % [8]. PMT's are about 100 to I 000 times as sensitive to light as a photodiode in a diode array and possess about I 000 times the surface area. These properties allow the detection of low light levels and thus require less intense light sources in an absorption photometer. This can be important in preventing photolysis of the sample gas.

Figure 4 – A DUAL BEAM MULTIWAVELENGTH SPECTROMETER

The light sources used in these analyzers are hollow cathode lamps, which emit characteristic emiSSion lines determined solely by the metal from which the cathode is constructed. The center wavelengths of the emission lines are known to have high accuracy and precision, and due to the nature of the discharge in these lamps, the center wavelength does not drift due to temperature, pressure, or any other external influence. The emission band is a fundamental property of the material. The bands are also extremely narrow, which contributes to the linearity of the absorbance measurement. For example, the magnesium emission line which is often used for the measurement of SO₂ occurs at precisely 285.214 nanometres and is 0.16 nanometres wide at half maximum. These narrow emission bands at precise wavelengths allow the analyzer to achieve high resolution, wavelength accuracy and wavelength precision in the · spectrometer, · independent of variations in the filter bandpass.

The filter wheel can hold up to six bandpass filters, thus allowing for the measurement of up to 5 species simultaneously while using one filter for baseline stability correction. The dual-beam nature of the spectrometer results in an automatic correction of the signals for variations in the light source, while the use of an additional analytical wavelength provides baseline correction for variations in amplifier or photomultiplier tube gain. The net result of this is that the concentrations of the five measured species are corrected for any variations which may have occurred in the photometer itself. Even variations in the wavelength selective filters are removed by the dual-beam nature of the photometer. The filters typically have a 5 to 10 nanometre full width at half maximum bandpass. The bandpass is centred on the analytical line of interest. Any variation in the bandpass characteristics of the filter will result in a change in the light intensity reaching the beam splitter. However, since the beam is divided into two separate beams at this point, the dual-beam nature of the photometer will eliminate the effect of these variations in the calculation of the absorbance.

The narrow bandwidth of the emission lines contributes directly to the linearity of the measurement. Photodetectors perform a linear average all of the incident light on the detector. If the range of wavelengths incident on the detector encompasses a region in which the species of interest has some variation in absorptivity, this averaging process creates some degree of non-linearity in the measured absorbance. This is because the actual light intensity varies exponentially with the concentration of the gas, but the detector responds linearly to the light flux. A detailed mathematical description of the absorption process explaining the nonlinearity of various types of detection systems is given in reference 7. The use of spectral radiation of minimal bandwidth (such as the emission lines from hollow cathode lamps) results in a very narrow range over which the averaging process occurs, and thus ensures that linearity can be achieved over a wide range of gas concentrations.

Conclusions

Single beam dual-wavelength devices have seen the greatest application in process and emissions monitoring. However, the drift specifications and non-linearities for these devices are often limiting factors when modem applications, which often require high accuracy and low .measurement ranges, are considered.

Single beam multi-wavelength devices, such as those based on diode array detectors, have the advantage of acquiring data simultaneously over a multitude of analytical wavelengths. This has been their strength in laboratory applications, where they have been used for qualitative analysis in the interpretation of unknown samples, and quantitatively in the calculation of the concentration of ?pies. of known composition. In such applications, difficulties with drift and wavelength calibration could easily be overcome by zeroing the analyzer before and after sample analysis to minimize baseline drift and running a calibration sample to minimize the effects of wavelength shifts and span drifts. Baseline drift has been a problem for these analyzers in process and emissions monitoring applications, with many suppliers specifying that a one-hour auto zero is required. In process applications, this implies that the analyzer is unavailable for process monitoring almost ten percent of the time since a typical zero operation will last five minutes. In addition, span drift and cross-interference due to non-linearities have been observed.

Dual-beam analyzers outperform single-beam devices on all of the pertinent specifications related to zero drift, span drift and linearity. Dual-beam multiwavelength photometers, based on narrowband emission sources, are capable of weeks of continuous operation with minimal baseline drift. In addition, the accuracy and narrowness of the spectral emission bands result in exceptional linearity and thus dynamic range. The high photometric accuracy provided by the dual-beam design and the photo-multiplier tube detectors allows the analyzers to measure low concentrations of analyte gas. Their only detriment is that with only six analytical wavelengths available, the photometer is not suitable for the analysis of mixtures of unknown composition. Since the composition of tail and stack gas streams is known, however, the net result of this is a photometer designed for process applications, which possesses the stability and linearity for on and off-ratio analysis of tail gas streams [10], and the cross-talk rejection required to perform detailed analysis of stack gases.

References