Printer Friendly

An intelligent spectrophotometer for monitoring chemical vapours.

Advances in microprocessor technology achieved during the last decade or so are being coupled to conventional analytical instruments to develop intelligent, compact, and inexpensive chemical monitors for a variety of applications. These developments are being fuelled by a strong market demand for intelligent analytical instruments for industrial hygiene and process control, environmental monitoring, biomedical analyses, remote sensing, and scientific research. According to a 1987 discussion paper on chemical sensors by the Science Council of Canada, "the opportunity exists to conduct research leading to the development of new sensors and new sensing systems. The potential market is inviting. Already, the existing world market for conventional sensors is estimated at $2.2-billion (US) per annum. As cheaper, better, and more capable sensors become available, this market will grow fast."

In general, most analytical techniques (electrochemical, chromatographic, spectroscopic, etc.) are benefiting from microprocessor technology, but the marriage is most effective when operator sophistication needs to be reduced to a minimum and/or the technique is based on a physical principle capable of yielding a wealth of information given the required computational power. Optical spectroscopy is ideal with respect to both of the above criteria. Unlike most analytical techniques that usually depend on one dimension to identify the unknown species (eg., retention time in gas chromatography or voltage in an electrochemical device), optical spectroscopy is multi-dimensional. For example, in absorption and emission spectroscopy molecules exhibit spectra that are a complex function of frequency and may also exhibit polarization effects. In emission spectroscopy, the intensity and frequency of the emitted light depend on the frequency of the exciting light; also, the decay lifetime is a characteristic parameter of the emitting molecule. With the right hardware and software, all these dimensions can be exploited to develop an intelligent instrument capable of analysing complex mixtures with little or no sample preparation.

At the Whiteshell Nuclear Research Establishment of Atomic Energy of Canada Ltd., we are deploying recent advances in ultraviolet spectroscopy and microprocessor technology to develop a portable, versatile and inexpensive monitor capable of identifying and measuring a large number of chemical vapours present at concentrations as low as a few parts per billion in air. Public concerns about the quality of ambient air, due to chemicals advertently or inadvertently released into the atmosphere, as well as several other industrial applications, point to a need for such a monitor. We expect industry, environmental authorities, emergency planning organizations, and others to use this monitor.

The Optical System

The monitor we are developing is a single-beam, ultraviolet spectrophotometer, and is shown schematically in Figure 1 [omitted]. The light source is a xenon flash lamp capable of providing reproducible light pulses at a rate of 0.1 Hz to 10 Hz in the wavelength range from 200 nm to 400 nm. A multiple-pass absorption cell with an effective path length of 5 m is used. A flat-field concave grating disperses the light onto a multi-element photodiode array. The light intensity impinging on the photodiode array is converted by a diode array controller board into a train of pulses whose amplitudes are proportional to the intensity of light striking the respective diode. An analogue to digital AID) converter converts these pulses to digital information that is then processed by the microprocessor.

The Microprocessor

The main functions of the microprocessor are to acquire, analyse and report the data, and to perform diagnostics on system components. In summary, the microprocessor, after a predetermined number of light flashes, calculates the absorption spectrum of the species present in the sample cell compartment. This is done by comparing the intensity ([I.sub.s]([lambda])) of the sample beam at each wavelength to the intensity ([I.sub.R]([lambda])) of the light source and calculating the absorption spectrum A([lambda]) = log [[I.sub.R]([lambda])/[I.sub.S]([lambda]).] The absorption spectrum is then compared to absorption spectra stored in the microprocessor's memory, and a decision is made on which file spectrum or, in the case of more than one component, which combination of file spectra best matches the observed spectrum. Once the absorbing species are identified, the name(s) and concentration(s) of the absorbing species are displayed.

The monitor uses a custom-designed microprocessor. The hardware is based on the Zilog Z280 processor and is designed in CMOS to ensure low power consumption. This microprocessor incorporates on a single chip many of the functions that are often assigned to peripheral circuitry. A memory management unit allows the system to address up to 16 megabytes of memory. An on-chip instruction cache of high-speed random-access memory buffers the most recently used instructions. The use of this cache enhances the execution speed of the numerical analysis routines. The Z280 also includes three 16-bit counter/timers, four 24-bit Direct Memory Access controllers, and a serial port with on-chip baud rate generation. The data acquisition system incorporates two A/D converters. A 12-bit A/D converter is used to read the data from the photodiode array, while a multiplexed 8-bit A/D converter is used to monitor the ambient temperature and the status of the battery pack. Additional peripherals include a power supply for the light source, glue logic for timing and control, and system memory for data and program storage.

Several features are incorporated into the instrument to make it more functional. For example, a power-on self-testing procedure is installed to check for faults in the A/D converters, flash-lamp power supply, and data acquisition circuitry. For user interaction, a full alphanumeric keypad is backlit for use in dim lighting conditions. The instrument is powered by a battery, which can be used for up to eight hours without recharging. (Space is also available for a second battery to extend the operation to 16-hours.) In order to increase battery life, the CPU can turn off power to sections of the instrument when they are not in use. Also, the instrument is equipped with an RS232 port to download data to another computer for archiving and/or further analysis. The data are downloaded in LOTUS 123 file format for direct retrieval into a spreadsheet.

The data acquisition is performed in a two-step process. First, a reference spectrum for clean air is obtained. The reference spectrum is stored for later use in battery supported memory, which enables the instrument to be turned off, transported to another site if necessary, and turned on again without losing the reference information. Second, sample spectra are obtained through the instrument's menu-driven software, which makes the initiation of this process as simple as touching a single key. The instrument turns on an internal sampling fan to draw in the atmosphere to be analysed. After about 10 volume changes of the sample cell, the fan is turned off and data acquisition begins. In both the case of the reference and sample data, multiple measurements of the spectrum are taken and signal averaged. The signal averaging procedure leads to high signal-to-noise ratios, enabling us to measure absorbances down to 0.001.

The reference and sample spectra thus obtained are used to calculate an absorption spectrum, which serves as the fingerprint of the chemical species present. The absorption spectrum is analysed by assuming it is a linear combination of the m library spectra of pure species stored in the memory of the device.

A([lambda])= [C.sub.1].[S.sub.1]([lambda]) + [C.sub.1].[S.sub.2]([lambda]) ....[C.sub.m].[S.sub.m][lambda] (1)

where [C.sub.j] is the concentration and [S.sub.j] is the absorption spectrum of the jth pure species. The absorption spectrum is compared to the library of spectra for pure species stored in memory and a decision is made as to which species are present.

There are many methods of analysing data of the form expressed in Equation (1), including regression methods, linear programming, and pattern recognition. We have chosen to use a multiple regression approach due to its inherent simplicity, the wealth of statistical information it provides, and the rapid execution times. In describing this approach, it is convenient to consider the matrix representation of the problem specified in Equation (1):
 [S] [C] = [A] (2)
 nxm mxl nxl


[graphic omitted]

Equation (2) can be solved by conventional K matrix least-squares techniques. However, this method will often lead to negative elements in the solution vector and/or over-fitting of the data. A negative element in the solution vector would indicate a negative concentration, which has no physical meaning. A more appropriate method of solving Equation (2) is to use an iterative, feed-forward least-squares procedure, with the constraint that the solution vector C cannot be negative, C(j) [is greater than or equal to] 0 for j =l,m. In order to minimize round-off errors, we avoided using explicit inverse methods to solve this least-squares problem. Instead, we used a more robust method involving triangularization of the coefficient matrix. Finally, we chose a feed-forward method, which analyses the data using a minimum number of parameters. This method is less likely to overfit the data and, thus, identify species that are not actually present.

Once data analysis is completed, the results are presented on the liquid crystal display and are also saved in an indexed file in system memory. These files are stamped with the date and time (also an optional header may be added to the file), and thus serve as a logbook of the work done with the instrument.


Figure 2 [omitted] shows an example of the instrument's capabilities. (The actual spectra will not be displayed in the final instrument.) In this case, a file containing the spectra of ozone, benzene, toluene, 2-butanone (MEK), and ortho-, para-, and meta-xylene was first created by metering known concentrations of each compound in a stream of air flowing through the instrument's absorption cell. Each spectrum in the file was identified with the compound's name and absorption coefficient. The instrument was then tested with an air mixture containing benzene, para-xylene, and 2-butanone concentrations of 1.7, 2.4, and 6.6 parts per million by volume (ppmv), respectively. The instrument was successful in identifying the true components of the mixture and measuring their concentrations with a fair degree of accuracy, namely, benzene, para-xylene, and 2-butanone concentrations of 1.9 [+/-] 0.3, 2.3 [+/-] 0.1, and 6.1 [+/-] 0.6 ppmv, respectively, and 0.0 ppmv for all other compounds.

The analysis of a more challenging unknown sample is shown in Figure 3 [omitted]. Using the same file of spectra as in the previous example, the instrument was asked to identify a sample consisting of ortho-, para-, and meta-xylene concentrations of 9.0, 9.1, and 9.2 ppmv, respectively. Based on subtle variations in the spectra of these isomers, the instrument was relatively successful in identifying the presence of the xylene isomers at concentrations of 8.3 [+/-] 0.2, 10.2 [+/-] 0.1, 10.0 [+/-] 0.2 ppmv, respectively, for ortho-, para-, and meta-xylene.


Work to date in our laboratories shows that ultraviolet absorption spectroscopy and microprocessor technology can be deployed successfully to develop an intelligent monitor for identifying and measuring a large number of chemical vapours in air. Needless to say, the success of identifying and measuring the components of an unknown mixture depends on well-known parameters such as the number of components present, the magnitude of their absorption coefficients, their relative as well as absolute concentrations, and how dissimilar their absorption spectra are. The examples presented here are representative of what can be achieved with an instrument designed to be portable.


The authors wish to acknowledge the many helpful discussions on this endeavour with D.R. McLean and members of the Research Chemistry Branch.


H.G. DeYoung, High Technology, 3(11), 41 (1983). J.R. Stetter, P.C. Jurs and S.L. Rose, Anal. Chem. 58(4),860 (1986).

D.M. Mayes and J.B. Callis, Appl. Spectry. 43, 27 (1989).

J.T.Y. Yeh, Chem. Eng. 93(2), 55, (1986).

S. Katz, Can. Chem. News 40(3), 13 (1988).

A.V. Nero, Jr., Sci. Am. 258(5), 42 (1988).

The Guide to Z280 Applications, Zilog applications note, January 1989.

J.B. Gayle and H.D. Bennett, Anal. Chem. 50,2085 (1978).

F.J. Kerrigan, Anal. Chem. 38, 1677 (1966).

T. Blaffert, Anal. Chim. Acta 161, 135 (1984).

D.M Haaland, R.G. Easterling and D.A. Vopicka, Appl. Spectry. 39, 73 (1985).

E.M.L. Beale, Symposium on Linear Programming, Vol. 2 1955.
COPYRIGHT 1990 Chemical Institute of Canada
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1990 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Vikis, A.C.; Harris, P.; MacFarlane, R.; Driver, P.A.; Reynolds, N.P.
Publication:Canadian Chemical News
Date:Jun 1, 1990
Previous Article:Flowsheet simulation in undergraduate chemical engineering education: now there's a controversy.
Next Article:Desktop publishing for spreadsheets.

Related Articles
Mapping chemical changes in brain.
Color measurement: instruments keep on shrinking in size and price.
10 ways to go green in your house.
Tennessee airport to test new explosives detector.
Compounders: now you can monitor color inside the extruder!

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters