From Olfactory Models to Electronic Nose. New Possibilities of Parallel Analytics.

From olfactory models to an electronic nose. New possibilities of parallel analytics.

FROM OLFACIAL MODELS TO THE “ELECTRONIC NOSE”.
NEW POSSIBILITIES OF PARALLEL ANALYTICS

    Source: magazine «Special Equipment»

Rapid progress in the field of electronic vision and hearing devices poses the task of scientists and engineers to develop the market of technical means similar to the human nose. The traditional approach is to increase the selectivity of highly sensitive sensors. It is in this direction that practically significant results were obtained in the field of enzymatic, immunoenzyme and ligand-receptor analysis. A fundamentally new approach to solving the problem is associated with the use of multiple systems of relatively non-selective sensors.

This approach is based primarily on the advanced development of computer technology, providing real-time processing of multiparameter information. With a multisensory approach, it becomes possible to extract information with a known accuracy on both the composition and concentration of individual components in multicomponent mixtures. The selectivity of individual sensors to the measured components is not of decisive importance; on the contrary, it is important that such sensors are characterized by significant cross-sensitivity. The idea of ​​a multisensory approach has proven to be quite fruitful, which is why a new type of artificial analytical systems has been formed — the «electronic nose». Currently, a number of companies have already made serious applications for the application of the developed analytical tools for assessing the quality of food products, environmental monitoring and medical diagnostics (by smell), as well as very unconventional applications in forensics, security and special military equipment.

The use of modern electronics and technologies to solve various problems related to determining the quality of odor is undoubtedly extremely relevant. The analytical capabilities of modern gas and liquid chromatographs and mass spectrometers make it possible to obtain a variety of information about the qualitative and quantitative composition of odors of products and environmental objects, but such studies are often unreasonably expensive, require a lot of time and, with rare exceptions, are not applicable for research in non-laboratory (“field”) conditions. It is for this reason that the development of simpler, cheaper and, most importantly, faster analyzers, the so-called “electronic noses,” is becoming a priority. The latter are understood as multisensory systems for express assessment of the quality of odors in practical conditions.

Portable sensors for many components of gas mixtures are well known and have found wide application in analytical practice. However, modern technologies and “know how” in the field of “electronic nose” have provided a qualitatively new level of analytical instrumentation. The use of a line of parallel receivers in combination with modern capabilities for processing multiparameter signals (olfactory images — “pattern”) in real time gives researchers a powerful tool, and the main aspirations of the developers, without a doubt, are aimed at conquering the market of devices for non-laboratory research.

SMELL AND PARALLEL ANALYTICS

It should be noted that parallel analytics does not yet exist. More precisely, there is no officially recognized, established term, and its existence occurs as if in a different (parallel) dimension and a different (parallel) quality in relation to classical analytics — analytics of sequential processes and measurements.

Nevertheless, parallel analytics, understood as analytics of simultaneously occurring physical and chemical processes during parallel (in real time) processing of incoming multidimensional information, exists without any doubt. Modern data allow us to assert that it is on the principles of parallel analytics that humans and animals perceive the world of smells. A typical example of the practical application of parallel analytics to modern problems of microanalysis are the “electronic nose” devices.

AT THE CRADLE

Unlike the physical senses (sight, hearing, tactile perception), which were adequately developed and conceptually formalized, the organs of the chemical senses were surrounded by an aura of mystery and were not fully investigated. Due to the inherent internal difficulties in understanding the nature of chemical senses, only sporadic attempts to reproduce them artificially were made for many years. Such attempts primarily include the works of Davis, Drevniks, Moncrieff, Wright, Wilkins and Hartman, Eymour, and others, in which some essential properties of olfactory perception were reproduced using various physicochemical mechanisms and electronic devices. The main obstacle to the practical use of the first olfactory models was their frankly low selectivity and technological imperfection. The situation changed qualitatively only in the late eighties. It was then, in the development of the idea that parallel microprocessor processing of signals from multiple non-specific sensors could lead to the extraction of specific information, that the first encouraging results were obtained in various areas of human activity associated with the key role of smells and taste.

OLFARINE MODELS AND THEORIES

SERIES OF SENSORS

It was in 1980 that researchers at the University of Warwick in Coventry, England, first proposed a line of sensors for detecting odors. Having initially focused on the sensory aspect of the problem, and in particular on the stability of sensor characteristics, the researchers used metal-oxide sensors as primary receivers. In further developments, the results obtained were extended to receivers using conductive polymers. It should be noted that in both technologies, the perception of chemical signals was carried out by changing the conductivity of the sensitive layer of the sensors.

These early developments, despite significant imperfections, gave rise to several commercial projects. In August 1991, the first symposium in this area was held under the leadership of the NATO Research Center. The symposium attracted considerable interest from other research groups around the world, and at present the number of research teams and commercial firms working on problems and technologies in the field of “electronic-nose” instrumentation is steadily increasing.

It was the University of Warwick specialists who came up with the good idea of ​​giving a name to the new direction. Despite the rather primitive analogy with the human nose in terms of technical implementation, the brand name “electronic nose”, or “E-nose”, has taken root and is currently generally accepted. Subsequent works by Gardner and Bartlett (1992), Kress-Roger (1996) and a number of other researchers summed up the second wave in olfactory modeling and the creation of new-generation artificial systems.

WHY AND WHY ELECTRONIC?

First, let us define the properties of gas sensors used in “electronic nose” systems. These sensors, firstly, must be technologically advanced in manufacturing, reliable in operation and ensure monitoring of objects over long periods (hours, days, weeks and even months). The noted properties are currently achievable exclusively on the basis of modern microelectronic technologies.

On the other hand, fundamental to the “electronic nose” is the idea that each sensor in the matrix of primary receivers must be characterized by different partial sensitivities with respect to the space of analyzed odors. Each sensor in the sensor matrix or line has its own characteristic response profile in response to the presentation of the spectrum of tested odors. The resulting response pattern of all sensors is quite complex and can be used to identify and/or describe a given odor in a generally accepted terminology understandable to humans only with the use of modern electronic computing tools.

And finally, the “electronic nose” allows, in principle, to bypass a lot of problems associated with the use of specially trained people in various fields: tasters and perfumers. These problems include: the spread of individual parameters, the adaptation of a living human nose to long-term perception, the influence of fatigue, various infections, toxic substances, physical condition on the acuity of smell; subjectivity in assessing perception and a number of other factors. There is an opinion that the “electronic nose” can create a reproducible image of odors that surpasses in its identification parameters both the capabilities of the human nose and modern analytical equipment, in particular GC/MS.

HOW DOES IT WORK?

As a rule, the “electronic nose” is a complex system consisting of 3 functional units operating in the mode of periodic perception of odorous stimuli: a sampling and sample preparation system, a line or matrix of sensors with specified properties and a processor unit for processing the signals of the sensor matrix. In a typical device, the sample is sucked by an air pump through an inlet pipe into a thermostatted cuvette compartment with a line of sensors installed in it. At the next stage, the sensors are exposed for some time to vapors of volatile substances that make up the odor, while the odorous substances (OS), interacting on the surface and/or penetrating into the volume of the active element of the sensor, form the total response of the system. During the measuring interval, the response of the sensor panel is analyzed and transmitted to the processor module. Then, vapors of a flushing gas (e.g. alcohol) are fed into the system in order to remove the odorous substance from the surface and from the volume of the active part of the sensor material. Finally, the carrier gas is supplied to the sensor cell in order to prepare the device for a new measurement cycle. The period of time during which the sensor is exposed to the PV vapors is called the response time. The second period (with the supply of the flushing gas to the cell) is called the recovery time (latent period).

SENSOR TECHNOLOGIES

As follows from Table 2, according to the operating principle, known sensors can be divided into five main categories based on the measurement of: conductivity, mass gain, characteristics of surface acoustic waves and optical parameters.

Metal oxide sensors

Metal oxide sensors are used much more often to solve various practical problems and, as a result, are much more accessible. The operating principle of such sensors is based on the change in conductivity of a number of wide-bandgap semiconductors based on oxides of tin, zinc, titanium, tungsten, indium and iridium, doped with metals with catalytic properties (palladium, platinum) at elevated temperatures in the presence of analyzed gases.

The sensor consists of a semiconductor material interacting with the PV molecules, located between two metal contacts on top of a resistive heating element, providing an operating temperature of the sensor in the range of 200 — 400 °C. In order to reduce the energy consumption of the device and excessive losses due to heat generation, the sensors are formed in minimum dimensions using microelectronic technologies. On one side of the sapphire substrate there is a thin-film platinum heater, and on the other — sensitive semiconductor elements and electrodes. An insignificant temperature gradient between the heater and sensitive layers allows maintaining a constant operating temperature with high accuracy by stabilizing the heater resistance. The main task solved by doping oxide materials is to obtain the maximum achievable specificity with respect to the target components of gas mixtures. Additional opportunities for increasing selectivity are provided by the correct choice of operating temperature.

MODERN TECHNOLOGIES IN MULTISENSORY ANALYSIS

As a rule, the detection limit of sensors based on oxide materials is within 5 — 500 ppm. A fairly high (interference) sensitivity of sensors of this type to water vapor and a tendency to drift of the baseline are noted. Compensation for such drift, determined by many reasons, is provided by algorithms embedded in the data processing unit. Metal oxide sensors also exhibit a tendency to poisoning (irreversible inhibition) due to volatile sulfur compounds and some other organic compounds. And yet, despite the noted shortcomings, the low cost and commercial availability of this type of sensors have determined its current widest distribution.

Polymer conductive sensors

Conductive organic polymers from the class of polypyrroles, thiophenols, indoles, anilines or furans are also widely used as active materials for conductivity sensors. When such polymers are exposed to PV vapors, various types of bonds (ionic associates, charge transfer complexes, etc.) can be formed, changing the nature of the electron levels. This affects the efficiency of electron transfer along the polymer chain, i.e., in other words, leads to a change in its conductivity. The effect of certain PVs on polymer conductivity is largely determined by the counterion selected for measurements, as well as by the functional groups with which the base polymer material is modified. The use of polymer sensors in “electronic nose” devices, as well as for metal oxide sensors, is based on the widespread use of microelectronic technologies (manufacturing electrode substrates with gaps between individual electrodes of 1 0 — 20 μm, etc.). Micron layers of polypyrrole can be formed from liquid monomer components by the method of electropolarization with voltage cycling from minus 0.7 V to 1.4 V. The required diversity of active materials for creating a line of sensors is achieved both by varying the cycling parameters and by using various (with pre-selected properties) polymer precursors. In accordance with the diffusion nature of the propagation of PV molecules in the sensitive layer, the response time of the polymer sensor is proportional to the thickness of the active zone of the polymer. To reduce it, they go along the path of reducing the zone size to a micron size. Polymer sensors are operational at room temperatures. Therefore, they are easier to adjust and operate as part of portable devices. The detection limit of PV can reach 0.1 ppm, but is usually in the range of 10 — 100 ppm.

The main disadvantages of existing technologies for creating polymer sensors are related to the complexity of the methods for forming sensitive layers, which require time and do not provide high reproducibility of the material properties in a series. However, given the rapid development of polymer physicochemistry in the direction of target “design”, the bottom type of sensors is, without a doubt, extremely promising. It is on this basis that fundamentally new technical modifications of the “electronic nose” can be proposed in the near future for the primary detection and identification of practically important substances and their mixtures (poisonous, potent, narcotic substances, etc.) in non-laboratory conditions.

Sensors based on mass increment measurements

The family of piezoelectric sensors for measuring mass increments, like the family of sensors based on conductivity measurements, is divided into two subtypes:

Quartz crystal microbalances (QCM) and surface acoustic wave (SAW) sensors. A QCM sensor is a quartz resonator disk several millimeters in diameter with metal electrodes on both sides. When excited by alternating current, the crystal is characterized by its own resonant frequency (e.g. 10 MHz or 30 MHz), which is determined, among other things, by its mass. According to the established dependence (Sauerbrey 1959), the change in resonant frequency with an increase in mass is:

l F=-2.3×106 F2 * l m/A

where l F is the frequency shift (Hz), F is the resonant frequency of the piezoelectric crystal (MHz), l m is the increase in crystal mass (g) due to adsorption of PV and A — area of ​​the active zone of the crystal (cm2).

When the sensors are exposed to PV vapors, the latter are adsorbed on the surface of the polymer coating. Subsequent exposure of the crystal to a gas that does not contain odorant molecules returns the resonant frequency to the original level.

Adaptation of the QCM to special technical applications is usually achieved by using a special polymer coating. This task can be significantly simplified by using known selective phases used in gas chromatography. Some proposals are being considered for using specific antibodies as a sorbing phase, for example, to detect explosive vapors. Positive results are known from studies by military specialists on creating devices for detecting trace amounts of toxic and poisonous substances with a detection limit of 1 pg. A characteristic feature of the QCM associated with the linearity of the calibration curve in a wide dynamic range is noted. The response time and recovery time of selective resonance structures are minimized by reducing both the size and mass of the quartz crystal and the thickness of the sorption layer. It should be noted that this is a fairly common property of all devices in which developers actively use microelectronic technologies in the manufacture of sensors. Indeed, when moving to the submicron level of manufacturing elements of measuring devices, the surface/volume ratio increases, and certain instabilities are introduced into the devices, worsening the signal/noise ratio and, ultimately, reducing the accuracy of measurements. This pattern is true for almost all types of devices manufactured with a high degree of microminiaturization.

Surface Acoustic Wave (SAW) Sensors

SAW sensors are the closest “relatives” of sensors manufactured using QCM technology. Surface acoustic waves, as follows from the definition, are waves that propagate along the surface of a device without penetrating into the volume. SAWs operate at significantly higher frequencies than QCM (typical SAWs operate at a frequency of hundreds of MHz). At the same time, SAWs can generate significantly larger changes in the frequency of the recorded signal.

Being planar in nature, SAWs are implemented using modern photolithographic technologies developed in microelectronics. This determines the technological and cost advantages of SAWs compared to other types of sensors, especially in cases where the line of sensors in the “electronic nose” must be sufficiently representative according to the conditions of the practical problem being solved.

Selectivity of sensors manufactured using SAW technologies is imparted (as well as to sensors in the CCM technology) using special polymer coatings. The differential measurement method allows to get rid of systematic errors introduced by changes in humidity, temperature, etc. For example, two closely located SAW sensors, one of which has a special polymer coating, react identically to changes in temperature, which allows to automatically take into account its influence in the difference signal.

A certain disadvantage of the modifications of sensors based on the measurement of mass increment (KKM and SAW) is a more complex circuit implementation compared to conductivity sensors, but this disadvantage is often compensated by a lower detection limit of target odor components. “Aging” of sensitive membranes (active zones of sensors) also represents a certain technical problem, the solution of which is achieved by software methods that ensure timely adjustment of the device.

Sensors based on catalytic field-effect transistors (MOCFT)

The operation of odor-sensitive metal-oxide-silicon field-effect transistors (MOSFTs) is based on chemical reactions of volatile organic compounds in the active zone of the sensor, accelerated by catalytically active metals. The mobile reaction products, diffusing through the MOSFT gate, change the electrical parameters of the transistor, which leads to the appearance of an amplified analytical signal. A typical MOSFT structure includes a p-type semiconductor structure with two n-doped regions and metal contacts. The sensitivity and selectivity of the device are ensured by varying the thickness of the active zone and the type of catalysts, as well as by selecting the operating temperatures at which the system elements operate. One of the disadvantages of MOSFT sensors is directly related to the principle of their operation, according to which the product of the catalytic reaction (e.g., hydrogen) must diffuse through the catalytically active layer to affect the charge-sensitive structure. For this purpose, the sensor design provides for a kind of permeability “window” between the catalytically active layer and the transistor gate. It is quite difficult to meet these requirements technologically, which is why the use of MOCPT sensor devices is currently limited mainly to laboratory studies.

Optical fiber sensors

Optical fiber sensors (OFS) are another modern type of sensors used in “electronic nose” type devices, using glass microfibers coated with a chemically active material on the end or side surface as sensitive elements. The chemically active material is created on the basis of specially selected or synthesized fluorescent dyes immobilized in a polymer matrix. A beam of light, propagating along the optical fiber, produces a kind of interrogation of the chemical coating. When interacting with volatile components of odors, the polarity of the dye environment changes and they respond to the stimulus with corresponding changes in the fluorescence spectrum.

The resolution of the excitation light lines and the fluorescence response of the sensor is provided by either purely spectral or spectral-temporal methods. The advantage of fiber optic sensors is the commercial availability of a very large range of fluorescent dyes previously developed for various scientific and technical applications. This gives developers a wide range of coatings and allows the implementation of various types of FBW devices. The disadvantages of FBW technology include a certain complexity of devices of this type as a whole: the need for a stabilized source of excitation light, a monochromator, a detector, etc., which increases the cost of the device, its energy consumption and weight and size characteristics. It should also be mentioned that a significant number of fluorescent dyes have a limited lifetime, which is associated with their photodestruction.

SIGNAL PROCESSING AND PATTERN RECOGNITION

The main task solved by the “electronic nose” is to identify the sample odor and, if possible, to establish the concentration of the odorant, which is associated with data processing and identification of a multidimensional picture of sensory signals (“odor image”). As a rule, the task is solved in four successive stages: preliminary data processing, extraction of distinctive features, classification and decision making. At the preliminary processing stage, sensor drift is eliminated, sensory data are compressed taking into account transient processes and relative errors are minimized. In this case, traditional signal processing techniques are used, for example, in chromatography: accounting for zero line drift, normalization of sensory responses for a full line of sensors, etc. Extraction of distinctive features pursues two goals: reducing the dimensionality of the measurement space and extracting information necessary for recognizing the olfactory image. For example, if a sensor line contains 32 elements, then the measurement space is characterized by 32 components, which is quite complex both at the time of creating a sufficient database and during further statistical processing of the results.

Since sensors are characterized by cross-sensitivity, in most practical cases such a number of them is obviously excessive. Therefore, it is justified to reduce the dimensionality of the measurement space by identifying the most informative sensory elements. These operations are carried out using the mathematical apparatus of the principal component analysis (PCA) or linear discriminatory analysis (LDA). The PCA method ensures finding the direction of maximum discrimination of sensory response patterns and is most often used in the linear approximation. However, this method is not optimal for solving classification problems. The LDA method is more often used to solve problems related to classification problems. This method allows finding the direction in which the greatest differences between samples with different odors are achieved, while minimizing the differences between samples with the same odors. Since the operation of the “electronic nose” may involve fairly large deviations from linearity, methods capable of processing data without a priori knowledge of the functional dependencies between input signals and output parameters, i.e. nonlinear and nonparametric methods, are required to obtain correct quantitative information. A number of research groups have proposed nonlinear transformations, such as nonlinear Sammon maps and self-organizing Kohonen maps. Sammon maps transform data into a 2- or 3-dimensional space that preserves the distance between each pair of samples in the original n-dimensional sensor space. Kohonen maps transform the n-dimensional sensor space into a 2-coordinate space of processing elements called neurons.

Neighboring neurons are arranged and combined in such a way as to provide self-organizing learning of the system in accordance with some neurobiological principles. At the final stage of classification, after the spaces of neural responses are obtained, the actual identification of the odor is carried out. Classical methods for performing this procedure are: the method of K nearest neurons (KBN), Bayesian classifiers and artificial neural networks (ANN).The last method of signal processing for “electronic nose” devices is, according to many researchers, the most promising. Neural networks are a computer simulation of interacting neurons in the human brain and are an example of nonparametric methods of pattern recognition. Pattern recognition is usually carried out in three stages: extraction of characteristic features from an array of data, classification and identification. A neural network consists of a number of simple information-processing units connected to each other — neurons. Layers of neurons that receive external information are called input, those that output the final result — output, intermediate layers — internal or hidden. In this case, each neuron has several inputs and only one output.

The properties of a neural network as an ensemble are determined not only by the properties of neurons and input values, but also by the mutual arrangement of neurons and the connections between them, i.e. the network topology. It is the topology and the values ​​of the weighting factors that determine the main properties of the network and its “knowledge”. The goal of training a neural network is to minimize errors in output signals, for example, the concentration of components of the analyzed mixture. Training consists of optimizing the values ​​of all weighting, temperature factors, and partial sensitivity factors when working with a data set in samples of known composition. Such a data set must be representative, i.e. cover the entire range of concentrations of all components being determined, since neural networks have poor extrapolation ability. The number of samples required for complete network training depends on the complexity of the problem being solved; it is quite large and is usually determined empirically.

WHAT? WHERE? HOW MUCH?

A large number of “electronic noses” are currently commercially available devices.

Since the pioneering work has been a British priority in the use of new technologies in various fields of application, many of the commercial devices are also of British origin. These include the products of such companies as Aroma-Scan, Bloodhound Sensor and EEV Chemical Sensor Systems. Other European companies include the German Lennartz, the French Alpha M.O.S. and the Swedish Nordic Sensor Technologies. Syrano Sensors, Electronic Sensor Technologies, Hewlett-Packard and Microsensor Systems are currently operating in the United States. Several companies are also known in Japan developing “electronic nose” technologies. The largest number of devices presented in Table 2 cost from $20 to 100 thousand, and a significant decrease in the cost of the products is predicted as the technology for manufacturing the sensors themselves improves. Until now, there are practically no portable models on the market for this product. A technological breakthrough should be expected in this direction.

CLOSER TO PRACTICE

The “electronic nose” has already found numerous applications and helps solve many problems related to food quality assurance, healthcare, environmental monitoring, pharmaceuticals, indoor air quality control, security and military affairs. However, it should be emphasized that in order to successfully solve many of the listed problems, the technologies used in the manufacture of the “electronic nose” must be much more advanced. The “electronic nose” can be effectively used in the food industry to assess the freshness of products, control quality and check the quality of incoming materials, optimize the operation of bioreactors and minimize product variations from batch to batch, monitor accidental or intentional contamination or inconsistencies with the trademark of food industrial products. For example, at North Carolina State University, the “electronic nose” was used to assess the smell of several brands of coffee beans. After preliminary “training” with the use of professional tasters, the prospects for using devices of this type not only for objectifying organoleptic assessments, but also for selecting promising aromas of coffee products were demonstrated. Another equally important application of the “electronic nose” device is the assessment of the freshness of food products, especially since today neither taste nor olfactory tests solve this problem to the required extent. The individual odor of human excrement has long been used in classical medicine as an important diagnostic feature. The “electronic nose” can undoubtedly provide significant assistance in providing objective and clinically and forensically significant assessments of such objects with a characteristic odor as exhaled air, sweat, urine and feces. In clinical studies, express diagnostics of acute infections by the quality of the odor of exhaled air is undoubtedly extremely important. Such a technique can be based on the individual characteristics of the odor of bacterial cultures that are pathogenic for humans. Monitoring human body odor can be extremely important in terms of providing first aid, including at home. In the pharmaceutical industry, the “electronic nose” is designed to provide screening of incoming components for the release of final products, to ensure quality control over the technological process, and to ensure safety requirements for storage of products.

Among the promising forensic applications of the «electronic nose» should apparently be attributed the use of odor information in the investigation of murders and other crimes against the person, and in particular the determination of the gender and individual characteristics of a person's odor by various objects — odor carriers (sweat, blood, hair, excrement, crime weapons and other indirect sources of odor). It is assumed that the «electronic nose» can be used to ensure effective control of employees of institutions with respect to alcohol and drug abuse.

Researchers from the University Medical Center in Durham (USA) used a commercial «electronic nose» with 32 sensors based on conductive polymers to detect sources of unpleasant odors in pharmaceutical products. This development turned out to be devoid of the shortcomings of the human nose, which was unable to effectively screen drugs due to extremely rapid fatigue. Developers of a subsidiary of the DAIMLER-BENZ AEROSPACE concern (Germany) based on the KKM technology created a multi-purpose electronic air analyzer. Its prototype was a device for automatic control of the state of the atmosphere in isolated volumes of manned spacecraft and orbital stations. The scope of the device depends on the type of piezoelectric sensor assembly. The device is capable of determining the condition and quality of food, chemical, perfumery products, identifying toxic compounds in products and waste, and automatically examining the atmosphere in closed volumes. The developers are convinced that a device of this type can be used for military purposes to detect toxic and biological agents, as well as to solve problems of airport security services and customs terminals, in particular, to detect explosives and narcotics during passenger checks, contraband medicines prohibited for transportation. The «electronic nose» can be successfully used to assess the quality of ambient air, control gas and water discharges from industrial and agricultural enterprises, as well as to solve numerous other environmental protection problems. Another application of the «electronic nose» can be early fire alarm systems operating on the principle of detecting volatile combustion products, automated safety control systems for complex and man-made hazardous industries (nuclear industry enterprises, etc.). Separately, it is also worth mentioning the development of systems for detecting planted mines and other munitions, as well as systems for detecting the aging processes of warheads and projectiles, based on the principles of the “electronic nose”.

“ELECTRONIC NOSE” ON THE RISE

Over the past few years, interest in the development of “electronic nose” technologies has been exponential.

It is reasonable to predict that in the very near future we will witness the appearance on the market of a whole family of multisensor devices integrated into portable special-purpose devices. In their main technical parameters, such devices will not be inferior to modern analytics devices. For the most part, these devices will be focused on the analysis of air samples, although ideas related to the creation of multisensor analyzers of various water samples are already being intensively developed. By analogy with the «electronic nose», such devices are called «electronic tongues» and have their main purpose — monitoring of water sources, food products, etc. Developed in a single microstructure (CHIP), such devices, as a rule, contain a sample collection micromodule, a line of sensors made in microelectronic technology and a processor for signal processing and data presentation. Microelectronic technology allows us to switch to the production of industrial products with reproducible parameters within the next 5 years. At the same time, intensive funding will be provided for research aimed at improving multisensor arrays to provide them with better selectivity and sensitivity. Priority is given to the development of miniature sampling and sample preparation devices that provide targeted transfer, purification and concentration of the sample. The ongoing research on integrating microfibrillar concentrating modules with microelectronic technology shows that serious achievements can realistically be expected in this area. Another pressing issue is related to the creation of highly sensitive sensors for special applications. For example, about 10 interdisciplinary working groups are currently working on creating an artificial nose for the instrumental determination of volatile secretions (including pheromones) of animals. The complexity of the problem is that the sensitivity of such sensors must be at least two orders of magnitude greater than the sensitivity of the human nose in relation to the specific odors being analyzed. It should be borne in mind that when approaching the solution of problems related to monitoring explosives, drugs and other contraband goods, the sensor response (and, accordingly, the recovery time) should be 10 — 100 times faster than is achieved at the current technological level. It is extremely important for practical applications of such sensors to achieve high reproducibility of responses (at the level of 5% during the service life), as well as stability of characteristics from device to device.

THIRD GENERATION

The most urgent directions in modernization of the «electronic nose» will undoubtedly be connected with reduction of weight, size and energy characteristics of devices. In addition to miniaturization, the problems of creation of new chemical interactive materials and new technologies improving performance characteristics of sensors in relation to classes of practically important substances will be on the agenda. It can be reasonably assumed that development of such technologies will promote convergence in output parameters of the «electronic nose» with its bioprototype — the olfactory organ of animals. It is quite possible that it is on the path of bionic modeling that a new (third) generation of the «electronic nose» will be created, which will rightfully take its rightful place in analytical instrumentation of the near future. In conclusion, we note that Russian researchers have a significant reserve and original developments in this area, theoretically and morally ready to master segments of this market.