Issue 33
		Print version
Category		Titel		Author
Guest Article		Electronic Noses		Rolf Daniels

The development of electronic noses has been inspired by the physiology behind human olfaction since biological olfactory systems contain many of the desired properties for electronic noses.

1. Physiology of olfaction

Olfaction is the result of a cascade of physiological events that are initiated by the first contact of an odorant with the nasal epithelium and end-up with the conscious perception and fine discrimination of smell.

First of all, odorants have to be carried by inhaled air into the nasal cavity and where they reach the nasal epithelium. This region contains nerve endings of the fifth cranial or trigeminal nerve. About 70 % of odors are said to stimulate the trigeminal nerve although, in general, they may be several times less sensitive than the olfactory receptors of the Regio olfactoria.

In humans, the olfactory epithelium is located in the roof of the nasal cavity, just below and between the eyes (Figure 1). The olfactory region of each of the two nasal passages in humans is a small area of about 2.5 cm². The olfactory epithelium is specially designed in order to allow the olfactory transduction upon the contact with an odorant. The olfactory epithelium is made up of three kinds of cells: supporting cells, basal cells, and receptor cells (Figure 2). The supporting cells produce the mucus which covers the olfactory epithelium. The basal cells are stem cells. They divide regularly producing fresh receptor cells to replace those that die. The receptor cells are ciliated sensory neurons embedded into the layer of supporting cells. The cilia of the sensory neurons are immersed in a layer of mucous. On the opposite side the neuronal cells form axons that send their signal to the brain via glomeruli in the olfactory bulb, mitral cells, and the olfactory nerve tract. Humans possess approximately 50 million primary sensory receptor cells each having 8 - 20 cilia that are whip-like extensions 30 - 200 microns in length. Interestingly, the sensory neurons of dogs which have a more developed sense of smell than humans have about 100 cilia.

Figure 1: Functional anatomy and structure of the human olfactory system

Enlarged version

The odorant receptors are located in the cilial membrane. They contain 7 helical transmembrane regions and are linked to a G-protein. However, before reaching the receptor, an odorant molecule has to dissolve in the mucous layer around the cilia and diffuse to the site of action. Passive diffusion of an odorant through the mocous layer depends on its physical and chemical properties. Alternatively, odorant binding proteins may support facilitated diffusion.

Figure 2: Structure of the olfactory epithelium

Enlarged version

Receptor binding of odor molecules leads to a number of events necessary for signal amplification (Figure 3). This includes G-protein activation and conversion of ATP to the "second messenger" cyclic AMP in the Cytosol. cAMP opens-up ligand-gated sodium channels for the facilitated diffusion of Na+ into the cell. The influx of Na+ reduces the potential across the plasma membrane.

Figure 3: Sensory transduction. Within the cilia of the olfactory sensory cells a cascade of enzymatic activity transduces the binding of an odorant molecule to a receptor into an electrical signal.

Enlarged version

If this depolarisation reaches threshold, it generates an action potential which is conducted along the olfactory nerve to the brain. Signal transduction means that a translation of a chemical into an electrical signa, is completed.

The brain evaluates the odor pattern and is able to associate it to other sensations, e.g. taste and appearance. This information can be stored, retrieved and matched with new input. So far it is not fully understood how the enormous repertoire of odors is detected. In the sixties, J.-E. Moore postulated the theory of the seven primary odors (ethereal, camphoraceous, musky, floral, pepperminty, pungent, putrid) for each having its own type of receptor. However, in the light of current research it is obvious that odor perception requires a combinatorial strategy. Most odor molecules are recognized by more than one receptor and most receptors recognize several odors, probably related by chemical property. Typical features of an odorant molecule are listed in Table 1.

Table 1: Essential physical and chemical properties of an odorant molecule

Sufficiently high vapor pressure

Low polarity

Surface activity

Some ability to dissolve in water and in fat

Molecular weight smaller than 300 (to date, no known odorant possesses a molecular weight greater than 294)

2. Electronic noses

Although the enormous sensitivity of the human nose is an ideal prerequisite for its use as an analytical tool its inherent subjectivity and poor reproducibility have been stimulating factors for the development of several alternatives for detecting volatile chemicals.

Some alternative methods are still using the human sense of smell as the basis for detecting an odor. One example is to measure odor concentration by olfactometry, comprising of an olfactometer and a human panel. The olfactometer dilutes a sample of odorous gas with a neutral gas (defined temperature, humidity, and purity) at a prescribed ratio and presents the mixture to panelists. The dilution at which the odor is first detected is referred to as the threshold dilution and is used to determine the concentration of the odor expressed as odor units.

An other alternative is the combination of gas chromatography and olfactometry (GC-O) in which the human nose plays the role of the detector. This technique is also referred to as sniffing GC. After chromatographic separation the smeller records the odor intensity at the sniffing port of the chromatograph. GC-O exhibits powerful capabilities that can be applied to flavors and perfumes, as well as to any odoriferous products (e.g. pollutants).

However, using the human nose as a sensor to measure odor concentration is labor-intensive, time consuming, presents difficulty if on-site measurements are desired and, requires a large panelist population in order to achieve confidence. Thus, the need for instrumentation to simulate the sensitivity of the human nose, as well as provide consistency in odor measurement remains without question.
Consequently, in England in the early 1980s Dodd and Persaud introduced the array concept of aroma detection. This was the first experimental electronic nose described in literature.

2.1 Definition

The electronic nose has previously been defined as "…an instrument, which comprises an array of chemical sensors with partial specifity and an appropriate pattern-recognition system, capable of recognizing simple or complex odors".

Accordingly, an electronic nose is composed of two main components: the sensing system and the pattern recognition system, capable of recognizing simple or complex odors. And an individual sensor used for the detection of a particular substance, e.g. CO-sensor, is thus no electronic nose.
The sensing system can be an array of several chemical sensors with partial selectivity. Not only electronic chemical sensors but also optical or other transducer signals have been described.

The number of different sensors used in an electronic nose is a compromise between

• quantity and complexity of the data collected and
• broad and over-lapping sensitivity of the array

Commercially available instruments use 6 to 48 sensors.

The terms electronic nose, artificial nose, mechanical nose and odor-sensing system are used as synonyms.

2.2 Comparison between human and electronic noses

Although electronic noses are systems that, just like the human nose, try to characterize different gas mixtures their performance is far from that of our olfactory sense. For example, they are not as sensitive to many odorous compounds as our nose.

Table 2 compares schematically some typical features of human and electronic noses.

A number of technologies is available, where chemical species are used to create an electrically or optically measurable signal. A few of the devices used in electronic noses are listed in Table 3, with a brief description.

Table 3: Examples of gas sensor technologies

Type	Sensitive material	Detection principle
semi-conducting metal oxides (M.O.S., Taguchi)	doped semi-conducting metal oxides (SnO2, GaO)	resistance change
quartz crystal microbalance (QMB); surface acoustic wave (SAW)	organic or inorganic layers (gas chromatography)	frequency change due to mass change
conducting polymers	modified conducting polymers	resistance change
catalytic field-effect sensors (MOSFET)	catalytic metals	workfunction change
pellistor	catalysts	temperature change due to chemical reactions
fluorescence sensors	organic dyes	light intensity changes
electrochemical cells	solid or liquid electrolytes	current or voltage change
infra red sensors	-	IR absorption

In an electronic nose a number of serially arranged singular sensors or an integrated sensor array may be used. Electronic noses using several transducer technologies are known as hybrid systems.

Most commercial odor sensing systems use the following transducer technologies (Figure 4):

• Sensor arrays based on metal oxide sensors (MOS)
• Sensor arrays based on conducting polymers
• Sensor arrays based on mass sensitive piezo-electric sensors
• Mass spectrometers

Figure 4: Photographs of different chemical sensors. (A) metal oxide sensor. (B) surface acoustic wave sensor. (C) quartz micro balance sensor. (D) conducting polymer sensor array.

Enlarged version

2.3.1. Quartz crystal microbalance (QCM)

A quartz crystal microbalance (QCM) sensor comprises a quartz crystal coated with a chemically-selective film. The crystal is configured in an oscillator circuit and typically has a base frequency of 10 MHz. When vapours from the sample interact with the coating film, some are absorbed, causing an increase in the mass of the film. This in turn leads to a decrease in the frequency of the sensor and this is measured as the response. The frequency shift is typically about some hundred Hertz. The advantages of QCMs include high selectivity, high sensitivity, stability over wide temperature ranges, low response to humidity, and good reproducibility.

The coatings are often materials used as stationary phases in gas chromatography. The huge amount of available coating materials allows to specifically adapt the selectivity across the full spectrum of organic compounds.

2.3.2. Surface acoustic wave sensors (SAW)

A surface acoustic wave (SAW) sensor is similar to a QCM sensor in the use of mass change on a piezoelectric crystal to indicate presence and concentration of the odorant molecules. However, SAW sensors operate at a higher frequency of oscillation. A SAW device consists of an input transducer, a chemical adsorbent film, and an output transducer on a piezoelectric quartz substrate. The input transducer launches an acoustic wave which travels through the chemical film and is detected by the output transducer. The device runs at a frequency in the range from 250 MHz to 1 GHz. Vapours from the sample which are absorbed by the coating film cause a reduction in the frequency of the sensor by increasing the mass of the film and changing its visco-elastic properties. SAW sensors offer the same advantages as QCMs but provide higher sensitivity. The disadvantage is the complexity in the interface electronics.

2.3.3. Metal oxide sensors (MOS)

A metal-oxide semiconductor (MOS) sensor is a resistive device made from a metal-oxide film (e.g., tin oxide). The odorant molecules undergo a reduction reaction on the film surface producing a conductivity change in the sensor. To remove the odorant molecules, an oxidation reaction must take place. Heaters within the sensors aid in the oxidation process.

Selectivity can be modified by doping the metal oxide (e.g. with Pd, Pt), modifying the operating temperature of the sensor, or varying the geometry of the sensor.

The advantages of metal oxides include low cost, longevity, low response to humidity, and electronic simplicity. The disadvantages include the necessity to operate at high temperatures, restrictive selectivity, high power requirements, and modest sensitivity (5-500 ppm).

2.3.4. Conducting polymer sensors

Conducting polymer sensors are another type of conductometric sensors. A conductive polymer (CP) sensor is a semi-conducting polymer film coated to adsorb specific species of molecules. When chemical vapours come into contact with the absorbent, the chemicals absorb into the polymers, causing them to swell. The swelling changes the resistance of the electrode, which can be measured and recorded. The amount of swelling corresponds to the concentration of the chemical vapour in contact with the absorbent. The process is reversible, but some hysteresis can occur when exposed to high concentrations.

Appropriate films of conducting polymers can be obtained by electrochemical deposition across narrow electrode gaps. Often the selectivity of the polymer films is altered by using polymer bilayers consisting of polypyrrole and polyaniline or poly (3-methylthiophene).

Selectivity may also be influenced by varying the distance between the electrodes.

The advantages of conductive polymers are wide selectivity, high sensitivity (0.1 - 100 ppm), stability, and operation at ambient temperatures. The biggest disadvantage is a strong sensitivity to humidity. Irreversible binding of odorant molecules to the conducting polymers will cause poisoning of the sensors which can be seen from a drift in the base resistance of the sensor.

2.3.5. Mass selective sensors

Mass selective sensors use the proven technology of mass spectrometers. The principle of the mass spectrometer is well known for detection of chemicals in the vapour phase. Sampled gas mixtures are ionised, and charged molecular fragments are produced. These fragments are sorted in a mass filter according to their mass to charge ratio. The ions are detected as electrical signals with an electron multiplier or a Faraday plate. Mass selective sensors record without previous separation the total ion current over a defined period of time. Most commercial instruments use a quadrupole mass spectrometer. The mass range is typically from 1 to 200 amu.

Mass selective sensors seems in particular to be advantageous when dominant matrices, e.g. water or alcohol, have to be analyzed. In such a case the corresponding peak(s) can be eliminated and thus the significance of the remaining mass peaks is increased. Mass selective sensors are also favorable when quantitative information about specific compounds e.g. pollutants or well defined off-odors are required.

2.4 Sampling

Apart from other technical requirements the reproducibility of the results is markedly influenced by a representative sampling procedure. The concentration of volatiles in a headspace depends on several factors. In order to yield constant partial gas pressure and thus constant results of repetitive measurements all these factors have to maintain constant. The main factors as known from headspace GC are:

• Sample size
• Headspace volume
• Temperature
• Equilibrium time
• Type of carrier gas, its quality and relative humidity
• Pressure of carrier gas

Measurements can be performed by either bringing the sensoric part of an electronic nose in the atmosphere which has to be analyzed or by transferring a representative fraction of the headspace to the sensor array. In this case again two opportunities can be distinguished: the static method uses a fixed sample volume which is brought to the sensor and remains there statically during a measuring cycle. Alternatively, a constant stream of the headspace is drawn across the sensor during a complete measuring cycle.

Sensitivity can be increased by several techniques, e.g. purge and trap technology, which is commonly known from gas chromatography.

3. Data analysis

The raw signal generated by an array of odor sensors is typically a collection of different electrical measure vs. time curves (Figure 5). These signals need to be processed in a more or less sophisticated manner in order to allow the recognition of a particular odor.

Figure 5: Typical sensor response of a conducting polymer sensor array to a certain odorant.

Enlarged version

A basic method for observing a data set is simply to plot all variables, or a subset of variables, in a bar chart (Figure 6). Another form of output is a scaled polar plot (Figure 7). Both can be obtained from the raw signal by integration of the curve over a distinct period of time. This way of visually displaying data is simple to interpret. Each vector on the polar plot represents the output from one sensor. As the relative response of each sensor changes when the sensor array is exposed to vapors from differing samples, the overall shape and appearance of the polar plot vary.

Figure 6: Bar graph generated from the raw data shown in Figure 5. A: Values taken as average signal from 15 to 75 sec. B: Values taken as average signal from 55 to 75 sec.

Enlarged version

Figure 7: Radar plot generated from the raw data shown in Figure 5. A: Values taken as average signal from 15 to 75 sec. B: Values taken as average signal from 55 to 75 sec.

Enlarged version

In order to express similarity or difference of two odors it may be useful to calculate the distance of the two corresponding data sets.

As a chemical sensor system provides several variables, a multivariate distance measure is therefore more appropriate than a simple univariate distance measure. A multivariate distance is calculated in the original or a reduced variable space. There are two main methods to calculate multivariate distances. The euclidic distance is the length of the vector connecting two points in the variable space.

The ED can be calculated according to

ED =

Where xa is the response of sensor number n produced by sample A and xb is the response of the same sensor upon contact with sample B.

However, the euclidic distance does not take the variation within classes into account. A more appropriate distance measure between classes is the statistical distance (also called Mahalanobis distance). The statistical distance is calculated as the ratio between the euclidic distance and the class variance in the direction of vector between class centres. Directions with high variance within the classes will thus give a low statistical distance.

3.1 Classification and dimension reduction

Classification is the task of making a model capable of assigning observations into different classes. A classification is often combined with a dimension reduction in variable space. A multi-sensor system produces data of high dimensionality, i.e. a large number of variables characterizing each observation. It is difficult to visualize more than three dimensions simultaneously. Hence, methods to reduce the dimensionality of multivariate data sets are important. The variable space is an essential concept in order to grasp the ideas behind many data processing techniques. In the variable space, variables are seen as orthogonal basis vectors. An observation corresponds to a point in the sensor space, and a whole data set can be seen as a point swarm in this space. A way to reduce the dimensionality is to find new directions in the variable space and use only the most influential directions as new variables. A basis change is made and a dimensionality reduction is performed. In a principal component analysis, a transformation (projection) in the variable space is made (Figure 8). Directions are found that explain as much of the variance in a data set as possible. These new directions, called principal components, are then used as the new variables. Keeping only principal components with high variation leads to a dimension reduction. There are other methods to reduce the dimensionality in variable space. All these methods operate by finding new directions optimizing a specific criterion, and only the most influential directions are kept for the following visualization and classification.

Figure 8: Schematic picture of how a principal component (PCA) score plot is made. The first principal component is the direction with most of the variance in the data set. The low-dimensional projection of the data can be used as a simple but good approximation of the data set.

Enlarged version

3.3 Artificial neural network

An artificial neural network (ANN) is an information processing paradigm that was inspired by the way biological nervous systems, such as the brain, process information. The key element of this paradigm is the novel structure of the information processing system. It is composed of a large number of highly interconnected processing elements (neurons) working in unison to solve specific problems (Figure 9). ANNs, like people, learn by example. An ANN is configured for an application such identifying chemical vapours through a learning process. Learning in biological systems involves adjustments to the synaptic connections that exist between the neurons. This is true of ANNs as well. For the electronic nose, the ANN learns to identify the various chemicals or odors by example.

Figure 9: Schematic of an artificial neural network. It consists of three interconnected layers of neurons. The computing neurons (hidden and output layers) have a non-linear transfer function. The parameters of the neurons are chosen through a minimization of the output error for a kown training set.

Enlarged version

The basic unit of an artificial neural network is the neuron. Each neuron receives a number of inputs, multiplies the inputs by individual weights, sums the weighted inputs, and passes the sum through a transfer function, which can be, e.g., linear or sigmoid (linear for values close to zero, flattening out for large positive or negative values). An ANN is an interconnected network of neurons. The input layer has one neuron for each of the sensor signals, while the output layer has one neuron for each of the different sample properties that should be predicted. Usually, one hidden layer with a variable number of neurons is placed between the input and output layer. During the ANN training phase, the weights and transfer function parameters in the ANN are adjusted such that the calculated output values for a set of input values are as close as possible to the known true values of the sample properties. The model estimation is more complex than for a linear regression model due to the non-linearity of the model. The model adaptation is made using the so-called back-propagation algorithm involving gradient search methods, where each weight is changed in proportion to the error it is causing.

4. Applications

In order to guarantee a certain standard of raw materials as well as final products, quality control in different industries becomes increasingly important. For many years, electronic noses have been in commercial use for a wide variety of odor and volatile compound applications.

Their most popular applications currently are:

• Quality control in the food industry
• Quality control of packaging material
• Medical diagnostics
• Environmental monitoring
• Perfume and aroma industry
• Control of beverages, e.g. wine and beer
• Tobacco industry
• Coffee industry
• Assessment of car interiors

The following examples shall demonstrate how electronic noses may be used in the cosmetical and pharmaceutical industry.

4.1 Classification of excipients

Figure 10 shows the PCA plot of different excipients frequently used to formulate topical products. The PCA plot shows that polyethylene glycol and propylene glycol can be easily distinguished whereas the less volatile excipients (caprylic/capric trigycerides and octyldodecanol) are found in the same cluster. Obviously the experimental set-up chosen does not allow to discriminate between all of these substances.

Figure 10: Discrimination of different excipients in the principle component analysis

Enlarged version

4.2 Assessment of essential oils

Quality control in aroma and perfume industry is often a time consuming and costly task. Routine analysis can be successfully performed using an electronic nose.

Figure 11 shows the result of an analysis of three different natural peppermint oils grown at different sites and one synthetic peppermint aroma. After dimension reduction, the different classes can be visualized showing a significant distance between the synthetic oil and the samples from natural origin. Although the difference between the three natural oils seems to be rather small, an artificial neural network was easily able to recognize their origin.

Figure 11: PCA score plot of natural and synthetic peppermint oils.

Enlarged version

Figure 12 shows the PCA plot of different batches of a synthetic peppermint aroma. Aroma 1 and 2 are very similar whereas aroma 3 is clearly distinguished. A sensorial analysis revealed a anis off-odor for this aroma.

Figure 12: PCA score plot of different batches of a synthetic peppermint aroma.

Enlarged version

4.3 Quantification of volatile components

The quality control of cosmetic and pharmaceutical preparations is an important task. The content of volatile components can be easily monitored using an electronic nose. Figure 13 shows the score plot of two series of mixtures containing different amounts of the volatile insect repellent DEET (N,N-Diethyl-3-methyl-benzamide). The content of DEET in a matrix of either polyethylene glycol 300 or caprylic/capric triglyceride could be estimated directly from the headspace of the samples. Even samples containing only 1 % of the active ingredient were clearly separated from the pure excipients.

Figure 13: Discrimination of two series of binary mixtures containing the insect repellent DEET. There is a clear separation between the pure components (N,N-Diethyl-3-methyl-benzamide (DEET); Polyethylene Glycole 300 (PEG); Caprylic/Capric Triglyceride (MCT)) and the individual mixtures.

Enlarged version

4.5 Screening of the physical stability of creams

The shelf live of creams is often limited due to physical instability, e.g. creaming, coalescence or changes in colloidal structure. As all these changes alter the distribution of volatile components within the formulation, the physical changes will also affected the "odor" of a cream. Thus, an electronic nose can be used as a tool to monitor the stability of a cream.

Figure 14 shows the PCA plot of a moderately stable o/w cream when stored in a freeze thaw cycle test (-5/40° C) or at constantly 40° C. It can be clearly seen that all samples are subject to changes during two weeks storage. This effect is more pronounced when the samples passed the cycling test. Stable preparations would have shown no difference in the odor profile of sample before and after storage.

Figure 14: Score plot of a moderately stable cream stored for 2 weeks at different conditions. Storing the samples at constantly 40° C has a smaller impact on the physical stability compared to a threeze-thaw cycle.

Enlarged version

5. Conclusions

The electronic nose mimics the behavior of the human nose. It uses many of the concepts from biological olfaction including the sniffing, chemical detection, and odor recognition processes. Although the human nose is often more sensitive to odors than an electronic nose this new technology provides a lot of advantages. In particular the possibility to evaluate odors objectively without getting tired is a great step forward. Thus, this new type of instrumentation should steadily open new fields of application in all industries.

6. Further literature

Amoore, J.E.: Molecular basis of odor; Ed. C.C. Thomas Publisher, Springfield, USA, 1970
Amoore, J.E.: Stereochemical theory of olfaction; Nature 198, 1963, 271-272
Aromascan: Dem Erfolg auf der Spur: Digitale Aromatechnologie; Euro Cosmetics 7-8/1995, 27-32
Boeckh, J.: Die Wahrnehmung des unsichtbaren Riechens und Schmeckens: Moleküle als Sinnesreize; Tagungsbeitrag EBC-Kongreß 1993
Buchbauer, G., Selos, S.: Neue Erkenntnisse über die molekularen Mechanismen des Riechens; Euro Cosmetics, 12/1995, 12-23
Buchbauer, G., Selos, S.: Neue Erkenntnisse über die molekularen Mechanismen des Riechens II; Euro Cosmetics, 1/1996, 22-37
Brunke, E.-J., Ritter, F., Schnaus, G.: Neue Ergebnisse zu sensorisch relevanten Spurenkomponenten in Blütendüften; Dragoco Report 1/1996, 5-21
Davide, F., Holmberg, M., Lundström, I.: Virtual olfactory interfaces: electronic noses and olfactory displays. In: Communications Through Virtual Technology: Identity Community and Technology in the Internet Age (G. Riva and F. Davide(Eds.)), IOS Press: Amsterdam, 2001, 193-220.
Firestein ,S.: How the olfactory system makes sense of scents. Natur 413, 2001, 211-218
Freeman, W.F.: Physiologie und Simulation der Geruchswahrnehmung; Spektrum der Wissenschaft, 4/1991, 60-69
Gardner, J.W., Bartlett, Ph.N.: A brief history of electronic noses; Sensors and actuators B, 18-19, 1994, 211-220
Göpel, W.: Chemical imaging: I. Concepts and visions for electronic and bioelectronic noses; Sensors and Actuators B 52, 1998, 125-142
Göpel, W.: Entwicklungstrends der Chemo- und Biosensorik; GIT 4/1996, 330-335
Hatfield, J.V., Neaves, P., Persaud, K., Travers, P.: Towards an integrated electronic nose using conducting polymer sensors; Sensors and actuators B, 18-19, 1994, 221-228
Hoefer, U., Felske, A., Schulz, G., Steiner, K.: SnO2-Multisensorsysteme für die Analyse von Gas- und Geruchsstoffgemischen; Tagungsbeitrag GMA-Fachtagung Feb. 1996, Düsseldorf
Horner, G.: Einsatz von Sensor-Arrays in der Olfaktometrie. Sensor 93, Vol. 2, Nürnberg 1993, 179-185
Horner, G., Vonach, B.: Ein intelligentes Sensorsystem erkennt Gerüche; Laborpraxis 4/1995, 28-30
Linder, M.E., Gilman, A.G.: G-Proteine; Spektrum der Wissenschaft, 9/1992, 54-62
Madsen, M.G., Grypa, R.D.: Spices, Flavor Systems and the Electronic Nose. Food Technology 54, 2000, 44-46.
Moy, L., Tan, T., Gardner, J.W.: Monitoring the stability of perfume and body odors with an electronic nose; Perfumer & Flavorist 19, July/August 1994, 11-16
Moy, L., Collins, M.: Elektronische Nasen und neuronale Netzwerke; Laborpraxis 5/1996, 14-18
Ohloff, G.: Scent and Fragrances, Springer-Verlag, Berlin Heidelberg, 1994
Pearce, T.C., Gardner, J.W., Friel, S., Bartlett, Ph.B., Blair, N.: Electronic nose for monitoring the flavour of beers; Analyst, 18, April/1993, 371-377
Persaud, K., Dodd, G.: Analysis of discrimination mechanisms in the mammalian olfactory system using a model nose; Nature Vol.299, 23, 1982, 352-355
Persaud, K.C., Pelosi, P.: An approach to an artificial nose; Trans. Am. Soc. Artif. Intern. Organs, Vol.31, 1985, 297-300
Persaud, K.C., Travers, P.J.: Arrays of broad specifity films for sensing volatile chemicals; Department of Instrumentation and Analytical Science, University of Manchester, Institute of Science and Technology
Rapp, M., Reibel, J.: Gasanalytik mit Sensorsystemen: Ein Weg zur elektronischen Nase?; Nachr. Chem. Tech. Lab., 44, 11/1996, 1088-1092
Restrepo, D., Brand, J.B.: Molekulare Mechanismen des Riechens; Dragoco Report 39, 1992, 122-127
Weimar, U., Göpel, W.: Chemical imaging: II. Trends in practical multiparameter sensor systems; Sensors and Actuators B 52, 1998, 143-161.

This article was published as follows: Rolf Daniels, Electronic Noses, Eurocosmetics 10 (2002/9) 20-29.

Author

Prof. Dr. Rolf Daniels



Prof. Dr. Rolf Daniels has a Ph.D. degree in Pharmaceutics. Before continuing his academic career, he worked for Pfizer in the department of pharmaceutical development for 2 years. In 1995 he became Professor of Pharmaceutics in the Institute of Pharmaceutical Technology at the Technical University of Braunschweig. His main interests are in the field of surfactant-free emulsions, stability assessment of semi-solids, and controlled delivery of insect repellents. Since 1997 he has been head of the department Dermocosmetics of the Society of Dermopharmacy (GD).

top