DE112011102013B4 - Method for ion mobility spectrometry - Google Patents

Abstract

Method of ion mobility spectrometry (IMS), in which - a known gas or gas mixture in the ion mobility spectrometer (1) is ionized and thus forms reaction ions (6), - an analyte or a mixture of several analytes, ie a substance to be analyzed or a corresponding substance mixture, in the injection region of an ion mobility spectrometer (1) is introduced and either also directly ionized or charge transfer from the reaction ions (6) takes place to the atoms or molecules of the analyte (5) to be analyzed, - the Reaktionsionen- and positively or negatively charged analyte mixture (5, 6) traverses the drift region between the injection and the detection region of the ion mobility spectrometer (1), - a time-dependent measurement series of IMS spectra of this reaction ion and positively or negatively charged analyte mixture (5, 6) is recorded in the detection range of the ion mobility spectrometer, - the IMS spectra of the Measuring series are analyzed so that qualitative and / or quantitative statements are made with respect to the analyte or analyte mixture, characterized in that - for each IMS spectral signal of the series of measurements in a preprocessing step, an adaptive smoothing and group delay equalization is performed and differential quotients of first and second order are formed , - from this, for each peak, its position, its height and its half-width are determined as characteristic parameters describing the peak, - the characteristic parameters peak position, peak height and peak half-width are normalized with the actual measurement conditions at the time of acquisition of the spectra, - an identification of the analytes the normalized, the peaks describing characteristic parameters for the currently known spectra sequences is performed by a pattern recognition method.

Description

The invention relates to a method of ion mobility spectrometry (IMS), in particular a corresponding method for the determination of the analytes. Ion mobility spectrometry, due to its high detection sensitivity, enables the detection and identification of environmental and industrial pollutants, chemical warfare agents, explosives and drugs even in the presence of minor outgassing. Ion mobility spectrometers are currently used primarily for warfare detection and in the workplace monitoring industry. Because of the specificities of this technology, which allows air to be used as the carrier gas, the devices are less expensive to produce than, for example, analytical instruments such as mass spectrometers or chromatographic analyzers. In addition, ion mobility spectrometers allow a high degree of miniaturization. To detect traces of explosives or drugs, more than 80,000 ion mobility spectrometers are currently in use worldwide, preferably at airports.

The physical principle of ion mobility spectrometry is based on the different drift velocities of ions in the electric field in air at atmospheric pressure. It is in the 1 also depicted. Via a gas inlet into the ion source, the ambient air or the air-analyte mixture is introduced into an ion source at the injection site in the ion mobility spectrometer and ionized. This can be done, for example, by means of a weak radioactive beta emitter, such as tritium, whose activity is approximately 50 MBq. If the analyte or the analyte mixture is not introduced together with the ambient air, but separately, this is usually done by means of a transport gas. This arrives at a membrane through which the still neutral analyte molecules migrate and thus likewise enter the ion source. During ionization, positive ions are generated from constituents of the air of the types NH ⁺ , NO ⁺ , (H ₂ O) nH ⁺ , which form the positive reaction ion peak (RIP ⁺ ). Negative ions of the type O ₂ ^- and (H ₂ O) ^- m form the negative reaction ion peak (RIP ^- ). These ions of air (RIP ⁺ ) and (RIP ^- ) are constantly available in the ion mobility spectrometer.

If other molecules, such as the analytes (eg amphetamines, pollutants such as organophosphorus compounds or halogenated hydrocarbons, aromatics, mercaptans, etc.) are present in the air or in the air-analyte mixture, charge transfer from the reaction ions to the reaction takes place Pollutant molecules instead. This charge transfer can be simplified as follows: RP ⁺ + M → RP and M ⁺ (positive mode) RP ^- + M → RP and M ^- (negative mode)

By means of electrical impulses on a control grid, the thus ionized air-analyte mixture from the ion source enters an electric field in which the ions are arranged according to their drift velocities. They walk in the drift tube the distance from the injection site to the detection site, ie a detector such. a Faraday plate with shielding grid. During this "migration", the ions are separated in a weak electric field, which happens according to their mobility against the flow direction of an inert gas. The time required by each type of ion for this "hike" is called its drift time. This time is specific to each type of ion, as long as the measurement conditions (such as pressure and temperature) are kept constant. The drift time is determined by various parameters such as the interactions of the ion species with its environment on the way from the injection site to the detection site or its mobility. Thus, the drift rates are dependent on the molecular size, the molecular charge and the molecular shape. The different drift rates lead to different types of ions reaching the detection site, ie the collector electrode, at different times.

The recorded at the detection location, referred to as ion mobility spectrum waveform allows an analyte separation, as ions of the same charge but different mass and / or structure in the signal-time curve generate maxima at different times.

A disadvantage, however, when using the now widely used methods for the determination and evaluation of IMS spectra that occur in the typical for many applications analysis of multicomponent mixtures matrix or coverage effects that can lead to significant misinterpretation.

This problem is troublesome, for example, in the investigation of mixtures, as they frequently occur in environmental analysis and are caused by the sole use of the usual methods of ion mobility spectrometry by the ongoing ionization reactions. The following example, which is in 2 illustrates this problem when examining mixtures such as For example, in the environmental analysis often occur, and is caused by the sole use of ion mobility spectrometry by the ongoing ionization reactions. In the example, three environmental chemicals (benzene, toluene and xylene) are dosed simultaneously into the IMS, only one peak is visible, the ion formation of the other compounds is suppressed due to their physico-chemical properties.

Analogous to the combination of mass and gas chromatography, ion-mobility spectrometers also have the option of gas chromatographic pre-separation of the analyte mixture with a gas chromatographic separator, e.g. by means of multi-capillary columns, whereby the effect mentioned can be avoided. The combination of the ion mobility spectrometry with the chromatographic separation technique allows the representation of the peaks as a function of the drift and the retention time in the form of three-dimensional diagrams: Instead of representing the peaks as a function of the pure drift time by the ion mobility spectrometer, the total retention time is used. This total retention time consists of the retention time within the gas chromatographic separation device and the drift time through the ion mobility spectrometer.

The representations in the form of so-called chromatograms are obtained when the signal magnitude is mapped as a function of drift and retention time or the total retention time composed thereof. In principle, it is possible to evaluate the "pattern" occurring in the chromatogram depending on the position using methods of image processing and to allocate the desired analytes. However, spectrometric methods are characterized by a high volume of generated data that can hardly be handled without computer-aided statistics and that requires effective data preprocessing under real-time conditions.

In particular, however, if there are changes in the measurement conditions, such as pressure and temperature during the acquisition of the IMS spectra, or even to improve the analysis capability, e.g. by temperature-induced separation of gas components and the prevention of condensation effects at higher temperatures is desired, additional corrective measures are required in their analysis in order to normalize the measured spectra and with sufficient certainty to make qualitative and quantitative statements about the analyte or the analyte mixture, and to reduce the risk of misinterpretations. This normalization of the data requires an additional considerable amount of computation. For this there is still no satisfactory solution. The problem is exacerbated when the measurement methodology is extended to the three-dimensional case by comparing the measured IMS chromatograms with corresponding reference chromatograms.

An approach is in DE 10 2007 059 461 A1 described. In this method, the information is obtained by the evaluation of the complete chromatogram, wherein a threshold value method is used to reduce the numerical effort. On this basis, peaks in chromatograms can be described and identified by appropriate parameters. As a rule, the individual spectra belonging to the chromatogram are also subjected to preprocessing in these methods, with which measured data distortions due to noise, environmental and device parameters are to be minimized. Even after preprocessing the single spectra, the chromatogram contains millions of data points, most of which are in noise or redundant, so that only a comparatively small amount of information is needed to characterize the analytes. Another problem is that the complete analysis data has waveforms that strongly depend on environmental conditions and therefore need to be normalized before chromatogram analysis. In this case, for example, the chromatogram consisting of a sequence of 500 individual spectra with 2000 data points in each case is decomposed using a threshold value method such that associated peak sequences of successive individual spectra can be described and identified by ellipse parameters. The reaction ion peak serves as the reference value of the measured value sequences. For the application-relevant case of a completely vanishing reaction ion peak, no solution is explicitly stated here. Another disadvantage is that in the above-mentioned method, an analysis of the measurement results is possible only after the complete measurement data acquisition. This results in a minimum duration until the measurement result of approx. 8 min, in which changed environmental parameters are to be expected.

For the known application areas, such as explosives detection and process monitoring in general industry, the accuracy of the quantitative statements has been sufficient. With increasing sensitivity and the consequent new fields of application, e.g. In addition, in the medical field for respiratory diagnostics and biofilm detection, or in the semiconductor industry, the quality of the quantitative statements has to be decisively improved.

The moisture contained in each measurement sample has a significant influence on the quantitative measurement result. The possibility of using higher working temperatures would therefore also be desirable for this purpose.

Last but not least, the use of ion mobility spectrometry for many of the purposes listed above requires a mobile IMS device and, in particular, a corresponding method which, in spite of mobility, offers qualitatively and quantitatively high quality signal evaluation.

In the WO 2005/106450 A1 A method of ion mobility spectrometry is described to detect ionic species in a sample. For this purpose, representations that were generated by an ion mobility analyzer are graphically evaluated. For this purpose, selected parts or windows of a three-dimensional plot are compared with known patterns from a data table, and thus chemical compositions within an unknown sample are determined.

In the WO 2007/056488 A1 A method for sample analysis is also described based on ion mobility analysis. Therein, a differential mobility spectrometer (DMS) is described having a first ion mobility filter with a first flow path and a second ion mobility filter with a second flow path, wherein the ion streams of both paths are partially conducted as countercurrents in the other path.

In the CA 2 091 279 A1 describes a method for generating a gas chromatogram in which, by means of an ion mobility detector, sequences of ion mobility spectra are generated, which are correspondingly converted to produce a display. In the US 2001/0 030 285 A1 the possibility of determining a wide range of chemical compounds is described, which is to be achieved by the separate construction of the scanning units of the ionization units in an ion mobility spectrometer.

The US Pat. No. 7,078,680 B1 describes inter alia a method for sample analysis using ion mobility spectroscopy. In this case, a signature of the ion mobility spectrum is generated by means of a waveform decomposition and a statistical evaluation, which is compared with signatures of known substances in a neural network.

It is an object of the present invention to develop a method of ion mobility spectrometry, with which IMS spectra recorded under varying measuring conditions can be automatically evaluated in real time in such a way that qualitative and quantitative determination of the analytes is possible with high accuracy.

This object is achieved by a method of ion mobility spectrometry (IMS) with the features of claim 1. The claims 2 to 12 give embodiments of this inventive solution again.

In a method of ion mobility spectrometry (IMS), in which a known gas or gas mixture is ionized in the ion mobility spectrometer and thus forms reaction ions, an analyte or a mixture of several analytes, ie a substance to be analyzed or a corresponding mixture, optionally by means of a carrier gas, in the Injection region of an ion mobility spectrometer is initiated and either also directly ionized or charge transfer from the reaction ions to the atoms or molecules of the analyte to be analyzed takes place, and this reaction ion and positively or negatively charged analyte mixture the drift region between the injection and the detection area of the By traveling through the ion mobility spectrometer, a time-dependent series of measurements of IMS spectra of this reaction ion and positively or negatively charged analyte mixture in the detection range of the ion mobility spectrometer is recorded and the IMS spectra the series of measurements is analyzed so that qualitative and / or quantitative statements regarding the analyte or analyte mixture are given. According to the invention, adaptive smoothing and group delay equalization are carried out for each IMS spectral signal of the measurement series in a preprocessing step, and differential quotients of the first and second order are formed. From this, for each peak as the characteristic parameter describing the peak, the peak position, the peak height and the half width of the peak are determined. These characteristic parameters peak position, peak height and peak half-width are normalized with the measurement conditions of the spectra current measurement conditions and thus an identification of the analytes from the normalized, describing the peak characteristic parameters for the currently known spectra sequences performed by a pattern recognition method. The exclusive description of the Peak in the IMS single spectrum by the mentioned characteristic parameters on the one hand not to information losses in the subsequent pattern recognition or the immediate determination of the analyte or the components of the analyte mixture. On the other hand, this procedure also allows the description of partially hidden peaks. In addition, the required numerical effort for normalization and detection of the drift time under real-time conditions is minimized by this procedure.

Measurement signals with low signal power are usually superimposed by interference signals, which can lead to significant distortions of the characteristic parameters to be determined. Noise reduction techniques use differences in frequency or time domain as well as different statistical properties of both signal components.

Therefore, in an advantageous embodiment, the method is characterized in that for each IMS single spectrum first wavelet-based smoothing and then a dependent of Thresholding result for noise reduction, adaptive digital filtering for smoothing and group delay equalization is performed. Such preprocessing minimizes errors in the calculation of the characteristic parameters peak position and peak height. The limitation of the peak detection and peak description on absolutely relevant measurement information, as well as the possible calculation of the peak half-width from the peak height and the value of the difference ratio of the second order for each particular peak of the IMS spectrum assuming a Gaussian or cosine squared curve of the IMS peak bring about the desired redundancy reduction required for an instantaneous evaluation of the data and immediate determination of the analytes. Due to the process, this also gives estimates for partially hidden peaks, so that a complex baseline correction can usually be dispensed with.

• 1.) u(t) enthält in den Zeitintervallen 0 ≤ t ≤ t₁ und t₂ ≤ t ≤ t_M praktisch keine Messinformationen. Die Zeitpunkte t₁ und t₂ sind in der Regel unbekannt und müssen geschätzt werden. Die näherungsweise Bestimmung von t₁ und t₂ ermöglicht prinzipiell die Ermittlung statistischer Parameter für die Signalverläufe in den unterschiedlichen Auflösungsstufen j (auch Level genannt) der Wavelet-Transformation.
• 2.) Signifikante Signalanteile im Messsignal im Zeitintervall t₁ ≤ t ≤ t₂ sind in der Regel in allen Auflösungsstufen (Level) der Wavelet-Transformation deutlich von den Größen in den Zeitintervallen 0 ≤ t ≤ t₁ und t₂ ≤ t ≤ t_M zu unterscheiden.

Diese Eigenschaft ermöglicht die Steuerung des Anteils des bei der Signalrekonstruktion durch die Inverse diskrete Wavelet-Transformation (IDWT, also den Umkehrprozess der Wavelet-Zerlegung) zu berücksichtigen hochpassgefilterten Signalanteils mittels einer multiplikativen Bewertung mit einen Faktor g(t) im Bereich von null bis eins. Die Festlegungen für Verlauf und Größe von g(t) lassen sich empirisch aus den Schätzwerten für die statistischen Signalparameter der hochpassgefilterten Signalanteile gewinnen und anhand von Simulationsrechnungen optimieren. Investigations show that exactly time-limited wavelet systems with a low number of coefficients are needed to solve the problem. The Daubechies-4-Wavelet produced the best results in the extensive simulation calculations carried out. Daubechies wavelets are exactly time-limited. Although the spectrum is not band-limited, it falls off quickly to low and high frequencies, so that in addition to the good time selectivity and a good effective band limitation is realized. The shortest is the Daubechies 4 wavelet, it is relatively edgy. As the number of coefficients increases, the wavelets are rounded, increasing the frequency selectivity. Daubechies wavelets are among the most commonly used wavelet transforms for digital signal analysis and compression. Noteworthy is the ease of implementation as fast wavelet transformation (FWT). Based on the measured signals thus determined, a noise reduction in the IMS spectra is achieved by means of wavelet thresholding. In principle, the following assumptions are made for the signal - which can then also be realized by a corresponding parameter selection:

• 1.) u (t) contains practically no measurement information in the time intervals 0 ≦ t ≦ t ₁ and t ₂ ≦ t ≦ t _M. The times t ₁ and t ₂ are usually unknown and must be estimated. The approximate determination of t ₁ and t _{2 in} principle enables the determination of statistical parameters for the signal profiles in the different resolution levels j (also called level) of the wavelet transformation.
• 2.) Significant signal components in the measurement signal in the time interval t ₁ ≤ t ≤ t ₂ are generally clear in all resolution levels (levels) of the wavelet transformation of the quantities in the time intervals 0 ≤ t ≤ t ₁ and t ₂ ≤ t ≤ t _M to distinguish.

This property makes it possible to control the proportion of the high-pass filtered signal component to be taken into account during signal reconstruction by the inverse discrete wavelet transformation (IDWT) by means of a multiplicative evaluation with a factor g (t) in the range of zero to one , The definitions for the course and magnitude of g (t) can be obtained empirically from the estimates for the statistical signal parameters of the high-pass filtered signal components and optimized on the basis of simulation calculations.

In order to eliminate the increase in the "edginess" of the noise-reduced IMS measurement signal associated with the wavelet thresholding, a wavelet-based controlled low-pass filtering was also developed and realized by means of time-discrete convolution.

The group delay equalization is realized as follows:
The required digitization of analog source signals requires frequency-band-limited functional characteristics before analog-to-digital conversion. Scanned functions have a frequency spectrum which is periodically continued at the sampling frequency f _a . The unambiguous description of an analog signal by its samples assumes that the overlapping (aliasing) of the periodically continued spectra is reliably avoided. It is expedient to limit the signal in terms of frequency so that as far as possible with minimum bandwidth no measurement information is lost. It is possible to eliminate with the analog filter all spectral components that are above the significant for the measurement signal frequency limit. Thus, an additional noise reduction is achieved, also further not explained here disturbances are eliminated. The frequency-band limited input measurement signal by means of an analog filter is thus available as a discrete-time signal after an analog-to-digital conversion.

The evaluated with a Butterworth analog filter, so a filter for signal processing, which is designed so that the achieved frequency response in the transmission range is as flat as possible, magnitude spectrum largely corresponds to the requirements for a distortion-free transmission of a measurement signal of high bandwidth, while the phase response or the group delay behavior leads to significant linear signal distortions when the passband of the analog filter is matched to the significant part of the signal spectrum. A realizable within the following digital signal processing phase or group delay correction allows the minimization of the linear signal distortion. Since the signal magnitude spectrum can be considered optimal when using a Butterworth filter, an all-pass is required for the digital phase correction. Alpasses, ie electrical filters which, in the ideal case, have a constant absolute frequency response for all frequencies, while the phase shift depends on the frequency, are distinguished by a frequency-independent magnitude profile and a variation range depending on the filter level for the phase.

It is now particularly favorable to detect the position of the reaction ion peak in the process according to the invention as a function of pressure and temperature and to replace this reaction ion peak in the case of its complete disappearance over a limited period of time by a current temperature- and pressure-corrected moving average. In contrast to previously existing methods, this makes it possible to generate a reference value even in the case of temporary absence of the reaction ion peak, which corresponds to the respective temperature and pressure situation in the system, and to which reference can be made to the analytes' peaks measured at that time and thus their Peaklage can be determined exactly.

In particular, it is advantageous that the determined peak positions of the analytes are specified in the form of a difference distance to the reaction peak and this difference distance is likewise subjected to a pressure and temperature correction. Thus, the position of the peaks of the analytes can be specified independently of the measurement conditions: The same amounts are always obtained for the peak concentrations of the analytes for the same system to be analyzed, even if the temperature and pressure conditions differ greatly for different measurements. This makes measurements comparable to different temperature and pressure conditions comparable. In addition, temperature and pressure-independent reference values can also be generated and used again for subsequent measurements.

ist in der Regel fehlerbehaftet, da p und T einerseits nicht unmittelbar im Driftraum und anderseits nur mit einem entsprechendem Totzeitfehler bestimmt werden können. Deshalb ist es erforderlich, die aktuelle RIP-Position bei jeder Messung direkt dem IMS-Spektrum zu entnehmen. Im einfachsten Fall entspricht die Peakposition des Reaktionsionenpeaks der Peaklage des Maximums im IMS-Spektrum. In Abhängigkeit von der Probenkonzentration verringert sich jedoch die Anzahl der Reaktionsionen durch Ladungsübertragung, sodass die RIP-Position nicht mehr mit dem Detektorstrommaximum korreliert. Die Auswertung von Messreihen zeigt dabei, dass mehrere aufeinanderfolgende IMS-Spektren entstehen können, die an der zu erwartenden RIP-Position kein relatives Maximum besitzen. Diese Eigenschaft hat physikalische Ursachen und muss daher bei der automatischen RIP-Positionserkennung berücksichtigt werden. Ist die RIP-Position in einer Reihe der ermittelten IMS-Spektren nicht auffindbar, so muss sie durch eine adaptive Schätzung bestimmt werden. Dazu ist es zweckmäßig, folgende Peaktypen im IMS-Spektrum zu unterscheiden:

Peaktyp 0:

Maximalwert im IMS-Spektrum, U_pmax

Peaktyp 1:

relative Maxima im IMS-Spektrum mit U_p < U_pmax

Peaktyp 2:

zeitlich verdeckte Peaks auf einer fallenden oder ansteigenden Impulsflanke

The position or time position of the reaction ion peak (RIP) depends on a number of device parameters as well as on the measurement and ambient conditions. These influencing factors influence the position of the wanted peaks in the IMS spectrum to the same or at least similar extent. In order to increase the selectivity of the measurement method, the time intervals of the peaks in the IMS spectrum to the reaction ion peak (RIP) are virtually invariably determined in order to exclude a considerable proportion of errors of unknown size. Under constant measuring conditions (such as pressure and temperature) and with stable device parameters (such as the field strength E), the temporal position of the RIP position is practically constant and fluctuations are only in the order of magnitude of the temporal quantization error of ± 10μs. Pressure and temperature changes and the polarity change, however, lead to a non-negligible change in position of the RIP positions. The correction of the RIP position according to the derivable relationship

is usually faulty, since p and T can not be determined directly in the drift space on the one hand and only with a corresponding dead time error on the other hand. Therefore, it is necessary to take the current RIP position for each measurement directly from the IMS spectrum. In the simplest case, the peak position of the reaction ion peak corresponds to the peak peak deposition in the IMS spectrum. However, depending on the sample concentration, the number of reaction ions decreases due to charge transfer, so that the RIP position no longer correlates with the detector current maximum. The evaluation of measurement series shows that several consecutive IMS spectra can arise, which have no relative maximum at the expected RIP position. This property has physical causes and must therefore be taken into account in automatic RIP position detection. Is the RIP position in a row of detected IMS spectra can not be found, it must be determined by an adaptive estimation. For this it is useful to distinguish the following peak types in the IMS spectrum:

Peak type 0:

Maximum value in the IMS spectrum, U _pmax

Peak type 1:

relative maxima in the IMS spectrum with U _p <U _pmax

Peak type 2:

temporally hidden peaks on a falling or rising pulse edge

Peaks of types 0 and 1 are characterized by the zero crossing of the first order difference quotient u '(n) as well as negative values of the second order difference quotient u "(n). u "(n) <0 (2) │u '(n) │ _min = 0 (3)

For the peak type 2 applies u "(n) <0 (4) │u '(n) │ _min > 0 (5)

The RIP positions of the reference spectra in the positive (p) and negative (n) measurement modes as well as the corresponding pressure and temperature data are needed to calculate the RIP positions during the measurements. To determine the parameters of the reference spectrum, a sequence of approximately eight to ten spectra per measurement mode is required in order to reduce the errors in the temperature and pressure measurement by averaging. The respective last spectrum of the positive measurement mode or the negative measurement mode sequence is declared as a positive reference spectrum or negative reference spectrum. With the exception of the specific quantitative information, the algorithms are independent of the measuring mode (positive or negative).

• (1) Eingabe des Startwertes RIP_0(n,p) und der Grenzen RIP_h(n,p) und RIP_l(n,p) des Toleranzbereiches
• (2) Berechnung der gleitenden Mittelwerte von Druck p und Temperatur T
  
• (3) Druck- und temperaturabhängige Korrektur der unter (1) genannten Werte: RIP_korr(0,h,l) = RIP_(0,n,l)(1 + (p(i) – p)/p)(1 – (T(i) – T)/T) (8)
• (4) Begrenzung des Toleranz- bzw. Suchbereiches für den Erwartungswert RIP_ref.
• (5) Berechnung des Schätzwertes RIP_ref.

The following process steps are required:

• (1) Input of the starting value RIP _{0 (n, p)} and the limits RIP _{h (n, p)} and RIP _{l (n, p) of} the tolerance range
• (2) Calculation of moving averages of pressure p and temperature T
  
• (3) Pressure- and temperature-dependent correction of the values mentioned under (1): RIP _{corr (0, h, 1)} = RIP _{(0, n, 1)} (1 + (p (i) - p ) / p ) (1 - (T (i) - T ) / T ) (8th)
• (4) Limitation of the tolerance or search range for the expected value RIP _ref .
• (5) calculation of the estimate RIP _ref.

The calculation of the RIP position during the measurements of the analyte spectra is carried out according to a similar algorithm, but modifications must be carried out in such a way that in each case the RIP position of the reference spectrum must still be referenced and thus a double correction must be carried out.

A decisive step in the reduction of the data volume in the analysis procedure and thus for the particularly fast and reliable determination of the analytes consists in limiting the pattern recognition in the individual IMS spectra to the characteristic parameters peak position, peak height and peak half width, as well as to a selectable number of search areas within the Drift time, which can be the same or different interval width. These search ranges are also referred to below as drift time intervals. Depending on the analyte to be determined, there are always areas in the IMS spectra that are less relevant for the corresponding evaluation. In an advantageous embodiment of the method according to the invention, therefore, the total analyzable range is limited only to the part which is essential for determining the analyte. This makes it possible to evaluate the spectra on the essential peaks and Here, as described, to limit their characteristic parameters peak position, peak height and peak half-width. By eliminating the non-relevant analyte peaks, the analysis of which would also require effort, but which do not provide a decisive contribution to the desired analyte determination, so the computational effort, which is necessary for unambiguous determination of the analyte, further reduced and increases the efficiency of pattern recognition.

Multivariate methods, such as clustering and classification, methods of dimensionality reduction and pattern recognition, in conjunction with effective preprocessing of data for extracting specific properties of the individual spectra, make statistically reliable statements regarding the comparison of several series of measurements with similar multicomponent mixtures.

In a particular variant of the method according to the invention, the respective determination of the total retention time of an analyte under real-time conditions, i. during the current measurements, realized by a temporally successive evaluation of all search areas, ie the selected drift time intervals of each IMS single spectrum by means of a cluster based on the respective search area cluster analysis, where measured at a suitable location temperature and pressure conditions to normalize the retention time information are used, and this method is repeatedly applied to a sequence of IMS spectrums detected at defined time intervals. Here, the total retention time of an analyte is understood to mean the time required for the analyte to migrate from a defined starting location, the injection site located either within the ion mobility spectrometer or in a device upstream of the ion mobility spectrometer, to the detection site, and the Retention time in the ion mobility spectrometer upstream device and the drift time in the ion mobility spectrometer includes.

Sufficiently accurate pattern recognition by means of cluster analysis requires, for reasons of cost, restrictions on a small number of relevant features. This requirement is met by the proposed approach by dividing the drift time interval into a correspondingly selectable number of regions, possibly with different widths. This method concept makes it possible, in principle, to select reference data from running substance measurement series under real-time conditions. Ambient or measurement conditions are automatically taken into account and, if necessary, allow subsequent normalization for entries in a substance database. The principle can also be used for IMS measuring tasks in which gas chromatographic pre-separation can be dispensed with.

In a particular variant of the method according to the invention, the cluster analysis is applied to a sequence of IMS spectra at equidistant time intervals. Again, such periodic analysis simplifies the analysis in many cases, since it guarantees that the data will ideally be analyzed on a regular basis in which the method is capable of providing complete analysis data, and therefore not analytically important data on a randomly chosen basis get lost over a larger distance.

It is convenient and easy to implement with the methods according to the invention that the results of the cluster analysis with known entries in a substance database for these purposes Drift or retention time are compared and identified with appropriate agreement, the corresponding analytes. In addition, the respective analyte concentration can be estimated from the peak height or from the peak area approximately calculated from the peak height and the peak half-width. Thus, both a qualitative analysis of the analyte (s) in question and a corresponding quantitative analysis of the corresponding analyte are possible by the method according to the invention.

In particular, with the method according to the invention, for the purpose of completing and improving the analysis capability for subsequent analyzes of unknown analytes or analyte mixtures, measurement series with known analytes or analyte mixtures can be carried out, such that in the search ranges of the spectra obtained are typical for the known analytes Parameters are determined and selected and for multiple successively recorded IMS spectra with the same cluster features in one or more search areas and taking into account the real measurement conditions, these sets of characteristic for the respective analyte or analyte mixtures parameters as reference parameter sets for two- or three-dimensional spectra declared and into a Substance database to be entered.

Measurement series with unknown analytes or analyte mixtures can then be carried out with the method according to the invention such that their characteristic parameters in a selected Number of search areas of the same or different interval width, the drift time intervals are determined and selected taking into account the real measurement conditions, for multiple immediately consecutive IMS spectra with area-specific similar cluster features in one or more search areas, these search areas are defined as Referenzsuchbereiche for the analyte or the analyte mixture , - the corresponding characteristic parameters are declared as reference parameters for two- or three-dimensional spectra and are used for recognition and recognition of identical analytes or analyte mixtures and entered into a substance database.

In the case of completely unknown analytes or mixtures of analytes, it may be advantageous to have the search areas consecutively following one another, since it may not be apparent which drift time intervals for the determination of the analytes or of the analyte mixture will make the decisive contribution.

It is provided in a particular variant of the method according to the invention that the analyte or the analyte mixture before entering the ion mobility spectrometer passes a gas chromatographic separation device upstream of the ion mobility spectrometer. The prerequisite for this is that the analyte or the analyte mixture can be evaporated without decomposition. This makes it possible, in particular with analytes or analyte mixtures whose peaks are in direct use of an ion mobility spectrometer without upstream gas chromatographic separation device very close to each other or even partially overlapped to better differentiate the peak times. In the subsequent determination of the peaks, the fact that a gas chromatographic separation device has been connected upstream is of course taken into account, taking at this point the total retention time, which is composed of the retention time within the gas chromatographic separation device and the drift time through the ion mobility spectrometer. The temperature and pressure conditions in the gas chromatographic separation device can be used to normalize the retention time information. If a gas chromatographic separation device is connected upstream of the ion mobility spectrometer, the connection between the gas chromatographic separation device and the ion mobility spectrometer is generally maintained throughout the entire time of analysis of the IMS spectra. In special cases, however, it is also possible to separate or restore the connection between the gas chromatographic separation device and the ion mobility spectrometer according to a predetermined schedule.

The solution according to the invention will now be explained with reference to application examples and corresponding figures.

The 1 shows the structure of a typical ion mobility spectrometer according to the prior art, but how it can be used for the inventive method,

2 demonstrates the problem of assignability of IMS spectra to analyte mixtures, as is usually the case with prior art methods. These two figures, which also serve to illustrate the state of the art, have already been explained above.

3 first shows a typical IMS spectrum of an ion mobility spectrometer,

4 represents an IMS spectrum with a pronounced reaction ion peak and the first and second order difference ratios after the wavelet-based noise reduction and the group delay correction, and the relative extreme values calculated and identified therefrom,

5 shows a sequence of impulse responses of low-pass filters with and without discrete-time all-pass component,

The 6a to 6c show possible typical IMS spectra, where the 6a an IMS spectrum with the maximum peak reaction ion peak, the 6b an IMS spectrum with a shaped and labeled reaction ion peak, but which does not represent the maximum of the IMS spectrum, and 6c represents an IMS spectrum with a labeled reaction ion peak with undetectable peak height.

7 represents an IMS spectrum with a pronounced reaction ion peak and the first and second order difference ratios after the wavelet-based noise reduction and the group delay correction and the relative extreme values calculated therefrom and referred to the RIP position,

8th shows a typical table with the estimates for the peak half-widths as a function of the drift time of in 2 illustrated IMS spectrum,

9 shows a diagram with the drift times of unknown analytes with respect to the position of the reaction ion peak, whereby the characteristic parameters peak position and peak height were determined by measurement for two selected drift time intervals,

10 shows a diagram with the drift times of known, registered in the substance database analytes relative to the position of the reaction ion peak, wherein for two selected drift time intervals the characteristic parameters Peaklage and peak height are shown as reference values with tolerances for the Peaklage,

11 1 shows a diagram with the characteristic parameters Peaklage and peak height, which are represented as a function of the drift time of the reaction ion peak, and the corresponding reference values of an IMS single spectrum at the times T_retent_1 and T_retent_2 for two selected drift time intervals,

12 shows the values calculated separately for both drift time intervals for the quadratic Euclidean distance as a result of the cluster analysis at two different measurement times T_retent_1 and T_retent_2,

13 shows five diagrams with the characteristic parameters Peaklage and peak height, which are determined metrologically as a function of the drag time based on the reaction ion peak, for two selected drift time intervals at different times:

13a : Time T_n for two selected drift time intervals and the characteristic parameter determined at time T_n-1 as reference values,

13b : Time T_n + 1 for two selected drift time intervals and the characteristic parameters determined at time T_n as reference values,

13c : Time T_n + 2 for two selected drift time intervals and the characteristic parameters determined at time T_n + 1 as reference values,

13d : Time T_n + 3 for two selected drift time intervals and the characteristic parameters determined at time T_n + 2 as reference values,

13e : Time T_n + 4 for two selected drift time intervals and the characteristic parameters determined at time T_n + 3 as reference values,

The 14 shows the values for the quadratic Euclidean distance calculated separately for both drift time intervals as a result of the cluster analysis at five different measurement times T_n to T_n + 4,

The 15 represents in a matrix the assignment of detailed consecutive similar patterns per drift time interval in a continuous measuring process of an unnamed analyte mixture,

The table of 16 illustrates in the form of a table the assignment of the column number j in 15 to the drift time interval t_j_min to t_j_max,

and the 17 shows by way of example a relative frequency distribution.

In a known, in 1 shown ion mobility spectrometers described above 1 are produced by coupling with a gas chromatographic device shown here only schematically, such as a multi-capillary, known, consisting of a series of IMS spectra three-dimensional data structures, which are often given in the form of IMS chromatograms also not shown here. In 3 a typical IMS spectrum is shown. The ambient air or the air sample mixture enters an ion source and is detected by means of a weak radioactive beta emitter, such. For example, tritium having an activity of 50 MBq is ionized. It produces positive ions from components of the air of the types NH ⁺ , NO ⁺ , (H ₂ O) nH ⁺ , which form the positive reaction ion peak (RIP ⁺ ). Negative ions from Type O ² and (H ₂ O) ^- m form the negative reactant ion peak (RIP ^-). These air (RIP ⁺ ) and (RIP ^- ) ions are in the ion mobility spectrometer 1 always available.

Each IMS spectrum of the measurement series is smoothed so that the difference ratios of the first and second order can be formed with sufficient accuracy from the signal. 4 shows an IMS spectrum with a pronounced reaction ion peak and the course of the first and second order difference quotients, according to a wavelet-based noise reduction method and an effort-optimized adaptive smoothing with group delay correction and zero-line correction, and the relative extreme values calculated therefrom and identified.

The noise reduction by means of wavelet thresholding leads to an increase in the "angularity" of the noise-reduced IMS measurement signal, which can be realized for reasons of cost expediently by low-pass filtering by means of time-discrete convolution of the noise-reduced IMS signal with an impulse response derivable from the low-pass. Depending on the signal properties derivable from the wavelet thresholding, the selection of a suitable impulse response is adaptively carried out. Expediently, impulse responses are used, in which additional phase frequency responses have been calculated, which effect a group delay correction and thus compensate for the linear distortions of antialiasing filters and other precursors. Surprisingly, the additional group delay correction only leads to an insignificantly higher numerical effort. 5 shows exemplary impulse responses of digital low-pass filters with and without time-discrete all-pass component.

The RIP position can be advantageously determined by an adaptive method for arbitrary amplitudes. It is expedient to distinguish between shaped and partially hidden peaks and to determine the peak position of the maximum value. Within a tolerance range to be determined in each case, which is corrected for temperature and pressure, the RIP position has to be defined, whereby this is done weighted according to the peak height and the degree of peak shaping. In addition, there is a moving averaging of the calculated peak positions, which is defined as the RIP position with temperature and pressure corrected as the amplitude disappears. 6a to 6c show by way of example the method for each different qualities of the reaction ion peak: While in the IMS spectrum of the 6a the reaction ion peak simultaneously represents the maximum value of the spectrum, the reaction ion peak in the IMS spectrum of 6b Although still shaped, but has a very low peak height, and in the IMS spectrum of 6c If the reaction ion peak is no longer detectable, it is necessary to resort to a calculated position of the reaction ion peak, which is marked accordingly in the diagram.

The RIP position forms a suitable reference base for the pressure, temperature and device parameter-related normalized representation of the metrologically recorded characteristic parameters of the analyte peak position and peak amplitude in the IMS spectrum in 7 , To determine the analyte concentration, the estimates of the peak half - value widths calculated from the respective peak height and the value of the difference quotient of the second order are required here in the table in 8th are registered. The estimate for the peak half width is analytic for the Gaussian and cosine square pulses from the relationship t_HW = b · sqrt (u (n) / u "(n)) (9) With 2.22 ≤ b ≤ 2.355, where the two limit values of b in each case result from the estimates based on cosine square pulses or Gaussian pulses.

The individual preprocessing steps lead to the generation of data sets which according to the invention sufficiently describe the respective current IMS spectrum with just a small amount of information, without loss of measurement information and taking into account the measurement conditions and current device parameters, as well as all previously acquired IMS spectra belonging to the data set.

According to the invention, the IMS-typical drift time is divided into a selectable number of drift time intervals of the same or different interval width, and the analyte identification is carried out by pattern recognition immediately after each other separately for each drift time interval. For this purpose, a comparison is made of the characteristic parameters of the current IMS measurement which are relevant for the drift time interval corresponding reference parameters taking into account permissible tolerance ranges. In 9 For example, assuming two drift time intervals 1 and 2, the characteristic parameters of the current measurement at time T_retent_1 are shown.

10 shows by way of example the corresponding reference parameters which are independent of T_retent_1 and identify the permissible tolerance ranges for the peak action.

11 shows the corresponding relationships at a time T_retent_2 compared to the time T_retent_1. In this case, there is no pattern conformity in the drift time interval D1, in contrast to a pattern matching in the drift time interval D2. In 12 Then the calculated values for the relative quadratic Euclidean distance of the cluster analysis carried out separately for the two drift time intervals D1 and D2 are given. By way of example, for the drift time interval D1, a good match between the measured and the reference parameters of the 10 for the retention time T_retent_1 diagnosed, in the drift time interval D2 there is little or no similarity. Also in 12 the values for the relative quadratic Euclidean distance, calculated for the time T_retent_2, of the cluster analysis carried out separately for the two drift time intervals D1 and D2 are given. By way of example, for the drift time interval D2, a good match between the measured and the reference parameters is diagnosed; in the drift time interval D1 there is no or only a slight similarity.

The retention time difference for the analytes to be assigned to the patterns in the drift time domains D1 and D2 is defined by the relationship Delta_T_retent_2,1 = T_retent_2 - T_retent_1 (10) given. Retention times or retention time differences can be corrected or normalized by taking into account the operating temperature in the gas chromatographic separation device.

Furthermore, the results of the cluster analysis, which are based on the characteristic parameters used, can be compared with corresponding entries for the drift and retention time and, if they match, the corresponding analytes can be identified. In addition, the analyte concentration can be estimated from the peak height or peak area, which is approximately determined by product formation from the peak height and peak half-width.

In this case, characteristic parameters of reference spectra can be selected from current measurement series of known analytes or analyte mixtures, which can be assigned multiple consecutive IMS spectra with the same cluster features taking into account the current measurement conditions and therefore represent corresponding reference parameters for two- or three-dimensional spectra. As such, they can also be used for entries in a substance database.

13a to 13e shows diagrams with the characteristic parameters Peaklage and peak height measured for different times T_retent_n = T_n to T-retent_n + 4 = T_n + 4 for two selected drift time intervals D1 and D2, measured as a function of the drift time.

Multiple temporally successive plots of the relative quadratic Euclidean distance to the pattern match in equal drift time intervals in 14 include the statement that these IMS spectra represent reference spectra with respect to the drift time intervals and therefore assign the characteristic as reference parameter to the corresponding measured information parameters.

Furthermore, characteristic parameters of a selectable number of advantageously consecutively consecutive search regions of the same or different interval width, ie the drift time intervals, can be selected from current measurement series of an unknown analyte or analyte mixture taking into account the measurement conditions and in the case of multiply immediately consecutive IMS spectra with domain-specific homogeneous cluster features in one or more several search areas, these search areas are defined as reference search areas for the analyte or the analyte mixture. These corresponding characteristic parameters can then be declared as reference parameters for three-dimensional or in special cases for two-dimensional spectra and used for entries in a substance database for the purpose of recognizing identical analytes or analyte mixtures.

That's how it shows 15 shows in the form of a matrix the assignment of detailed consecutive similar patterns per drift time interval in a continuously running measuring process of an unnamed analyte mixture. The respective IMS spectrum corresponds to the line number i, the column number j characterizes the drift time interval within the IMS spectrum. The 16 illustrates in the form of a table the assignment of the column number j in 15 to the drift time interval t_j_min to t_j_max.

The matrix elements m_ij in 15 indicate the number of immediately consecutive IMS spectra to which identical patterns within a drift time interval j can be assigned as a result of the cluster analysis. By way of example, the spectra i = 3 to i = 20 can be assigned identical patterns with respect to the drift time interval j = 7, which are derived from the characteristic parameters determined for this drift time interval, therefore the value of 18 is assigned here to the matrix elements m_3 7 to m_20 7 by way of example.

The determination of the frequency that can be realized by forming the column sums determines the selection of relevant drift time intervals for the unnamed analyte mixture. The 17 shows by way of example a weighted relative frequency distribution, which assigns a high relevance to the drift time interval j = 7.

The characteristic parameters which are detected in the relevant drift time intervals are then assigned to the analyte mixture as reference parameters and thus enable the recognition of similar analyte mixtures.

LIST OF REFERENCE NUMBERS

1

Ion-mobility spectrometer

2

gas inlet

3

gas outlet

4

Drift gas inlet

5

analyte molecules

6

reagent ions

7

ionized analyte molecules

8th

transport gas

9

Flow direction of the inert gas

10

ion source

11

drift tube

12

detector

13

membrane inlet

14

control grid

15

external auxiliary equipment (valves, pump, flow meter)

16

Gas chromatographic separator

Claims (12)

Method of ion mobility spectrometry (IMS), in which - a known gas or gas mixture in the ion mobility spectrometer ( 1 ) is ionized and thus reaction ions ( 6 ), an analyte or a mixture of several analytes, ie a substance to be analyzed or a corresponding substance mixture, into the injection region of an ion mobility spectrometer ( 1 ) and either is also directly ionized or a charge transfer from the reaction ions ( 6 ) to the atoms or molecules of the analyte to be analyzed ( 5 ) takes place, - the reaction ion and positively or negatively charged analyte mixture ( 5 . 6 ) the drift region between the injection and the detection region of the ion mobility spectrometer ( 1 ), a time-dependent series of measurements of IMS spectra of this reaction ion and positively or negatively charged analyte mixture ( 5 . 6 ) is recorded in the detection range of the ion mobility spectrometer, - the IMS spectra of the measurement series are analyzed so that qualitative and / or quantitative statements regarding the analyte or analyte mixture are given, characterized in that - for each IMS spectral signal of the measurement series in a preprocessing step adaptive smoothing and group delay equalization are performed and first and second order difference quotients are formed, from which for each peak its position, its height and its half width are determined as characteristic parameters describing the peak, the characteristic parameters peak position, peak height and peak half width with the Recording time of the spectra can be normalized to current measurement conditions, - An identification of the analytes from the normalized, describing the peak characteristic parameters for the currently known spectra sequences is performed by a pattern recognition method. A method according to claim 1, characterized in that for each IMS single spectrum first a wavelet-based smoothing and then a dependent of Thresholding result for noise reduction, adaptive digital filtering for smoothing and group delay equalization is performed. A method according to claim 1 or 2, characterized in that the position of the reaction ion peak is detected pressure and temperature-dependent and in the case of complete disappearance of the reaction ion peak over a limited period this is replaced by a current temperature and pressure corrected moving average. Method according to one of claims 1 to 3, characterized in that the determined peak positions are given in the form of a difference distance to the reaction peak and this difference distance is also subjected to a pressure and temperature correction. Method according to one of claims 1 to 4, characterized in that the data volume is reduced in the analysis method in that the pattern recognition in the IMS single spectra on the parameters Peaklage, peak height and peak half-width and a selectable number of search areas the same or different interval width within the Drift time, the so-called Driftzeitintervallen is limited. Method according to one of claims 1 to 5, characterized in that the respective determination of the total retention time of an analyte, ie the time that the corresponding analyte for its migration from a defined starting place, the injection site, either within the ion mobility spectrometer or in a the ion mobility spectrometer upstream device is required to the detection point is realized under real-time conditions that - a temporal successive evaluation of all search ranges of each IMS single spectrum by means of a cluster based on the respective search area cluster analysis, where measured at a suitable location temperature and pressure conditions be used to normalize the retention time information, - this method is applied repeatedly to a detected at defined intervals sequence of IMS spectra. Method according to one of claim 6, characterized in that the time intervals of the detection of the sequence of IMS spectra are equidistant. Method according to claim 6 or 7, characterized in that the results of the cluster analysis are compared with known entries in a substance database for drift or retention time and - if the corresponding analytes are identified, - from the peak height or from the peak height and peak half width Approximately calculated peak area, the respective analyte concentration is estimated. Method according to one of claims 6 to 8, characterized in that series of measurements are carried out with known analytes or mixtures of analytes, - the characteristic parameters are determined and selected in the analytical typical search areas of the obtained spectra, - for multiple consecutive IMS spectra with the same cluster features in one or several search areas and, taking into account the real measurement conditions, these sets of characteristic parameters for the respective analytes or analyte mixtures are declared as reference parameter sets for two- or three-dimensional spectra and entered into a substance database. Method according to one of claims 6 to 8, characterized in that measurement series are carried out with unknown analytes or mixtures of analytes, - whose characteristic parameters in a selected number of search areas of the same or different interval width, the drift time intervals are determined and selected taking into account the real measurement conditions, For multiple immediately consecutive IMS spectra with domain-specific similar cluster features in one or more search areas, these search areas are defined as reference search areas for the analyte or the analyte mixture, - the corresponding characteristic parameters are declared as reference parameters for two- or three-dimensional spectra, - used for recognition and recognition of the same analytes or analyte mixtures and entered into a substance database. A method according to claim 10, characterized in that the selected search areas follow each other without gaps. Method according to one of claims 1 to 11, characterized in that the analyte or the analyte mixture before entering the ion mobility spectrometer ( 1 ) an ion mobility spectrometer ( 1 ) upstream gas chromatographic separation device ( 16 ) happens.

Images