Previous Topic Back Forward Next Topic
Print Page Frank Dieterle
 
Ph. D. ThesisPh. D. Thesis 5. Results – Kinetic Measurements5. Results – Kinetic Measurements 5.3. Makrolon – A Polymer for Time-resolved Measurements 5.3. Makrolon – A Polymer for Time-resolved Measurements 5.3.2. Time-resolved Measurements5.3.2. Time-resolved Measurements
Home
News
About Me
Ph. D. Thesis
  Abstract
  Table of Contents
  1. Introduction
  2. Theory – Fundamentals of the Multivariate Data Analysis
  3. Theory – Quantification of the Refrigerants R22 and R134a: Part I
  4. Experiments, Setups and Data Sets
  5. Results – Kinetic Measurements
    5.1. Static Sensor Measurements
    5.2. Time-resolved Sensor Measurements
    5.3. Makrolon – A Polymer for Time-resolved Measurements
      5.3.1. General Properties of Makrolon
      5.3.2. Time-resolved Measurements
      5.3.3. Thickness of the Sensitive Layer
      5.3.4. Influence of the Carrier Gas
    5.4. Conclusions
  6. Results – Multivariate Calibrations
  7. Results – Genetic Algorithm Framework
  8. Results – Growing Neural Network Framework
  9. Results – All Data Sets
  10. Results – Various Aspects of the Frameworks and Measurements
  11. Summary and Outlook
  12. References
  13. Acknowledgements
Publications
Research Tutorials
Downloads and Links
Contact
Search
Site Map
Print this Page Print this Page

5.3.2.   Time-resolved Measurements

The interesting property of Makrolon is the microporous structure, which allows a discrimi­nation of analytes due to different sizes of the analyte molecules [220]-[222]. Small analytes show a very fast sorption (of the Langmuir type) into the pores, medium sized analytes slower kinetics and bigger molecules hardly sorb into the pores of the polymer. For the two analytes R22 and R134a, the sensor responses of the SPR device are plotted in figure 20 during 60 seconds of exposure to analyte with a relative saturation pressure pi/pi0 of 0.1 and afterwards during exposure to synthetic air. The smaller molecules of R22 immediately sorb into the pores of the polymer and desorb very fast resulting in a rectangular shape of the sensor response. The bigger molecules of R134a show a slower kinetics resulting in a curvature of the sensor response. A quantitative correlation of the different interaction kinetics with the absolute volume of the analyte molecules is difficult, as different mechanisms play a role. For the diffusion of the molecules within a microporous polymer, often the kinetic volumes of the molecules are used. These kinetic volumes can be calculated with the help of the Van-der-Waals parameters, the critical pressure, the critical temperature, the Lennard-Jones potential  and the Stockmayer potential [223]-[225]. The kinetic volume is 0.120 nm3 for R22 and 0.140 nm3 for R134a. On the other hand, for the inclusion of the molecules into the pores, the easiest description is the Van-der-Waals volume, which is 0.048 nm3 for R22 and 0.056 nm3 for R134a. For the sorption process into the pores, which can be regarded as a combination of both volumes, the effective volume of the molecules consequently depends on the thickness of the layer as the path length of diffusion increases with the thickness of the layer. This effect is exploited for tailoring the kinetics of a specified analyte by changing the thickness of the sensitive layer (see section 5.3.3).

figure 20:  SPR sensor responses for 60 seconds of exposure to analyte with a relative saturation pressure of 0.1 and after that for exposure to air.

For a quantitative determination of analytes in mixtures, which is based only on the different kinetics of the analytes, the kinetics has to differ for the complete concentration range of interest. In figure 21, the sensor responses of the sorption of two analytes are plotted versus the time of sorption (x-axis) and versus the concentration of the analytes (y-axis). Both analytes show a linear relationship between the sensor response and the concentration. In addition, the relationship between the sensor responses and the time of sorption are also linear for both analytes. The similar shape of the sensor responses along the time axis means that both analytes cannot be discriminated by the kinetics although analyte 2 has twice the sensitivity of analyte 1. For example, the sensor response recorded over time is similar if analyte 1 is present with a concentration of 0.8 or analyte 2 is present with a concentration of 0.4 rendering a simultaneous quantification of both analytes impossible. For a successful discrimination, the sensor responses of the analytes have to show a different shape along the time axis and not a different slope. In mathematical terms, the analytes have to show different partial derivates of the second order (twice derived by time) over the complete concentration range for at least as many time points as analytes to be quantified:

 
(24)

With zi as sensor response of the analyte i and tj as time point j.

 

figure 21:  Linear sensor response of two analytes for a sorption process versus the concentration and the time.

To test the feasibility of the quantification of R22 and R134a in mixtures by the use of only 1 sensor, the sensor responses of the SPR device were time-resolved recorded for 20 single analyte measurements with relative saturation pressures between 0 and 0.1. The sensor responses of these measurements are shown in figure 22. It is obvious that the kinetics of sorption and desorption differs for both analytes over the complete concentration range enabling the quantitative determination of both analytes in binary mixtures on the basis of the kinetics recorded with only 1 single SPR sensor. Thereby the sensor responses at definite time points are used as input variables for a multivariate calibration. These time points, which are for example displayed as dots and boxes in figure 20, have also been referred to as "virtual sensor array" in literature. In contrast to typical setups for chemical sensing, many variables can be generated by one single sensor and not only 1 variable per sensor. The number of variables generated depends on the scanning speed (digitization speed) for the recorded sensor response and on how long the sensor responses are recorded. In contrast to some approaches, which try to fit a model to the sensor response and which use the model parameters as input variables, the scanning of the sensor response allows an accuracy only limited by the sensor recording system, whereas the model approach often has to deal with lacks of fit or instable parameters.

The time-resolved approach does not need to wait until an equilibrium state between the analyte and the analyte sorbed in the polymer has been reached (R22 in contrast to R134a in figure 22) proposed in static approaches [226]. This allows shorter measurements and the application of a broader variety of polymers and analytes.

In figure 22 it is also visible that the sensor responses of many time points are very similar. This means that the variables are highly correlated rendering the application of simple calibration methods like the MLR useless (see section 2.5).

figure 22:  SPR sensor response during 60 seconds of exposure to analyte and afterwards during exposure to synthetic air. The x-axis represents the time and the y-axis represents different relative saturation pressures of the analyte.

The experimental design of multivariate calibrations highly depends on the interactions of the analytes in mixtures with the sensitive layer compared to the interactions of the single analytes with the sensitive layer. A first approach is the assumption that the sensor responses of the analytes in mixtures are additive of the sensor responses of the single analytes. In this case, the experimental design only needs few levels of concentrations for a successful calibration. In figure 23, the sensor responses of the measured binary mixtures of the refrigerants (both of equal concentrations) minus the sensor responses of the single analytes of the corresponding concentrations are plotted versus the time and versus the total concentration of both analytes. It is visible that the plot strongly deviates from the ideal case of additive signals, which would result in a plane. The signal during exposure to analyte is lower than expected. This effect reinforces with higher concentrations of the analytes and can be ascribed to a saturation of the sensitive layer with analyte. Additionally a faster kinetics directly after the beginning and after the end of exposure to analyte is visible for higher concentrations of the analytes. The deviations from the signals being additive make an extensive calibration like a several-level full factorial design of the mixtures necessary.

figure 23:  Deviations from additive signals of both analytes when measured in mixtures of equal concentrations of both analytes.

A systematic variation of the size of the molecules with other molecular properties like polarity being similar can be found when a homologous series of analytes is investigated. In figure 24, the sensor responses of the SPR device are shown for 120 seconds of exposure to analyte of the homologous series of low alcohols and after that for exposure to air. Similar to the refrigerants, the shape of the sensor response changes from a rectangular to a wave with an increasing size of the alcohol molecules. The differences of the sorption and desorption kinetics should allow the quantification of quaternary mixtures using only the time resolved measurements of this single SPR sensor.

figure 24:  SPR sensor response during 120 seconds of exposure to analyte and afterwards during exposure to synthetic air. The x-axis represents the time and the y-axis represents different relative saturation pressures of the analyte.

 

Page 82 © Frank Dieterle, 03.03.2019 Navigation