Applications of Guided Microwave Spectroscopy and Multivariate by dsp14791

VIEWS: 57 PAGES: 29

									   Applications of Guided
Microwave Spectroscopy and
 Multivariate Techniques for
the Analysis of Solid Samples
             Adrie Dane
  CPACT, Department of Chemistry,
  University of Hull, HU6 7RX, UK
                         Contents
• Microwave Spectroscopy
  – Guided Microwave Spectroscopy
  – Analysis Of Solid Non-uniform samples
• The Determination Of Moisture In Tobacco
• The Analysis Of Fat In Crisps
• Conclusions



      A.D.Dane@chem.hull.ac.uk, CPACT                        2
      Department Of Chemistry, University Of Hull, HU6 7RX
    Microwave Region

                                           The Dielectric Constant
                                          ε’=(vvacuum/vmixture)2




A.D.Dane@chem.hull.ac.uk, CPACT                                  3
Department Of Chemistry, University Of Hull, HU6 7RX
Polarisation In Microwave Region


                          By Orientation
                                                            Ionic




                                                                Electronic




     A.D.Dane@chem.hull.ac.uk, CPACT                                         4
     Department Of Chemistry, University Of Hull, HU6 7RX
Amplitude Reduction
     Energy loss as heat caused by: friction of the
            polarising/relaxing elements.

               ’                      Resonance Frequency




      ” (electric loss factor)


                     Frequency

A.D.Dane@chem.hull.ac.uk, CPACT                              5
Department Of Chemistry, University Of Hull, HU6 7RX
       Water GM Spectrum      4
                           x 10


                     1.6

 Cut-Off
                     1.4
Frequency
            Signal
                     1.2             Amplitude
                      1
                                     Reduction:
                     0.8


                     0.6

                             500   1000        1500      2000    2500   3000
                                               Frequency [MHz]

                                          ’
                                          ”


       A.D.Dane@chem.hull.ac.uk, CPACT                                         6
       Department Of Chemistry, University Of Hull, HU6 7RX
                         GMS


                                                         a=4.8cm


                                              b=10.2cm



                   c= 2a = 9.6 cm
                   fc=vvacuum/2a’
A.D.Dane@chem.hull.ac.uk, CPACT                                    7
Department Of Chemistry, University Of Hull, HU6 7RX
           Water GM Spectrum
                      4
               x 10


         1.6



         1.4



         1.2
Signal




          1

                          Cut-Off Frequency
         0.8                 fc=360MHz
                               λ=9.6 cm
         0.6

                  500       1000   1500       2000        2500   3000
                                   Frequency [MHz]


          A.D.Dane@chem.hull.ac.uk, CPACT                               8
          Department Of Chemistry, University Of Hull, HU6 7RX
GMS Instrumentation




A.D.Dane@chem.hull.ac.uk, CPACT                        9
Department Of Chemistry, University Of Hull, HU6 7RX
  Non-uniform Solid Samples
• Spectral parameters are averages for
  samples, air, sample holder
• Packing density
  – As uniformly as possible




       A.D.Dane@chem.hull.ac.uk, CPACT                        1
       Department Of Chemistry, University Of Hull, HU6 7RX   0
Moisture In Tobacco




                      1
                      1
                                    Moisture in Tobacco
                               55
                                         57       64
                               50        64       61            68         65
                                                                72                  73
                                                                           67
                                                                                    77
                               45
                                         63       62            67.5
                               40        56       64            65
                                                                           58       67
          Moisture level [%]




                                                                           57       66
                               35
                                         57
                                                  68            61.5
                                         55
                                                                62                  58
                               30                 55                       56
                                                                           51       61

                               25
                                         46       53
                                         48                     59         51
                                                  54                                56
                               20                               56         52       55
                                         46
                               15        47
                                                  50
                                                  54
                                                                60         53
                               10                                                   60
                                                                56         53       65
                                5
                                     A        B             C          D        E
                                                       Tobacco type

*Each   symbol represents 8 measurements: 4 measurements at different
positions for 2 samples at each tobacco type/moisture level. Numbers are
sample weights (gram). Reference moisture levels are obtained by weight
         loss upon oven-drying (NDC infrared Engineering, Essex).
                                        Spectra
                       4
                x 10
           2



         1.5

                 Increasing Moisture
           1
Signal




         0.5



           0
                                                                   8.85
                                                                   19.78
         -0.5                                                      29.40
                                                                   38.47
                                                                   47.76
          -1
                           500   1000      1500      2000   2500     3000
                                         Frequency [MHz]


           A.D.Dane@chem.hull.ac.uk, CPACT                                  1
           Department Of Chemistry, University Of Hull, HU6 7RX             3
         Multivariate Analysis
• Principal Component Analysis
  – Visualisation
       • Detect trends and outliers
• Multivariate Calibration
  –   Multivariate Linear Regression (MLR)
  –   Partial Least Squares (PLS)
  –   Non-linear PLS (n-PLS)
  –   Artificial Neural Networks (ANN)

         A.D.Dane@chem.hull.ac.uk, CPACT                        1
         Department Of Chemistry, University Of Hull, HU6 7RX   4
                                 Scores Plot
                                      mean centred data
               0.15
                                                                               50


                0.1                                                            45


                                                                               40




                                                                                    Moisture Content [%]
               0.05
                                                                               35
PC 2 (6.94%)




                  0                                                            30


                                                                               25
               -0.05
                                                                               20

                -0.1
                                                                               15


                                                                               10
                          -0.1     -0.05         0      0.05   0.1      0.15
                                           PC 1 (89.4%)


                 A.D.Dane@chem.hull.ac.uk, CPACT                                                           1
                 Department Of Chemistry, University Of Hull, HU6 7RX                                      5
                                  Scores Plot
                                          mean centred data
               0.15



                0.1



               0.05
PC 2 (6.94%)




                  0



               -0.05
                           A
                           B
                -0.1       C
                           D
                           E
                           -0.1      -0.05         0          0.05     0.1   0.15
                                             PC 1 (89.4%)



                A.D.Dane@chem.hull.ac.uk, CPACT                                     1
                Department Of Chemistry, University Of Hull, HU6 7RX                6
                                       PLS predictions
                                  60
                                       RMSE=1.615
                                        2
                                       R =0.9931
                                  50
   Predicted moisture level [%]


                                  40



                                  30



                                  20
                                                                                        A
                                                                                        B
                                  10                                                    C
                                                                                        D
                                                                                        E
                                  0
                                   0      10       20      30       40        50   60
                                               Reference moisture level [%]



A.D.Dane@chem.hull.ac.uk, CPACT                                                             1
Department Of Chemistry, University Of Hull, HU6 7RX                                        7
        Tobacco Conclusions
• It is feasible to use GMS for determination
  of moisture in tobacco.
• The sample weight is an important variable
  and needs to be measured more accurately
  than was done here.
• It is expected that the predictive ability of
  the models can be increased by further
  optimising the experimental set-up.
• Anal. Chim. Acta, 2001, 429/2. 185-194
       A.D.Dane@chem.hull.ac.uk, CPACT                        1
       Department Of Chemistry, University Of Hull, HU6 7RX   8
                  Crisp Samples

Walkers Crisps                                       Fat (%)
                                                Regular    Light
Ready Salted                                      34         22
Cheese & Onion                                    33         21
Salt & Vinegar                                    33         21



       A.D.Dane@chem.hull.ac.uk, CPACT                             1
       Department Of Chemistry, University Of Hull, HU6 7RX        9
 Non-polar molecules



                                                       Ionic




                                                           Electronic



A.D.Dane@chem.hull.ac.uk, CPACT                                         2
Department Of Chemistry, University Of Hull, HU6 7RX                    0
Guided Microwave Spectra
                        4
                 x 10
            2



          1.5



            1
 Signal




          0.5



            0



          -0.5



           -1
                            500   1000     1500      2000   2500   3000
                                         Frequency [MHz]

          A.D.Dane@chem.hull.ac.uk, CPACT                                 2
          Department Of Chemistry, University Of Hull, HU6 7RX            1
                                  Mean Centring
                      2000


                      1500


                      1000
Mean Centred Signal




                       500


                         0


                       -500


                      -1000


                      -1500
                               500        1000       1500      2000        2500   3000
                                                   Frequency [MHz]

                         A.D.Dane@chem.hull.ac.uk, CPACT                                 2
                         Department Of Chemistry, University Of Hull, HU6 7RX            2
       Multivariate Analysis
• Principal Component Analysis
  – Visualisation - Detect trends/outliers




       A.D.Dane@chem.hull.ac.uk, CPACT                        2
       Department Of Chemistry, University Of Hull, HU6 7RX   3
                                     Scores PC 1-PC 2
                                                    Frequencies>1.7GHz (autoscaling)
                              6
                                                                                                            10.2
                                       5.3     5.5                                           9.9
                              4                                                                 9.9

                                              5.55.5
                              2
                                                                                       9.6
                                              5.5                                 9.2 9.6
                                                                            9.2
   LV 2 (X 8.13%, Y 2.46%)




                                        5.3                                                            10.2
                              0
                                                                                         9.6

                              -2                                      6.1                                     10.2
                                                           5.7                                9.6
                                                          5.5         6.1
                              -4


                              -6
                                                    5.5                                                    10.2

                              -8                                5.7

                             -10
                               -20    -15       -10           -5        0        5           10       15          20
                                                            LV 1 (X 69.4%, Y 91.1%)

Salt & Vinegar, Ready Salted, Cheese & Onion, =Regular,                                                     =Light
                Numbers: (fat%/100)  sample weight
                              A.D.Dane@chem.hull.ac.uk, CPACT                                                          2
                              Department Of Chemistry, University Of Hull, HU6 7RX                                     4
                                       Scores PC 2-PC 3
                                                        All Frequencies (autoscaling)
                           15


                           10                                                     9.6
                                                                                           9.6           5.5
                                                                                        9.2       9.2
                                                              10.2                               5.5
                                                                          9.6
                             5
 LV 3 (X 13.2%, Y 2.42%)




                                                                                9.6
                                                              5.7
                                                 10.2
                             0                                                                                 5.5
                                                 6.1

                                           6.1                                    5.5              5.3
                                           5.7                                                                       9.99.9
                            -5                                      5.5                          5.5

                                                                                5.3          10.2
                           -10


                           -15
                                                                                           10.2

                           -20
                             -20        -15         -10         -5         0                 5             10           15
                                                          LV 2 (X 18.8%, Y 2.25%)
Salt & Vinegar, Ready Salted, Cheese & Onion, =Regular,                                                               =Light
                Numbers: (fat%/100)  sample weight
                                 A.D.Dane@chem.hull.ac.uk, CPACT                                                               2
                                 Department Of Chemistry, University Of Hull, HU6 7RX                                          5
     Crisps PCA Conclusions
• Trend of changing fat content has been
  observed suggesting that modelling is
  feasible
• Different flavours show different behaviour
  – If this is a problem model each flavour
    separately
• There is a rather large spread between
  replicates.
  – Packaging
       A.D.Dane@chem.hull.ac.uk, CPACT                        2
       Department Of Chemistry, University Of Hull, HU6 7RX   6
                        Packaging
• Averaging
  – Multiple stacking
  – Multiple Orientations
  – Larger Volume (no Sample Holder)
• Some measure of packing density
  – Weight
• Multivariate Calibration
  – Weight as an extra variable
  – Model (%/100)  sample weight instead of %
• Sieving or Grinding
       A.D.Dane@chem.hull.ac.uk, CPACT                        2
       Department Of Chemistry, University Of Hull, HU6 7RX   7
     Conclusions and outlook
• It is demonstrated that GMS in combination
  with multivariate techniques can be used to
  obtain both qualitative and quantitative
  results for solid samples.
• More research is on going, also on liquid
  samples.




       A.D.Dane@chem.hull.ac.uk, CPACT                        2
       Department Of Chemistry, University Of Hull, HU6 7RX   8
          Acknowledgements
• Gerry Rea, Tony Walmsley and Steve Haswell
• Chris Walker and Steward MacKenzie (VG Gas)
• Julian Iredale, Jane Waller, Bob Hammond and
  Richard Warren (NDC Infrared Engineering)
• Professor F. Alder (UMIST)
• Eduard Derks (Quest International, Naarden, NL)




       A.D.Dane@chem.hull.ac.uk, CPACT                        2
       Department Of Chemistry, University Of Hull, HU6 7RX   9

								
To top