Multivariate analysis of ToF-SIMS datasets
Wednesday, 20 September 2017 1
Gustavo Ferraz Trindade
SIRG Group Meeting 05/02/2016
Outline
Wednesday, 20 September 2017 2
MVA methods applied to ToF-SIMS
Examples
(Gustavo & SIRG members)
MVA methods
Wednesday, 20 September 2017 3
m1 m2
I1
I2
1
m1 m2
I1
I2
2
m1 m2
I1
I2
3
m1 m2
I1
I2
4
I1
I2
2
1
4
3
A Spectrum
Wednesday, 20 September 2017 4
I1
I2
2
1
4
3
PC1
PC2
“Rotation of axis” (keeps the orthogonal angle)
PCA
Wednesday, 20 September 2017 5
m1 m2
PC1
m1
m2
PC2
PC1 explains all the data
PC2 noise
(variables randomly distributed)
PCA
Wednesday, 20 September 2017 6
m1 m2
PC1
“Loadings” “Scores”
1 2 3 4 5 6 7
samples
PCA
Spectra with similar scores are from
similar samples
Wednesday, 20 September 2017 7
PC1
PCA
With more PCs it is usual to represent
the scores with scatter plots
PC2
From thousands of possible scatter
plots with the original variables, PCA
enables the whole dataset to be
shown in one single graph
Wednesday, 20 September 2017 8
Purespectrum
Purespectrum
Concentrations Concentrations
Samples (observations)
Fragments (variables)
+
NMF: Non-negative Matrix Factorization (aka MCR)
- does not allow negative solutions
- Easier to interpret than PCA
- Multiple solutions
NMF
Wednesday, 20 September 2017 9
M A*BT
+= =
(+ error)
NMF
(2 “pure compounds”)
(3 “pure compounds”)(4 “pure compounds”)(5 “pure compounds”)
and so on…
Wednesday, 20 September 2017 10
In SIMS, the different MVA observations or samples are always
spectra and they can be:
- Of different samples or technical repeats
- Of different pixels in an imaging data set of one sample
- Of different levels in a depth profile of one sample
The variables are the different intensities of mass peaks
(or channels, or time-of-flight)
MVA Methods
Wednesday, 20 September 2017 11
PCA as an
exploratory tool
Organic coating + microtome (R. Tshulu)
Analysis of organic coating
Wednesday, 20 September 2017 12
Interface fragments
Organic coating + microtome (R. Tshulu)
Analysis of organic coating
Wednesday, 20 September 2017 13
Organic coating + microtome (R. Tshulu)
Linescan to summarize
PCA results Interface
Analysis of organic coating
Wednesday, 20 September 2017 14
Analysis of pine wood
PLUS standard lignin and cellulose samples
Carefully selected peak list
Analysis of wood growth regions
Wednesday, 20 September 2017 15
- NMF with 2 compounds
- Standard samples enabled “relative quantification”
- Small differences between early/late wood observed
- Possibly wood anatomy + extractives
Analysis of wood growth regions
Wednesday, 20 September 2017 16
Next steps:
- add data of more samples (J. Banuls)
- Include extractives peaks and try relative quantification with more compounds
PC1
PC2
(extractives?)
PC3
PDMS
73
147
Analysis of wood growth regions
Wednesday, 20 September 2017 17
Automotive grade polypropylene
Very similar set of spectra
PCA to identify
“unknown” sample
U Unknown
Wednesday, 20 September 2017 18
Automotive grade polypropylene
Unknown is A
Scores of PC1 vs. PC2
Wednesday, 20 September 2017 19
NormalisedIntensity
 Automotive grade polypropylene under 150 C (IONTOF H/C stage)
 Not a “depth profile” but a “time profile”
 Surface is changing due to species segregation
 POSSIBLY DAMAGE (I have to do less scans per second to confirm)
Time / s (at 150 C)
NMF to separate
behaviours of
different compounds
Automotive grade polypropylene
Wednesday, 20 September 2017 20
NMF after 100 iterations (3 pure compounds)
Time / 20s (at 150 C)
NormalisedIntensity
Polypropylene PP + Irganox 1010 PDMS
Automotive grade polypropylene
Cold seal + BOPP
Adhesive/polymer interaction
Wednesday, 20 September 2017 22
Cold seal
BLUE
RED
Adhesive/polymer interaction
Wednesday, 20 September 2017 23
Reverse print
Adhesive/polymer interaction
RED
BLUE
Wednesday, 20 September 2017 24
Transfer of material
- Helped to understand the process spatially and aided
spectral analysis
Adhesive/polymer interaction
RED
BLUE
NMF to help on first guess on XPS peak fitting
Example: C 1s peaks of dicarboxylic acids with different chain lengths (J. Ferreira)
- Scales interpolation (spectra have different shifts)
- Resampling number of channels
- Smoothing
- NMF
etc…
Example: C 1s peaks of dicarboxylic acids with different chain lengths (J. Ferreira)
Results for 3 compounds
NMF to help on first guess on XPS peak fitting
PCA and NMF can be very useful for ToF-SIMS analysis
Conclusions
- Results only depend on pre-processing
- One solution
- Quicker
- Results more difficult to interpret/present
- No straightforward quantification
PCA NMF
- Needs number of compounds/algorithm
- Multiple possible solutions
- Takes longer
- Results easier to interpret/present
- Allows relative quantification
I recommend simsMVA!!
Conclusions
s i m s M V A
www.mvatools.com
Conclusions
xkcd
Thank you

Multivariate analysis of ToF-SIMS datasets

  • 1.
    Multivariate analysis ofToF-SIMS datasets Wednesday, 20 September 2017 1 Gustavo Ferraz Trindade SIRG Group Meeting 05/02/2016
  • 2.
    Outline Wednesday, 20 September2017 2 MVA methods applied to ToF-SIMS Examples (Gustavo & SIRG members)
  • 3.
    MVA methods Wednesday, 20September 2017 3 m1 m2 I1 I2 1 m1 m2 I1 I2 2 m1 m2 I1 I2 3 m1 m2 I1 I2 4 I1 I2 2 1 4 3 A Spectrum
  • 4.
    Wednesday, 20 September2017 4 I1 I2 2 1 4 3 PC1 PC2 “Rotation of axis” (keeps the orthogonal angle) PCA
  • 5.
    Wednesday, 20 September2017 5 m1 m2 PC1 m1 m2 PC2 PC1 explains all the data PC2 noise (variables randomly distributed) PCA
  • 6.
    Wednesday, 20 September2017 6 m1 m2 PC1 “Loadings” “Scores” 1 2 3 4 5 6 7 samples PCA Spectra with similar scores are from similar samples
  • 7.
    Wednesday, 20 September2017 7 PC1 PCA With more PCs it is usual to represent the scores with scatter plots PC2 From thousands of possible scatter plots with the original variables, PCA enables the whole dataset to be shown in one single graph
  • 8.
    Wednesday, 20 September2017 8 Purespectrum Purespectrum Concentrations Concentrations Samples (observations) Fragments (variables) + NMF: Non-negative Matrix Factorization (aka MCR) - does not allow negative solutions - Easier to interpret than PCA - Multiple solutions NMF
  • 9.
    Wednesday, 20 September2017 9 M A*BT += = (+ error) NMF (2 “pure compounds”) (3 “pure compounds”)(4 “pure compounds”)(5 “pure compounds”) and so on…
  • 10.
    Wednesday, 20 September2017 10 In SIMS, the different MVA observations or samples are always spectra and they can be: - Of different samples or technical repeats - Of different pixels in an imaging data set of one sample - Of different levels in a depth profile of one sample The variables are the different intensities of mass peaks (or channels, or time-of-flight) MVA Methods
  • 11.
    Wednesday, 20 September2017 11 PCA as an exploratory tool Organic coating + microtome (R. Tshulu) Analysis of organic coating
  • 12.
    Wednesday, 20 September2017 12 Interface fragments Organic coating + microtome (R. Tshulu) Analysis of organic coating
  • 13.
    Wednesday, 20 September2017 13 Organic coating + microtome (R. Tshulu) Linescan to summarize PCA results Interface Analysis of organic coating
  • 14.
    Wednesday, 20 September2017 14 Analysis of pine wood PLUS standard lignin and cellulose samples Carefully selected peak list Analysis of wood growth regions
  • 15.
    Wednesday, 20 September2017 15 - NMF with 2 compounds - Standard samples enabled “relative quantification” - Small differences between early/late wood observed - Possibly wood anatomy + extractives Analysis of wood growth regions
  • 16.
    Wednesday, 20 September2017 16 Next steps: - add data of more samples (J. Banuls) - Include extractives peaks and try relative quantification with more compounds PC1 PC2 (extractives?) PC3 PDMS 73 147 Analysis of wood growth regions
  • 17.
    Wednesday, 20 September2017 17 Automotive grade polypropylene Very similar set of spectra PCA to identify “unknown” sample U Unknown
  • 18.
    Wednesday, 20 September2017 18 Automotive grade polypropylene Unknown is A Scores of PC1 vs. PC2
  • 19.
    Wednesday, 20 September2017 19 NormalisedIntensity  Automotive grade polypropylene under 150 C (IONTOF H/C stage)  Not a “depth profile” but a “time profile”  Surface is changing due to species segregation  POSSIBLY DAMAGE (I have to do less scans per second to confirm) Time / s (at 150 C) NMF to separate behaviours of different compounds Automotive grade polypropylene
  • 20.
    Wednesday, 20 September2017 20 NMF after 100 iterations (3 pure compounds) Time / 20s (at 150 C) NormalisedIntensity Polypropylene PP + Irganox 1010 PDMS Automotive grade polypropylene
  • 21.
    Cold seal +BOPP Adhesive/polymer interaction
  • 22.
    Wednesday, 20 September2017 22 Cold seal BLUE RED Adhesive/polymer interaction
  • 23.
    Wednesday, 20 September2017 23 Reverse print Adhesive/polymer interaction RED BLUE
  • 24.
    Wednesday, 20 September2017 24 Transfer of material - Helped to understand the process spatially and aided spectral analysis Adhesive/polymer interaction RED BLUE
  • 25.
    NMF to helpon first guess on XPS peak fitting Example: C 1s peaks of dicarboxylic acids with different chain lengths (J. Ferreira) - Scales interpolation (spectra have different shifts) - Resampling number of channels - Smoothing - NMF etc…
  • 26.
    Example: C 1speaks of dicarboxylic acids with different chain lengths (J. Ferreira) Results for 3 compounds NMF to help on first guess on XPS peak fitting
  • 27.
    PCA and NMFcan be very useful for ToF-SIMS analysis Conclusions - Results only depend on pre-processing - One solution - Quicker - Results more difficult to interpret/present - No straightforward quantification PCA NMF - Needs number of compounds/algorithm - Multiple possible solutions - Takes longer - Results easier to interpret/present - Allows relative quantification
  • 28.
    I recommend simsMVA!! Conclusions si m s M V A www.mvatools.com
  • 29.

Editor's Notes

  • #2 Ethane dioic (carboxylic)
  • #3 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #4 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #5 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #6 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #7 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #8 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #9 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #10 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #11 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #12 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #13 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #14 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #15 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #16 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #17 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #18 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #19 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #20 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #21 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #22 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #23 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #24 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #25 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #26 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #27 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #28 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #29 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.
  • #30 - Fazer um rapido overview da apresentacao. Quando falar de resultados mencionar que foram feitas analyses SEM, EDX, XPS e SIMS.