Nutritional Supplement and Diesel Fuel Application Development for Benchtop NMR

Systems Operating at 42, 60, and 80 MHz Equivalency with Supercon NMR

SpinMetrix

John C. Edwards1,2, Gonzalo Hernandez2,3, and Paul J. Giammatteo1

1) Process NMR Associates LLC, Danbury, CT USA. 2) SpinMetrix SRL, Montevideo, Uruguay, and 3) Vis Magnetica, Montevideo, Uruguay

Benchtop high-resolution NMR systems are available at a number of field strengths and probe configurations. However beyond the obvious academic instruction market for these instruments very few applications have been demonstrated
across all available platforms and thus proving the general applicability of benchtop NMR technology to industrial quality control. We will present two chemometric-based applications that have been developed at 4 different field strengths
utilizing Varian Mercury 300 MHz, Magritek Spinsolve 42 MHz, Aspect AI 60 MHz, and Thermo Picospin 80 MHz NMR systems. Partial-least-squares (PLS) regression correlations were obtained on all 4 platforms relating to:
1) Omega-3 fatty acid composition of samples taken from various points in a nutritional supplement manufacturing process. Excellent correlations were obtained on all 4 NMR instruments proving that NMR technology is applicable to in-lab,
at-line. or on-line analysis of fish oil derived omega-3 fatty acid supplements. The 40 second NMR analysis effectively replaces a 60+ minute GC analysis.
2) Physical and chemical property determination of diesel fuels where excellent correlations were obtained between 1H NMR variability and parameters such as density, aromatic content by GC, hydrogen content by 1H TD-NMR (ASTM D7171
method), and sulfur content. Many more physical and chemical properties can be correlated to the 1H NMR spectrum allowing a single 40 second NMR experiment to predict 10-15 parameters that each require dedicated analyzers.
Finally, we will present the concept and initial results from an independent server-based NMR application software that can be utilized in conjunction with the NMR software of the current benchtop NMR systems, or alternatively as a standalone application platform. This software would effectively make chemometric and direct measurement NMR application ubiquitous across all NMR platforms.

Docosahexaenoic Acid (DHA) 22:6(n-3)

DHA (Area %)
0.01
13.34
15.66
18.16
9.54
28.90
10.51
0.23
10.04
16.47
21.26
5.95
7.43
25.91
0.06
12.23
19.30
2.78
0.72
23.41
5.18
21.52
28.18
0.00
12.82

Sample Description
First Esterification
First Esterification
Clathration
Mol Dist
Pollock Oil
Separator
PolyUnsat Ester
First Esterification
First Esterification
Clathration
Mol Dist
MSC Pollock Oil
Separator
PolyUnsat Ester
First Esterification
First Esterification
Clathration
Clath Raffinate
First Esterification
Mol Dist
MSC Pollock Oil
Separator
PolyUnsat Ester
First Esterification
First Esterification

300 MHz 1H NMR Spectra of Samples at Different Points in the Process

300 MHz 1H NMR

Experimental
Varian Mercury - 300 MHz NMR Supercon Spectrometer
4 pulse on pure sample in 5 mm tube, Run Unlocked
In MNova 8.1.2 SPC Files Imported, Stacked, Binned at 3 Hz interval,
Area Normalized to 100, Saved as Transposed Ascii Matrix
For Peak Integrals Used Advanced Feature Create Integral Graph
from Stacked Plot
PLS Regression Performed Thermo Grams IQ and Eigenvector Solo

EPA

First round of PLS regression analysis

revealed concentration outliers that
were linked to limitations in the GC
method. This study was performed on
improved GC data values.
Final 80 Sample models were utilized
to validate the calibrations on a 24
sample validation set (below).
NMR ID
FO3h026
FO3h027
FO3h028
FO3h029
FO3h030
FO3h031
FO3h032
FO3h033
FO3h034
FO3h035
FO3h036
FO3h037
FO3h038
FO3h039
FO3h040
FO3h041
FO3h042
FO3h043
FO3h044
FO3h045
FO3h046
FO3h047
FO3h048
FO3h049
FO3h050

EPA (Area %) DHA (Area %)

44.55
19.30
16.69
9.89
6.87
4.53
62.79
19.43
37.29
22.03
9.71
0.38
32.00
15.10
38.79
26.75
41.87
23.25
35.49
43.99
0.30
0.00
34.09
8.09
15.44
9.82
60.08
24.52
8.77
5.75
12.41
57.59
36.36
21.49
3.79
0.15
29.23
13.78
45.99
33.51
12.10
5.69
24.57
14.32
45.86
33.61
6.39
3.68
58.13
24.66

EPA

DHA

600

500

Integration of Peaks to
Produce Multivariate Integration Spectra

400

300

200

100

-100

6
Variables

150

10

11

0.4
Leverage

0.6

0.8

-200
0
200
Scores on LV 1 (95.40%)

400

4
3

Q Residuals (0.11%)

NMR Processing for Multivariate Analysis

Sample Description
Clathration
First Esterification
Crude Pollock Oil
Separator
PolyUnsat Ester
First Esterification
First Esterification
Clathration
Mol Dist
Separator
First Esterification
Clathration
MonoUnsat Ester
PolyUnsat Ester
Crude Salmon Oil
Separator
PolyUnsat Ester
First Esterification
Clath Raffinate
PolyUnsat Ester
MSC Pollock Oil
Separator
PolyUnsat Ester
First Esterification
Clathration

Expansion of Correlated Region

Y Stdnt Residual 1

Eicosaoentaenoic Acid (EPA) 20:5(n-3)

EPA (Area %)
0.64
21.55
62.97
29.43
14.21
52.74
15.21
7.18
16.95
36.35
61.09
13.32
71.78
41.40
1.19
11.73
43.38
6.07
9.77
58.93
10.62
43.91
54.05
0.00
26.97

Data

NMR ID
FO3h001
FO3h002
FO3h003
FO3h004
FO3h005
FO3h006
FO3h007
FO3h008
FO3h009
FO3h010
FO3h011
FO3h012
FO3h013
FO3h014
FO3h015
FO3h016
FO3h017
FO3h018
FO3h019
FO3h020
FO3h021
FO3h022
FO3h023
FO3h024
FO3h025

50 sample initial data set from

all points in manufacturing process.
40 Samples used for 80 MHz dataset

100

50

2
1
0
-1
-2

-3
0

10
15
20
Hotelling T^2 (99.89%)

25

30

300 MHZ
Wt% DHA by GC
300 MHz NMR
Regression Vector
Wt% DHA by GC

Wt% EPA

60

40

R2 =

0.993
8 Latent Variables
RMSEC=1.20
RMSECV=1.68

20

Wt% DHA by GC

40

Scores on LV 2 (0.45%)

Y CV Predicted 1

80

0.2

20

40
Y Measured 1

60

20

-20

-40
-400

80

Variables/Loadings Plot for 1H NMR - 55 Fish Oil Samples - Processed - 8-9-13 - Transposed.xlsx
0.04

Omega-3 Ethyl Ester Sample obtained at 4 field strengths ppm scale

300 MHz NMR

Regression Vector
Wt% DHA by GC

0.03

Aspect AI 60.38 MHz

Magritek Spinsolve 42.5 MHz

4 pulse on pure sample in 5 mm tube

Locked on 1H NMR signal
In MNova 8.1.2
SPC Files Imported, Stacked, Binned at 3 Hz interval, Area Normalized to 1000
Saved as Transposed Ascii Matrix
For Peak Integrals Used Advanced Feature Create Integral Graph from
Stacked Plot
PLS Regression Performed Thermo Grams IQ and Eigenvector Solo

4 pulse on pure samples in 5mm tube

Data Processed in Mestrelab Mnova 9
Data converted to 0.05 ppm integration
bins
Spectral area Normalized to 1000
Saved as transposed matrix

Reg Vector for Y 1

0.02

0.01

-0.01

-0.02

-0.03

20

40

60

80

100

120

Variable

Thermo Picospin80 82.3 MHz

30

25

15
10

1
0
-1

80

15

-2

42.5 MHz
Wt% EPA

60

-3
5
10
15
Hotelling T^2 (99.91%)

20

30
80

PLS Regression Data

42 MHz NMR
60 MHz NMR
82 MHz NMR
300 MHz NMR
Peak Integral Data
Fused - NMR-FTIR

The 1H NMR PLS correlations to EPA and DHA content in the various manufacturing
streams and products are provided in Table I and demonstrate that all 4 NMR
systems produce excellent correlations. Good correlations can also be obtained from
gross integration region spectra indicating that model robustness may be improved
by data reduction of NMR spectra into wider chemically relevant integration bins.

60
20

60 MHz
Wt% EPA

15
40
10

R2=0.993
SECV=2.13 wt%
8 Latent Variables

20
5
00
00

5 20 10
15
20 60 25
40
Hotelling
T^2 (99.91%)
Y Measured
1

0.8

50

30

R2=0.992
SECV=1.13 wt%
7 Latent Variables

20
10

10

20
30
40
Y Measured 1

50

60

R2=0.988
SECV=2.20 wt%
8 Latent Variables
0

20

40

00

60

0
-5
-10
-15

42 MHz - EPA

-20

-20
-100

80

-50

Y Measured 1

80 MHz
Wt% EPA

42.5 MHz - DHA

42.5 MHz
Wt% DHA

60

0.8
200

50

20

-2
-10

0.2
0.40
0.6
-100
100
ScoresLeverage
on LV 1 (98.04%)

Scores on LV 1 (95.78%)

80

60 MHz - DHA

40

20

20

60 MHz
Wt% DHA

40

2
10

-4
-20
0 -200

30
80

60

0.6

60 MHz - EPA

Scores on LV 2 (0.36%)

EPA Content (Wt%) DHA Content (Wt%)

R2, SECV (Wt%)
R2, SECV (Wt%)
0.988, 2.20
0.989, 1.25
0.993, 2.13
0.992, 1.13
0.989, 2.17
0.992, 1.11
0.988, 1.68
0.991, 1.09
0.981, 2.71
0.964, 2.24
0.993, 1.62
0.992, 1.08

Y CV Predicted 1

Table I: Regression Results for Wt% DHA/EPA Content at 4 Field Strengths

(0.09%)
QYResiduals
1
CV Predicted

25

0.4
Leverage

4
20

2 (0.64%)
on LV
Scores
1
Residual
Y Stdnt

Omega-3 Ethyl Ester Sample obtained at 4 field strengths frequency scale

0.2

Y CV Predicted 1

10

40
Y CV Predicted 1

10

Scores on LV 2 (2.23%)

20

Scores on LV 2 (2.48%)

Y Stdnt Residual 1

Q Residuals (0.09%)

32 pulse average on pure sample

Data Processed in Mestrelab Mnova 9
Data converted to 0.05 ppm integration bins
Spectral area Normalized to 1000
Saved as transposed ascii matrix

R^2 = 0.989
20

8 Latent Variables
RMSEC = 0.75363
RMSECV = 1.2463

10

Calibration Bias = 0

-10

CV Bias = 0.098796
-20
0

10

20

30

40

50

-20
-100

60

Y Measured 1

-50

80 MHz
Wt% DHA

50

Scores on LV 1 (95.78%)

-10

-20

-200

-100
0
100
Scores on LV 1 (97.83%)

200

1H

benchtop NMR has great potential to increase the throughput of both routine
and emergency fuel sample analysis in refinery laboratories. Currently fuel
samples must be passed through multiple dedicated analyzers to obtain
information such as density, H-Content, aromatics, olefins, saturates, benzene,
Octane numbers, cetane index, cetane number, distillation curves, vapor
pressure, flash point, pour point, freeze point, cloud point, etc. Correlation of the
1H NMR spectra of these refinery fuel samples to these primary test results will
allow all parameters to be predicted in about 40 seconds from the 4 pulse
spectrum of the pure fuel. Here we have a few examples obtained on some diesel
fuels that were submitted to our lab for ASTM D7171 Hydrogen Content by TDNMR. We had density, H-content, and aromatics wt% by GC. Below are three
example correlation obtained on the Picospin 80 system (that requires 32 pulses
per sample due to the capillary sample size). The results were very similar for the
300, 60, and 42 MHz data obtained on the three other NMR system in our
laboratory. The comparative results are shown in Table II. The results are very
similar independent of the field strength of the NMR system. The data from all 4
NMR systems is provided in this section.

Diesel Fuels - 300 MHz

4 Pulse Average

Diesel Fuels - 80 MHz

32 Pulse Average

Diesel Fuels 42 MHz

4 Pulse Average

Diesel Fuels - 60 MHz

4 Pulse Average

SpinMetrix is a joint venture company between Process NMR

Associates and VisMagnetica developing an independent NMR
application implementation software for full spectral processing
automation, pre-processing for chemometric applications ,
chemometric model prediction, and results reporting. Here is an
example webpage where NMR FID data obtained on Varian 300,
Magritek Spinsolve, Picospin-80, and Aspect-60 can be uploaded
pre-processed and passed through server based PLS models for
multiple parameter prediction. Models can be accessed on-line or
locally on the spectrometer computer.

Table II: PLS Regression Results for Diesel Quality Parameters at 4 Field Strengths
1H

NMR and Diesel Parameter Correlation PLS Regression

Hydrogen Content (%H) ASTM D7171

TD-NMR
NMR Freq

Density (kg/L)

Aromatics (wt%)
Hydrogen Content (%H)

Aromatic Content (Wt% GC)

Density (Kg/L)

R2

SECV

LV

R2

SECV

LV

R2

SECV

LV

42 MHz

0.951

0.10

0.959

0.54

0.944

0.0021

60 MHz

0.952

0.09

0.962

0.53

0.951

0.0020

80 MHz

0.934

0.10

0.953

0.57

0.937

0.0022

300 MHz

0.974

0.07

0.953

0.57

0.937

0.0021

Conclusion
A wide range of PLS correlation models can be readily built
based on NMR data obtained on both superconducting and
benchtop permanent magnet NMR systems. Currently models
require that data be obtained on each individual spectrometer
system but it may be possible that various spectral deresolution techniques may make models obtained on one
system transferable between NMR systems at varying magnetic
field strengths.
At-line and in-line permanent magnet NMR systems can
yield the same high quality correlations as data obtained on
much higher field superconducting NMR systems. Very little
difference is observed in the quality of the correlations and
errors of prediction on models developed on the 4 systems in
our laboratory.