Szilveszter GERGELY Department of Applied Biotechnology and Food Science

(1)

The invisible world in my infrared eyes

MVDA aided (N)IR spectroscopy and imaging

Szilveszter GERGELY Department of Applied Biotechnology and Food Science

Budapest University of Technology and Economics

Guest lecture at PCCL from BUTE NIR Spectroscopy Group 13

th

of September, 2018 ● Leoben, Austria

2

The main topic of this lecture:

The invisible world in my infrared eyes

3

Newton’s experiments (over the white light)

(1704)

It was known earlier, but he wrote

it down...

(2)

4

Herschel’s experiments (below the red light)

11

^th

of February, 1800

Discovery of Uranus 13thof March,

1781 from court of

his house.

5

Coblentz’s experiments (the IR fingerprint)

(1905)

Ethanol CH3–CH2–OH Dimethyl ether H3C–O–CH3

C2H6O, but IR sp. ≠!

6

IR – heat – energy

An AC-130U gunship fires flares to emit masses of infrared and confuse heat seeking missiles.

http://www.wired.com/dangerroom/2012/10/infrared-obscurant/

Using IR against the

fire-and- forget missiles

The small

„stars” are more warm, than the engines

(3)

7

Developing of instruments:

from the military usage...

http://www.wired.com/dangerroom/2013/04/darpa-infrared-cameras/

A short movie?

DARPA:

„Cheetah”

Aims today:

smaller instrument,

better resolution

http://www.newsweek.com/2015/01/16/making-blood-draws-easier-friendlier-needle-296317.html 8 A short move?

(IDRC 2002 Baltimore, hospital ship) NIR 2007 Umeå (S):

painting

„H2O” on white paper by water

Developing of instruments:

... to civil applications

9

The spectrum is the sum total of changes of states:

chemical AND physical „fingerprint”

bending (deformation) changes in angle of bonds

symmetric stretching changes in length of bonds

asymmetric stretching changes in length of bonds

Vibrations of molecules

Paradigm shift with changes of millennium

NIR:

molecule / vibrational spectroscopy

That is why we can measure moisture content

(4)

10

Moisture content below 0.5% vs. NIR (artificial fertilizer, 1 ^st test)

0,15 0,2 0,25 0,3 0,35 0,4 0,45 0,5

2018.03.14 18:45 2018.03.15 12:00 2018.03.16 4:00 2018.03.16 20:00 2018.03.17 12:00 2018.03.18 4:00 2018.03.18 20:00 2018.03.20 2:00

KF víztartalom Perten víztartalom

Royal Society of Chemistry (RSC): Infrared Spectroscopy (IR). (2009) 11

http://www.rsc.org/learn-chemistry/resource/res00000283/spectroscopy-in-a-suitcase-ir-student-resources#!cmpid=CMP00001302

Vibrations of molecules

NIR:

molecule / vibrational spectroscopy

In case of MIR:

more sharp peaks S/N: 100×

Murray I.: Scattered information: philosophy and practice of near infrared spectroscopy. 12 In Near Infrared Spectroscopy: Proceedings of the 11th International Conference, Ed by Davies A.M.C., Garrido-Varo A., NIR Publications, Chichester, pp. 1–12 (2004).

Bonds have a vibrational “answer”

All we need:

water, proteins, lipids, carbo-

hydrates

(5)

13

A quick example for NIR spectra

0 1 2 3

1100 1300 1500 1700 1900 2100 2300 2500 Wavelength (nm)

OD

0 1 2 3

1100 1300 1500 1700 1900 2100 2300 2500 Wavelength (nm)

OD

Water (H

₂

O)

Ethanol (CH

₃

–CH

₂

–OH)

0 1 2 3

1100 1300 1500 1700 1900 2100 2300 2500 Wavelength (nm)

OD

Wine

10 mm 4 mm 1 mm 10 mm

4 mm 1 mm

10 mm 4 mm 1 mm

Gergely S., Farkas K., Forgács A., Salgó A.: Quantitative and qualitative differentiations of alcoholic beverages by near infrared spectroscopy.

In Near Infrared Spectroscopy: Proceedings of the 11th International Conference, Ed by Davies A.M.C., Garrido-Varo A., NIR Publications, Chichester, pp. 569–572 (2004).

Many variables,

small differences

► MVDA

14

»chemical and physical „fingerprint”«

Dahm D.J, Dahm K.D.: The Physics of Near-Infrared Scattering.

In Near-Infrared Technology in the Agriculture and Food Industries, Ed by Williams P., Norris K., Association of Cereal Chemists Inc., St. Paul, pp. 1–17 (2004).

15

»chemical and physical „fingerprint”«

Dahm D.J, Dahm K.D.: The Physics of Near-Infrared Scattering.

In Near-Infrared Technology in the Agriculture and Food Industries, Ed by Williams P., Norris K., Association of Cereal Chemists Inc., St. Paul, pp. 1–17 (2004).

This is only glass + air / water / oil.

What are inside a grain?

(6)

16

Particle size vs. NIR

0 0,1 0,2 0,3 0,4 0,5 0,6 0,7

950 1000 1050 1100 1150 1200 1250 1300 1350 1400 1450 1500 1550 1600 1650

A= lg (1/R)

Hullámhossz (nm)

2,0 mm 1,4 mm 1,0 mm 0,5 mm 0,3 mm

„Mixed feed”

Alja

17

How we can measure? – 3-in-1

dispersive (DS)

diode array (DA)

Fourier trans- formed (FT)

„bench-top”:

pre-

„process”:

post- dispersive

Robust; prompt data in the whole range (moving smpl, ON-LINE !!!) In this case:

pre- dispersive arrangement

Always post- dispersive

Sumriddetchkajorn S., Chakkrit Kamtongdee C.: 18 Optical penetration-based silkworm pupa gender sensor structure.

Appl. Optics51(4), 408–412 (2012)

How we can measure? – NIR LEDs

Non- destructive sexing of silkworms within pupa

(7)

Sumriddetchkajorn S., Chakkrit Kamtongdee C.: 19 Optical penetration-based silkworm pupa gender sensor structure.

Appl. Optics51(4), 408–412 (2012)

How we can measure? – NIR LEDs

20

How we can measure? – NIR lasers (tunable)

Pál MAÁK / Dept. of Atomic Physics, BME Non-

destructive sorting of watermelon

21

How we can measure? – NIR lasers (tunable)

Pál MAÁK / Dept. of Atomic Physics, BME

(8)

22

Step by step – (IQ)Q = IQ ²

IQ ► e.g. pharma

Q ► e.g. agro-food (overlapping:

e.g. moisture)

Identification

Who are you?

searching for sharp spectral differences

correlation, Euclidean distances etc.

Qualification

Where do you belong to?

define subpopulation based on more fine differences

principal component analysis (PCA), Mahalanobis distance etc.

Quantification

How many?

calibrations with reference parameters

partial least squares (PLS) method etc.

Main elements:

well-defined sample sets

accurate reference methods

reliable spectroscopic tools

exact mathematics and chemometry

Overview of calibration and validation

23

Calibration

to find multivariate correlation between spectroscopic and reference data sets

Validation

to estimate the performance of the calibration model using independent sample set or cross-validation

Overview of calibration and validation

24

(9)

X_cal y_cal

reference data for calibration spectroscopic

matrix for calibration spectrum selection spectrum transformation

representative sample set

scanning

statistical evaluation

calibration modell (MLR, PLS, ANN)

Calibration

reference measurements

25

independent sample set, cross-validation

spectroscopic matrix for validation

scanning reference

measurement

reference data (measured) reference

data (predicted)

y’val

y’_val

yval yval

+ ++++ ++ ++ ++ ++++++++

++++ ++++

+

Xval

comparison (scatter plots)

Validation

26

27

Analysis of reference data

Box & Whisker Plot Median 25%-75% Min-Max Raw Data

2,00 2,20 2,40 2,60 2,80 3,00 3,20 3,40 3,60 3,80 4,00

2,03 2,29 2,40 2,63 2,83 3,06 3,27 3,48 3,88 4,13

Elméleti hatóanyagtartalom [%] Mért (HPLC) hatóanyagtartalom [%]

1,5 2,0 2,5 3,0 3,5 4,0 4,5

3,67

Box & Whisker Plot Mean Mean±SD Mean±1,96*SD

2,00 2,20 2,40 2,60 2,80 3,00 3,20 3,40 3,60 3,80 4,00

2,03 2,29 2,40 2,63 2,83 3,06 3,27 3,48 3,67 3,88 4,13

Elméleti hatóanyagtartalom [%] Mért (HPLC) hatóanyagtartalom [%]

1,5 2,0 2,5 3,0 3,5 4,0 4,5

Bar/Column Plot of multiple variables CLX gélek analitikai eredményei in CLX gélek analitikai eredményei 5v*11c Elméleti hatóanyagtartalom [%] Mért (HPLC) hatóanyagtartalom [%]

2,00 2,20

2,40 2,60

2,80 3,00

3,20 3,40

3,60 3,80

4,00

2,03 2,29

2,40 2,63

2,83 3,06

3,27 3,48

3,67 3,88

4,13

1 2 3 4 5 6 7 8 9 10 11

0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0 4,5

Scatterplot: Elméleti hatóanyagtartalom [%] vs. Mért (HPLC) hatóanyagtartalom [%]

Mért (HPLC) hatóanyagtartalom [%] = -,0495 + 1,0368 * Elméleti hatóanyagtartalom [%]

Correlation: r = ,99925

1,8 2,02,22,4 2,62,8 3,03,2 3,43,63,8 4,04,2

Elméleti hatóanyagtartalom [%]

1,8 2,0 2,2 2,4 2,6 2,8 3,0 3,2 3,4 3,6 3,8 4,0 4,2

Mért (HPLC) hatóanyagtartalom [%]

0,95 Conf.Int.

SEL= 0,017044 R²MAX

= 0,999385

Never forget:

the reference laboratory has error, too : standard error of laboratory

(10)

Analysis of reference data

Differences between day-shift and

night shift:

controller is sleeping...

28 Box & Whisker Plot: nedvesség

Mean ±0,95 Conf. Interval ±SD

1 2

# 8,0 8,5 9,0 9,5 10,0 10,5 11,0 11,5 12,0

nedvesség

Histogram: nedvesség K-S d=,13839, p<,01 ; Lilliefors p<,01

Expected Normal

7,2 7,4

7,6 7,8

8,0 8,2

8,4 8,6

8,8 9,0

9,2 9,4

9,6 9,8

10,0 10,2

10,4 10,6

10,8 11,0

11,2 11,4

11,6 11,8

12,0 12,2 X <= Category Boundary 0

5 10 15 20 25 30

No. of obs.

Categ. Histogram: nedvesség

#: 1 nedvesség = 90*0,5*normal(x; 8,8744; 0,6454)

nedvesség

No of obs

#: 1 7,07,58,08,59,09,510,010,511,011,512,012,513,0 #: 2 0

10 20 30 40 50 60

Categ. Histogram: nedvesség

nedvesség

No of obs

#: 1 7,0

7,58,0 8,59,0

9,510,0 10,511,0

11,512,0 12,513,0 0

10 20 30 40 50 60

#: 2 7,0

7,58,0 8,59,0

9,510,0 10,511,0

11,512,0 12,513,0

Mathematical pretreatments

SNV / MSC;

g-s / S-G D1 / D2

29

Mathematical pretreatments

Double- edged sword!

„trial and error”

(next slide)

30

(11)

Mathematical pretreatments

1,5 31 2 2,5 3 3,5 4

850 900 950 1000 1050

hullám hos s z [nm ]

A = log(1/T)

-0,2 0 0,2 0,4 0,6 0,8 1 1,2 1,4

1100 1300 1500 1700 1900 2100 2300 2500 hullám hoss z [nm ]

A = log(1/R)

d < 315 µm 315 µm < d< 710 µm

710 µm < d

transmission (diffuse) reflection

What is

the aim?

Determination of protein content or particle size?

||: Double- edged sword! :||

The reason is physical:

granulation (corn grits).

Mathematical pretreatments

32 The effect on

calibrations for glucan content of corn fiber

„trial and error”

33

c o mmo d it y G × E in s tr u me n t

c ro p y e a r

Filtering of spectra (outliers),

searching for patterns

(12)

34

Drugs – building spectrum library

65 products, 390 batches

10 / 10 vs.

100 / 10

35

65 products, 390 batches

API from China, India

(e.g.

tetracaine).

Drugs – searching for patterns (PCA)

36 liqueurs

vermouths alcohol content

extract content low

low high

high

Alcoholic/hard drinks – searching for

patterns (PQS)

(13)

37

From „off-line” to „in-/on-line”

„[...] These measurements can be:

at-line:

Measurement where the sample is removed, isolated from, and analyzed in close proximity to the process stream.

on-line:

Measurement where the sample is diverted from the manufacturing process, and may be returned to the process stream.

in-line:

Measurement where the sample is not removed from the process stream and can be invasive or noninvasive [...]”

Because NIR is quick

and non- destructive

38

The Process Analytical Technology (PAT) tools

& within it: research fields of our academic group

In order to implement a successful PAT project, a combination of three main PAT tools is essential

Multivariate data acquisition and data analysis (MVDA) tools

Usually advanced software packages which 1) aid in design of experiments (DoE), 2) collection of raw data and 3) statistically analyzing this data in order to determine what parameters are Critical Process Parameters (CPP).

Process analytical chemistry (PAC) tools

In-line and on-line analytical instruments used to measure those parameters that have been defined as CPP. These include mainly near infrared spec- troscopy (NIRS); but also include biosensors, Raman spectroscopy etc.

Continuous improvement and/or knowledge management tools

Software packages which accumulate Quality Control data acquired over time for specific processes with the aim of defining process weaknesses and implementing and monitoring process improvement initiatives.

39

ICH Q10 approach of PAT

& within it: research fields of our academic group

Is it any faster than the light?

No. So the NIR light is good enough for us,

multivariate data analysis (MVDA) [calculating, solving]

near-infrared spectroscopy (NIRS) [off-line testing]

design of experiments

(DoE) [planning, projecting]

Summary:

time is money, so we need PAT.

(14)

40

The tool of in-/on-line – the fiber probe

Yes, it is breakable...

:(

41

How we can collect spectra?

The theory ...

diffuse reflectance (R)

trasmittance (T) transflectance (TR)

attentuated total reflectance (ATR)

42

How we can collect spectra?

... and the practice (with probes)

diffuse reflectance (R)

trasmittance (T) transflectance (TR)

attentuated total reflectance (ATR)

(15)

43

An example: from cuvette to tank

scaling-up from 1 mL to 5000 L

0 2 4 6 8 10

0 3 6 9 12 15 18 21

Time (hour) Gly (g L-1)

0 2 4 6 8 10 12

0 3 6 9 12 15 18 21

Time (hour) Ace (mmol L-1)

Black dots:

reference data by hours

Colour lines:

NIR based predicted data

by 4 mins!!!

44

An example: from cuvette to tank

scaling-up from 1 mL to 5000 L

From „off-line” to „in-/on-line”

off-line / at-line on-line / in-line

the same optics = quick calibration transfer

Is it any moving part?

No:

fix grating.

Automatic change of source = continous operating digital VIS camera, IP65, OPC, modbus ASCII etc.

(16)

46 Bühler AG: NIR Multi Online Analyzer MYRG.

http://www.buhlergroup.com/global/en/products/nir-multi-online-analyzer-myrg.htm#.V34_q_NPqWg

From „off-line” to „in-/on-line”

47

Industrie 4.0

You can find this sign on your bank card...

Budapesti Kereskedelmi és Iparkamara (BKIK): Ipar 4.0 (szakértői tanulmány).

http://bkik.hu/iparitagozat/ipari-tagozat/osztalyok/bkik-vii-hirkozles-informatika-osztaly/vii-osztaly-hirei/ipar-4-0/.

Király Olívia: Ipar 4.0 avagy beléptünk a jövőbe – 5 fogalom, ami segít az eligazodásban.

http://konzervtelefon.blog.hu/2017/07/12/ipar_4_0_avagy_beleptunk_a_jovobe_5_fogalom_ami_segit_az_eligazodasban

48 Lengyel Adrienn: A Győri Gyufagyár (Várostörténeti puzzle – 8. rész)

http://gyoriszalon.hu/index.php?mact=News,cntnt01,detail,0&cntnt01articleid=2375

Industrie 1.0 – Győr / Raab / Arrabona

Match factory, smoking chimney

(17)

49 id. Konok Tamás (1951, képszám: 43348)

http://www.fortepan.hu

Industrie 2.0 – Győr / Raab / Arrabona

Textile factory

50 Bauer Sándor (1975, képszám: 109843)

http://www.fortepan.hu

Industrie 3.0 – Győr / Raab / Arrabona

Vending machine

51 AUDI AG (2015, képazonosító: 146eaa22)

https://images.audi.hu/gallery/tn2/146eaa22.jpg

Industrie 4.0 – Győr / Raab / Arrabona

It is not a vending machine

(18)

52

Industrie 4.0 – here we are

Perten Instruments: DA 7440 On-line NIR.

https://www.perten.com/Products/DA-7440-On-line-NIR/

||: = IoT internet of things :||

We need more and more dedicated solutions

53

Industrie 4.0 – resulting...

Király Olívia: Ipar 4.0 avagy beléptünk a jövőbe – 5 fogalom, ami segít az eligazodásban.

http://konzervtelefon.blog.hu/2017/07/12/ipar_4_0_avagy_beleptunk_a_jovobe_5_fogalom_ami_segit_az_eligazodasban

Macro/micro imaging

54 T (trans- mittance)

(19)

Genuie steroid tablet Fake steroid tablet

Tablets – NIR microscopy

PCA Abs.

55

Package materials – „as is”

Péter GORDON / BME EFI-Labs

56

Package materials – light microscopy

metal layer

57 Péter GORDON / BME EFI-Labs

(20)

58

Package materials – IR microscopy

59

Package materials – identification

PET

60

Package materials – identification

PA

(21)

61

Package materials – identification

PE

Biodegradable polymers (TPS/PLA)

62

Biodegradable polymers (TPS/PLA)

63

(22)

Thank you for your kind attention