APPLICATION OF

OES-AD SOFTWARE PACKAGE FOR

DEVELOPING AUTOMATED TECHNIQUES

E.V. Shabanova1, I.E. Vasilyeva1, and I.L. Vasiliev2

1 Vinogradov Institute of Geochemistry SB RAS, Favorsky Str., 1a, 664033, Irkutsk, Russia e-mail: shev@igc.irk.ru

2 Institute of System Dynamics and Control Theory SB RAS, Lermontov Str., 134, 664033, Irkutsk, Russia

Spectral equipmentSpectral equipment

Multichannel analyzer

Diffraction spectrograph

Arcjet

Vertical direct-current

source

OE spectrum of sample with simple compositionOE spectrum of sample with simple composition(CRM of soil SDPS-1)(CRM of soil SDPS-1)

KNOWLEDGE BASE

Hierarchical tables of

wavelengths of analytical

lines of elements

Search and identification of analytical

lines

Choices of taking into account of

background and AP

calculations

Chemometric selections of

optimal conditions of AP processing and

constructing calibrations

Type of n-dimensional calibration:a) with /without taking into account of spectral interference; б)  with /without taking into account of nonspectral influences; в)  with /without blank

Forming of training and

test sets

Processing sample spectra

Selection of optimal APs

and calibrations

Comparison of result quality of

learn and test sets

It isn't different

It's d

iffere

nt

Desired condition compliance

No

RETRAINING

Results of determining content

of i-th element

DA

TA

BA

SE

Yes

Requirements to analytical

results (qualitative,

semiquantitative, quantitative)

Table of element

contents of CRMs, theirs

stuff is available in Lab

Measured spectra of CRMs and samples

Scheme of program complex for OEA executingScheme of program complex for OEA executing

OE-Techniques developed by the softwareOE-Techniques developed by the software

OES-AD software package is used to obtain and process spectra when determining: – impurities in quartz and crystalline silicon and

silanes; – Au, Ag and noble metals in nonsoluble

carbonaceous substance from black shales;

– cadmium in crystals of BaF2;

– lanthanum and yttrium in barium fluoride crystals;– boron and tin in granitoids.

Technical approachesTechnical approaches

The primary studies such as:– selection of analytical lines (determination of upper

and low bounds of intensity measuring);

– selection of analytical parameter for each spectral lines (calculation method of intensity);

– selection type of n-dimensional calibration for every analyte or group of analytes.

Table of Ba, La and Y contents in BaFTable of Ba, La and Y contents in BaF22 crystals crystals

RCM Ba content, % La content, % Y content, %

Mixture 1 78.3 - -

Mixture 2 70.5 7.09 -

Mixture 3 62.66 14.18 -

Mixture 4 39.16 35.45 -

Mixture 5 23.5 49.63 -

Mixture 6 15.67 56.72 -

Mixture 7 7.83 63.81 -

Mixture 8 - 70.9 -

La 1 77.23 1 -

La 2 77.97 0.33 -

La 3 78.21 0.11 -

La 4 78.3 0.012 -

Y 1 65.48 - 10

Y 2 71.9 - 5

Y 3 75.12 - 2.5

Y 4 77.04 - 1

Y 5 77.69 - 0.5

Hierarchical tablesHierarchical tables of some elements and analytical linesof some elements and analytical lines

Spectral adjectives of analyte and major element linesSpectral adjectives of analyte and major element lines

Compound / Element

Melting and vaporization temperature

s, С

Wavelength, nm

Detection limit,

%

Line intensity by

Harrison

Excitation

potential, eV

Analytical parameter

BaF2 / Ba 1368; 2250 233.526 II263.478 II264.138 259.6678 I

0.0021–

10

60R305

40

6.017.43

–4.75

Calculation of

intensity as

squares

under the

curve,

the

background

under peak

took into

account

LaF3 / La 1493; 2300 333.749 II324.513 II324.935 II261.034 I289.307 II

0.00010.0010.0010.010.01

800400300106

4.123.993.995.666.94

YF3 / Y 1155; 2230 297.4588 I317.307 II291.905 I276.01 I

0.0010.010.010.03

12201810

4.177.434.244.57

OE spectrum of BaFOE spectrum of BaF22

Some lines of Ba and LaSome lines of Ba and La

Way of calculating their analytical parametersWay of calculating their analytical parameters

Calibration typesCalibration types

Calibration typeCalibration type DescriptionDescription

OLS Ordinary Least Squares (one-dimensional)

to each analytical line with following averaging of results for every element

GLS Ordinary (General) Least Squares (n-dimensional)

to all selected analytical lines

PCR Principal Component Regression

to all selected analytical lines

Information coefficient Q=85%

NW Neural network

to all selected analytical lines

Rank criterion from range of relative bias Rank criterion from range of relative bias

using the Rj-values calculated, the range of relative bias is computed as

All ranges obtained for each calibration were ranked by increase. If some ranges of relative

bias vary by 5 %, then the average of numbers of places, on which they get, is calculated.

Thus, the best calibration has the least rank.

%,100.

..

cert

certcalj C

CCR

The relative bias of analyte determination in the each j-th sample of training set is provided by

where Ccal. and Ccert. are calculated and certificated contents, accordingly.

For the entire set of CRMs within the interval minmax CCC

.minmax RRR

Calibrations used for Ba Calibrations used for Ba and La determination and La determination

in BaFin BaF22 crystals crystals

 Calibration

Element line composed of calibration& Its wavelength, nm

NW PCR GLS OLS

Ba II 233.527 + + + +Ba II 263.478 + + + +Ba II 264.138 + + + +Ba I 259.666 + + + +La II 333.749 + + + +La II 324.513 + + + +La II 324.935 + + + +La II 261.034 + + + +La II 289.307 + + + +

Calibrations used Calibrations used for Ba and Y determination for Ba and Y determination

in BaFin BaF22 crystals crystals

 Calibration

Element line composed of calibration& Its wavelength, nm

NW PCR GLS OLS

Ba II 233.527 + + + +

Ba II 263.478 + + + +

Ba II 264.138 + + + +

Ba I 259.666 + + + +

Y I 297.459 + + + +

Y II 317.306 + + + +

Y I 291.905 + + + +

Y I 276.01 + + + +

Calibration selecting for Ba and La determination Calibration selecting for Ba and La determination by rank criterionby rank criterion

CalibrationΔ C, % Δ R Rank of Δ R

Sum of ranksBa La Ba La Ba La

OLS-1 OLS-1 to each analyte line

7.8-78.3

0.01-70.932 27 3 3 6

NW NW to all lines of Ba and La

1717 2424 11 22 33

PCR PCR to all lines of Ba and La

0.01-63.8253 58 4 4 8

GLS GLS to all lines of Ba and La

2323 1010 22 11 33

Calibration selecting for Ba and Y determination Calibration selecting for Ba and Y determination by rank criterionby rank criterion

CalibrationΔ C, % Δ R Rank of Δ R

Sum of ranksBa Y Ba Y Ba Y

OLS-1 OLS-1 to each analyte line

65.5-77.7 0.5-10

3 29 3.5 4 8

NW NW to all lines of Ba and Y

0.5 9 1.5 2.5 4

PCR PCR to all lines of Ba and Y

5 12 3.5 2.5 6

GLS GLS to all lines of Ba and Y

0.40.4 11 1.51.5 11 33

Relative bias Relative bias for determining Ba and Y in training setfor determining Ba and Y in training set

Y5 Y4 Y3 Y2 Y1

-4

-2

0

2

4

6

GLS NW

Re

lativ

e b

ias

of

Ba

, %

Train sample

0

10

20

30

40

50

60

70

80

C (Ba), % C (Y), %

Con

tent

, % m

ass.

Y5 Y4 Y3 Y2 Y1

-4

-2

0

2

4

6

GLS NW

Re

lativ

e b

ias

of

Y, %

Train sample

0

10

20

30

40

50

60

70

80

C (Y), % C (Ba), %

Con

tent

, % m

ass.

Relative bias Relative bias for determining Ba and La in training setfor determining Ba and La in training set

La4 La3 La2 La1 M1 M2 M3 M4 M5 M6 M7-30

-20

-10

0

10

20

30 GLS NW

Re

lativ

e b

ias

of

Ba

, %

Train sample

0

10

20

30

40

50

60

70

80

90

100 C (Ba), % C (La), %

Con

tent

, % m

ass.

La4 La3 La2 La1 M1 M2 M3 M4 M5 M6 M7 M8-30

-20

-10

0

10

20

30 GLS NW

Re

lativ

e b

ias

of

La

, %

Train sample

0

10

20

30

40

50

60

70

80

90

100 C (La), % C (Ba), %

Con

tent

, % m

ass.

Technique of determination of La, Y and Ba in BaFTechnique of determination of La, Y and Ba in BaF22

• The spectral equipment: diffraction spectrograph DFS-

458, arcjet "Vesuvius", and multichannel analyzer MAES

• 5 mg of sample and 5 mg of buffer (graphite powder) is

completely evaporated with the direct current is 15 A

• La content area is 0.01-63 %

while Ba content changes 8-78 %

• Y content area is 0.5-10 %

while Ba content changes 66-78 %

• Reproducibility of La, Y and Ba measurements, sr , are

0.08-0.12; 0.10-0.15;and 0.05-0.07, accordingly