Bortoleto G. G., Verbi F. M., de Morais L., Onaga H., Ferreira M. M. C., Bueno M. I. M. S., "X- Ray Fluorescence and Multivariated Calibration for Determination of Metals". Córdoba, Argentina, 25-29/10/2004: IX Seminario Latinoamericano de Análisis por Técnicas de Rayos X (SARX 2004) [IX Latin American Seminary of Analysis by X-ray Techniques]. Avances en Análisis por Técnicas de Rayos X [Advances in Analysis by X-ray Techniques], 13 (2004) FRX27. Poster FRX27.
X-RAY
FLUORESCENCE AND MULTIVARIATED CALIBRATION FOR
DETERMINATION
OF METALS
Gisele G. Bortoleto*,
Fabíola M. Verbi, Laís S. de Morais, Helena Onaga, Márcia
M. C. Ferreira e Maria
Izabel
M. S. Bueno
DQA – IQ – UNICAMP
– C.P.: 6154 – CEP: 13083 970 Campinas, SP - Brazil * giseleb@iqm.unicamp.br
ABSTRACT
X-ray
fluorescence spectrometry is considered a
well established analytical
method. Nevertheless
research is still undergone to improve accuracy since the major
drawback in
this technique is the interelemental interferences.
For example, it is
uncommon to
stablish a linear correlation between X-ray characteristic
fluorescence
lines and the
elemental concentrations in a natural sample. Other elements contributes
to the
signal as well. For many years,
many mathematical methods are being
developed to
solve this problem. However this “drawback” can be very well explored
when
chemometric tools are used. In this work, Partial Least Squares (PLS)
method
was employed
to acquire a calibration model for Cr, Ni, Zn,
Fe, Cu, Co and Pb
determination.
A set standard water samples with known concentrations
of these
elements was
used.
The metals in aqueous solutions
with concentratios around mg L-1
were
measured in
an EDXRF equipment (Spectrace 5000) and the calibration correlations
were obtained
using The Unscrambler 7.6 program.
After many preprocessing and variable selections
an interesting result was
obtained
for Pb which suffered a strong effect
of the other elements. The best
calibration
model obtained for this metal did not need the use of Pb lines (La
and Lb),
but instead
only the line of the others elements.
This result shows the great
contribution
the multivariated calibration can provide.
In the work showed here,
chemometric
was able to consider all interelemental effects, with more reliable
results
than univariate
analysis.
___________________________________________________________________________________________________
FRX27