10th International Conference on Particle Induced X-ray Emission and its Analytical Applications PIXE 2004, Portorož, Slovenia, June 4-8, 2004 http://pixe2004.ijs.si/
Multivariate Statistics and Mean Atomic Number Classification of Dental Composites as Analyzed by PIXE E.A. Preoteasa*, Rodica Georgescu, C. Ciortea, Daniela Fluerasu, Horia Hulubei National Institute for Physics and Nuclear Engineering, P.O. Box MG-6 Magurele, RO-76900 Bucharest, Romania (*e-mail:
[email protected])
Elena Preoteasa Helident Dental Surgery Ltd., 10 M. Eminescu St., RO-2150 Campina, Romania ABSTRACT Dental composites have been approached by PIXE only recently. Using standardless 3 MeV proton thick target PIXE, 14 composites were analyzed. Relative concentrations of 14 elements were employed for classifying these very diverse biomaterials by multivariate statistics clustering in 2D and 3D projections and by mean Z. Keywords: PIXE, trace elements, biomaterials, dental composites, classification, multivariate statistics.
1.
INTRODUCTION
Dental composites brought substantial innovation in restorative dentistry and develop at a high rate [1], but they have been studied by PIXE only recently [2-4]. During dental use, they are submitted to chemical and mechanical degradation, and they convey in the organism foreign elements that may cause adverse effects. They contain trace elements [2-4], that may be valuable fingerprints for identification, quality control, and understanding restorations-dental tissues interface transfer; such processes may play a role in the formation of secondary caries around dental fillings. One dominant feature of dental composites is represented by a considerable variety, both in their dentistry-relevant attributes [1] and in their multielemental composition, as evidenced by PIXE [2-4]. To date, 22 different dental composites have been already studied by PIXE, including 14 biomaterials by our group [2-3] and 8 by other investigators [4]. Accordingly, we here meet a situation which is confronted with the issue of selecting a lower number of appropriate biomaterials for addressing the above dental research questions. A similar matter may emerge if one would plan to study dental composites by methods associated to more sophisticated experimental facilities, such as a synchrotron or a proton microprobe. There, user beamtime is rather rigorously restricted, and only a limited number of well-selected specimens could be examined. In both cases, we are challenged with the problem of classification of dental composites as a prerequisite for a suitable selection. Such a classification must take into account simultaneously the multielemental compositions of the biomaterials, as both our team [2,3] and the other one [4] positively detected by PIXE at least 18 major, minor and trace elements in the analyzed composites. This can be done for instance by Principal Component Analysis (PCA), a multivariate statistics technique [5] which computes linear combinations of the concentrations or of their logarithms (‘generalized parameters’ or ‘main coordinates’), to evidence relationships between the materials by projections in 2D and 3D subspaces, or by estimation of the mean atomic number Z values for PIXE-detected elements. We previously [3] estimated relative concentrations of the latter and used them to calculate the mean Z and to cluster the composites by PCA in 2D projections, and we present here further progress by 3D projections.
2.
MATERIALS AND METHODS
Flat disk-shaped samples were prepared by (photo)polymerization from 10 types of Western composites, some with two color shades, making a total of 14 biomaterials, labeled with Roman numbers as 943.1
943.2
previously described [2]. These thick targets were covered with a thin carbon foil. PIXE was performed with 3 MeV protons from a tandem van de Graaff accelerator using a hyper pure Ge detector. For the dental composites, appropriate reference materials are generally lacking and, for our present purpose, corrections for matrix effects [6] were attempted only to a very limited extent. A unique X-ray yield curve was calculated for a hypothetical light element thick target matrix and this was used to evaluate preliminary relative concentrations for all biomaterials included in the study. Estimation of detection limits was necessary to use them in the PCA calculation; they were established imposing that the minimum detectable number of pulses in a peak, Np, must satisfy the relation Np > 3 Nb1/2, where Nb is the number of pulses in the background. The classification procedures of biomaterials included: A) a PCA computation of linear combinations of the relative concentrations of 17 selected elements (given in Tab. 1), to evidence relationships between the biomaterials in 2D and 3D subspaces; B) the same as before, but using logarithms of concentrations; and C) estimation of mean Z values for PIXE-detected elements (‹Z›PIXE). The light elements not detected by PIXE were not considered in the evaluation of ‹Z›PIXE; this parameter, although roughly proportional to the true physical one, should be considered only a relative value defined for classification purposes. As PCA required definite concentration values for all 17 elements considered in each of the 14 analyzed specimens – even if certain elements were not detected in a given material – the values of the corresponding detection limits were used instead of concentrations for such missing elements.
3.
RESULTS AND DISCUSSION
The PIXE spectra of the 14 dental composites evidenced altogether up to 24 elements: 18 of them – Si, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, Sr, Y, Zr, Ba, Yb, Hf – were positively identified as major, minor and trace elements, while other 6 (Ag, Nd, Ho, Au, Hg, Pb) were possible traces which would require further confirmation due to the very weak intensity of their peaks in the spectra. It is obvious that only the first 18 elements may be relevant for the classification of the analyzed composites, but among them Si appears to be less reliable, as with our experimental setup we could not detect this element in TABLE 1. PIXE estimated relative concentrations of elements in 14 dental composites1. Elem. Cl K Ca Ti V Cr Mn Fe Ni Cu Zn Sr Zr Cd Ba Yb Hf
I.a 3.1 0.40 0.18 2.4
I.b 3.0 0.23 1.2 1.5
0.15 0.14 0.82 0.42 1.9 0.49
0.16 0.074 1.2 0.54 1.9 0.18
‹Z›PIXE 22,7 22,6
II
III
IV.a
IV.b
V
VI.a
VI.b
VII
0.33 0.092 0.022 0.050 0.45 0.23 0.084 8.5 0.22
5.6 1.5 0.009 0.22 0.13 2.0 0.17 0.15 0.17
0.62
0.45
8.6
0.076 0.026 0.029
0.071 0.024 0.033
0.008
0.024 0.013
0.020 0.008
28,2
22,1
0.22
0.03 0.25
0.003 0.005
3.0 6.2
2.9 6.4
0.08 1.3
61,2
62,8
26,9
9.3
VIII.a
VIII.b
IX
X
0.023 0.019 0.006 0.024
0.021 0.019 0.030
0.024 0.019 0.021
0.030 0.032
9.3
0.021
0.20
0.034 0.028 0.064 0.075 0.062 0.029
0.045 0.018 0.005
0.40 0.19
9.7
9.7
9.7
9.9
38,3
55,4
55,4
54,6
54,9
9.3
0.56
0.55
41,1
41,5
1
The relative concentrations are given in atomic %, and normated so that the sum over the PIXE-detected elements equals 10 % (a/a). In fact the light elements not detected by PIXE represent the largest part of dental composites; for instance, a complete analysis of material IV.a performed by combining PIXE and ERD gives only 14 % (w/w) for the heavier elements detected by PIXE [3]. The uncertainties of concentrations due to counting errors were estimated, but they are not shown due to space limitations. Proceedings of the 10th International Conference on Particle Induced X-ray Emission and its Analytical Applications , Portorož, Slovenia, June 4-8, 2004
943.3
some materials which, according to the producers, contained quarz particles. Thus all our classification attempts were based on the relative concentrations of the remaining 17 elements, in Table 1. he classification procedures led to several partitions of the 14 dental composites in 5-7 groups. The latter
A
B VIIIa+VIIIb+IX+X
2
III 3
V
V
1 2
Ib
VIa+VIb
Ia
II
-1
PC5
PC3
III
IVa+IVb
0
VIa+VIb 1
IVa+IVb
II
VII
-2
0
Ib
-3
VIIIa+VIIIb+IX+X
-1 -4 -2 0 2 4 6
-3 -4
-2
-1
0
1
2
3
Ia
4 -2 0 2 4 6
-3
-2
-1
0
1
2
3
4
-4
FIGURE 1. Three-dimensional plots showing projections of all PCA data obtained from the relative concentrations of 14 PIXE-analyzed elements (Cl, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, Sr, Zr, Cd, Ba, Yb, Hf) in 14 dental composites. The subspace PC1 x PC2 x PC3 (A) accounts for 70.6 % of the data variation, and the subspace PC! X PC2 x PC5 (B) for 66.8 %. The components are determined by Ti, Ni, Zn, Cr, Cl, K, Fe, ... (PC1); by V, Hf, Zr, Yb, Sr, ... (PC2); by Cd, Sr, Ba, Yb, ... (PC3); and Ca, Fe, Ba, K, ... (PC5).
largely overlapped, but some of their features were procedure-specific. Thus procedure (B) – in accordance with correlations between various elements ratios – succeeded to discriminate some cases where (A) and (C) failed. Fig. 1 illustrates 3D projections obtained by PCA with relative concentrations. They suggest the following 5-group classification: (Ia – Ib – II – VIa – VIb), (III), (V), (VII), (IVa – IVb – VIIIa – VIIIb – IX – X); however, note that Fig. 1 does not account for Cu and Mn (determining the PC4 axis, not shown). Using 2D projections computed with concentration logarithms as inputs, we obtained before [3] a slightly different, 7-group partition – more compatible with the mean Z – namely: (Ia – Ib), (II – III), (IVa – IVb), (V), (IVa – IVb), (VII), (VIIIa – VIIIb – IX – X).
4.
CONCLUSION
The classification is relevant given the remarkably diverse nature of dental composites and may be useful in simplifying the problem of choice when planning more complex experiments. Thus our approach may bring a contribution in making PIXE get the most out of this promising new applicative field, and in seeking answers to some questions related to the long-term use of composites in dentistry mentioned in the Introduction.
REFERENCES 1. 2. 3. 4. 5. 6.
Restorative Dental Materials, 10th ed., edited by R.G. Craig, St. Louis, Mosby Year Book, Ltd., 1997. Preoteasa, E.A., Ciortea, C., et al., Nucl. Instr. Meth. Phys. Res. B 189, 426-430 (2002). Preoteasa, E.A., Georgescu, R., et al., Anal. Bioanal. Chem., in press (2003). Vecchi R., Valli G., et al., These Proceedings, contributed paper number 941 (2004). Anderson, T.W., An Introduction to Multivariate Statistical Analysis, New York, Wiley, 1984. Campbell, J.L., Cookson J.A., Nucl. Instr. Meth. Phys. Res. B3, 185-197 (1984).
Proceedings of the 10th International Conference on Particle Induced X-ray Emission and its Analytical Applications , Portorož, Slovenia, June 4-8, 2004