The high energy X-ray spectra also contained several lines that
were of no practical use. Beside the pile-up lines of iron there were
also L lines of Ta induced in the exit window that reached the
detector thorough scattering in the air. For garnets that were
mounted in jewelry, the scattered protons also induced gold L
and silver K lines. A weak presence of these lines was also observed
for the stones of object 12/2009-4 (Fig. 1) that were removed from
the object due to restoration purposes and measured as such. Fortunately
none of these lines coincided with the K lines of yttrium,
which was actually the only reliable element deduced from the
hard X-ray spectra (Fig. 2). The X-ray intensities obtained from
the two spectra were normalized according to the iron Ka line, taking
into account the calculated transmission of air and aluminum
absorbers. The concentrations of all detected elements were calculated
by an iterative computer code that calculates the matrix elements
for X-ray and gamma ray production simultaneously in each
iteration step [12]. The sum of concentrations of all metal oxides
obtained was normalized to unity.
As a brief check of the procedure we used the argon line
induced in air and detected in the soft X-ray spectra. The sum of
all metal oxides was also calculated according to this line, using
the argon concentration in air as an internal standard. Large deviations
of the oxide sum from unity (more than 20%) indicated that the experimental geometry departed largely from our supposed
flat-target scheme. We then repeated the calculation varying the
X-ray emission angle in steps of 5 until the oxide sum departed
from unity by less than 20%; when this condition was met the
oxide concentrations were normalized to 100%. The adjusted emission
angles were in the range of 40–55 (65 in one case) and were
required for 21% of measurements. However, with this procedure
we only simulated the realistic X-ray escape lengths in the target.
We thus estimate that the accuracy of major elements in garnets
(including Y) was about 5%, while the uncertainty for the elements
present below a 1% level may worsen to 10–15%.
The high energy X-ray spectra also contained several lines that
were of no practical use. Beside the pile-up lines of iron there were
also L lines of Ta induced in the exit window that reached the
detector thorough scattering in the air. For garnets that were
mounted in jewelry, the scattered protons also induced gold L
and silver K lines. A weak presence of these lines was also observed
for the stones of object 12/2009-4 (Fig. 1) that were removed from
the object due to restoration purposes and measured as such. Fortunately
none of these lines coincided with the K lines of yttrium,
which was actually the only reliable element deduced from the
hard X-ray spectra (Fig. 2). The X-ray intensities obtained from
the two spectra were normalized according to the iron Ka line, taking
into account the calculated transmission of air and aluminum
absorbers. The concentrations of all detected elements were calculated
by an iterative computer code that calculates the matrix elements
for X-ray and gamma ray production simultaneously in each
iteration step [12]. The sum of concentrations of all metal oxides
obtained was normalized to unity.
As a brief check of the procedure we used the argon line
induced in air and detected in the soft X-ray spectra. The sum of
all metal oxides was also calculated according to this line, using
the argon concentration in air as an internal standard. Large deviations
of the oxide sum from unity (more than 20%) indicated that the experimental geometry departed largely from our supposed
flat-target scheme. We then repeated the calculation varying the
X-ray emission angle in steps of 5 until the oxide sum departed
from unity by less than 20%; when this condition was met the
oxide concentrations were normalized to 100%. The adjusted emission
angles were in the range of 40–55 (65 in one case) and were
required for 21% of measurements. However, with this procedure
we only simulated the realistic X-ray escape lengths in the target.
We thus estimate that the accuracy of major elements in garnets
(including Y) was about 5%, while the uncertainty for the elements
present below a 1% level may worsen to 10–15%.
การแปล กรุณารอสักครู่..
