The application of collision/reaction cell inductively coupled

plasma mass spectrometry to multi-element analysis in variable

sample matrices, using He as a non-reactive cell gas
Edward McCurdy* and Glenn Woods

Agilent Technologies Ltd., Lakeside, Cheadle Royal Business Park, Stockport, Cheshire,
UK SK8 3GR. E-mail: ed_mccurdy@agilent.com; Fax: 144 161 492 7173;
Tel: 144 161 492 7500

Received 2nd October 2003, Accepted 5th February 2004

First published as an Advance Article on the web 4th March 2004

The development of collision and reaction cells for inductively coupled plasma mass spectrometry (ICP-MS)
has extended the capability of the technique by allowing the selective attenuation or removal of previously
problematic spectral interferences. However, many of the reported applications of collision and reaction cell
ICP-MS have required the adoption of operating conditions that are so specific to the selected analyte and/or
target interference that the multi-element capability of the instrument has been compromised, or the conditions
have been applicable only to a well-defined and consistent sample type. This work demonstrates the application
of an ICP-MS, fitted with a collision/reaction cell, to perform multi-element analysis at the single ng ml21 level
in multiple sample types. A single set of operating conditions was used for all selected analytes across a diverse
range of sample matrices. The capability of an octopole-based collision/reaction cell ICP-MS, operated using
He as the inert collision gas, was demonstrated for the removal of unidentified polyatomic species arising from
variable Cl2, S2 and C-based synthetic matrices and round-robin samples comprising a range of clinical
matrices. Compared to the standard (no cell gas) mode, the He cell gas mode improved the accuracy of spike
recoveries at the 5 ng ml21 level for all the measured isotopes of all the transition metals investigated (Ti, V,
Cr, Mn, Fe, Co, Ni, Cu and Zn) in the synthetic matrix samples, using the same cell gas and voltage conditions
for all analytes in all matrices. Where reference values were available for the clinical samples, analyte recoveries
were typically within the range of the expected concentrations, when calibrated using simple (not matched for
sample matrix) calibration standards and using operating conditions that were constant for all analytes and all
matrices. Additional benefits of using an inert cell gas such as He were the complete absence of newly formed
interfering product ion species and freedom from analyte signal reduction through loss by reaction. This is in
contrast to the reported generation of new, cell-formed polyatomic ions and the loss of some analytes by
reaction, which can occur when highly reactive cell gases are used, and suggests that the use of an inert
collision gas may be suitable for the analysis of complex and variable sample matrices, where the identity of
any potential interfering species is not known in advance.

Introduction analyte (e.g., 40Ar35Cl1 on 75As1). The problematic interfering

polyatomic species may be derived from combinations of ions
Inductively coupled plasma mass spectrometry (ICP-MS) with from the plasma (e.g., 40Ar21 on 80Se1), the sample solution
a quadrupole mass spectrometer (sometimes referred to as ICP- (e.g., 40Ar16O1 on 56Fe1 in an aqueous solution) or the sample
QMS) has become a widely used analytical tool across many matrix (e.g., 40Ar23Na1 on 63Cu1 in a sea-water matrix).
industrial and research applications, with two of its principal While several attempts have been made to compile lists of
benefits being its ease of use and multi-element capability. The potential polyatomic ion interferences in ICP-MS,5,6 it is
widespread acceptance of the analytical strengths of ICP-MS important to recognise that many theoretically possible
has historically been tempered by the acknowledgement that polyatomic species are unlikely to survive in the plasma or
several interferences were difficult to address, resulting in to form during ion extraction, or are present at such low
poorer reported detection limits for the affected analyte intensities in the mass spectrum that they are analytically
elements. Most of the reported interference effects in ICP- irrelevant for most applications. It is also important to note
MS are well documented, having been discussed since the early that, prior to the development of collision/reaction cells in ICP-
papers on the technique, e.g., ref. 1. Such interferences may be MS, several methods were already employed to reduce
physical (e.g., sample viscosity and nebulisation), chemical polyatomic ions in the mass spectrum, including the use of
(e.g., inter-element reactions leading to analyte loss within the mathematical corrections, aerosol desolvation, matrix elimina-
sample introduction system), related to changes in the degree of tion, cool plasma conditions and high-resolution spectro-
ionisation of analytes (e.g., ionisation suppression from a high meters. While each of these methods may be effective in
concentration of easily ionised elements2,3), due to disruption reducing some polyatmic interferences, none is a universal
DOI: 10.1039/b312250f

in ion transmission (e.g., increased space-charge effects due to solution. Mathematical corrections may be complicated to set
the presence of a high concentration of high mass ions in the up, time consuming to maintain and may introduce errors if
extracted ion beam4), or spectral (e.g., polyatomic ion overlap). sample matrix components vary.7 Although aerosol desolva-
The principal sources of spectral interferences in ICP-MS are tion in the guise of spray chamber cooling8 is almost universally
isobaric overlap from a second element with an isotope at the used in commercial ICP-MS instruments, more elaborate
same nominal mass (e.g., 114Sn1 on 114Cd1), or from devices that use ultrasonic aerosol generation or heating and
polyatomic species at the same nominal mass as the target cooling stages to remove water vapour from the aerosol more

This journal is ß The Royal Society of Chemistry 2004 J. Anal. At. Spectrom., 2004, 19, 607–615 607
 efficiently may add excessive cost and complexity and analyte. Reactive cell gases may also provide very high
introduce new problems of poorer analyte washout and loss efficiency of interference signal reduction; up to 9 orders of
of volatile species.9 Cool plasma conditions are widely used in magnitude signal reduction for 40Ar1, from 2E9 counts per
the semiconductor industry, where samples typically have low second (cps) to 2 cps has been reported with H2 cell gas, for
and consistent levels of both matrix components and analyte example.27 Finally, reaction processes can offer a high degree
elements, but these conditions provide less efficient decom- of flexibility, since the wide range of different reaction gases
position of the sample matrix, and a reduced signal for poorly available means that specific conditions may be set up to
ionised analytes, and are therefore less suitable for the analysis address a variety of individual polyatomic overlaps.
of variable or high matrix samples.10 High resolution mass However, the reported problems of reactive gases, especially
spectrometers are more costly and more complicated to operate those that react with a large number of ions, include the loss of
than modern, automated, quadrupole-based ICP-MS systems, analytes by reaction28,29 (which reduces the number of analytes
but high resolution does have the benefit of absolute certainty that can be measured under a single set of conditions) and the
of interference removal, up to a required mass resolution of generation of new, cell-formed polyatomic species30 (which
about 10 000 m/Dm and with the proviso that ion transmission requires that the cell conditions be set up specifically for each
is severely reduced at high resolution settings.11 analyte in each sample so as to exclude these new interferences
One of the simplest approaches to reducing the occurrence of from the mass spectrum). In addition, the inability of a specific
polyatomic species in the mass spectrum of a quadrupole ICP- reaction pathway to address multiple interfering species of
MS is through judicious optimisation of the ICP conditions, different reactivity at the same analyte mass31 renders this
principally the selection of operating conditions that give a high approach unsuitable in many cases for the analysis of
effective temperature in the central channel of the plasma. unknown, complex sample matrices.
Plasma efficiency is typically monitored using the ratio of the The issue of loss of analyte signal and the generation of new,
CeO1 polyatomic ion to the Ce1 parent ion, which measures cell-formed polyatomic species also has implications for the
the ability of the plasma to dcompose this refractory molecular application of ICP-MS to the screening of unknown samples,
ion and can give an indication of the degree of decomposition using semiquantitative analysis. Semiquantitative analysis is an
of some other polyatomic species.12 However, plasma optimi- immensely valuable tool in ICP-MS, as it provides a rapid
sation alone is unlikely to be able to reduce most polyatomic overview of elemental concentrations, reporting almost every
species by much more than 90% or one order of magnitude element and covering a wide concentration range, but based on
(based on reducing the CeO1/Ce1 ratio from 3.0% to 0.3%), the measured response for only a small number of reference
which may be insufficient for many applications. elements. Semiquantitative calibrations in ICP-MS depend on
The development and commercialisation of collision and the fact that the relative sensitivity of the analyte isotopes can
reaction cell technology incorporated into ICP-MS instruments be predicted with a reasonable degree of accuracy, requiring
has allowed analysts to address several of the more commonly only the determination of the general ‘‘mass/response’’
reported and problematic interfering polyatomic species, allow- relationship for the instrument, followed by correction for
ing the adoption of an alternative approach to the analysis the abundance of the measured isotope and the degree of
of many interfered analytes. Following initial work, where ionisation of the element (using the Saha equation). Any
residual water vapour was identified as causing interference processes that disrupt this mass/response relationship as a
reduction through reaction processes, several publications have result of element specific (and matrix dependent) chemical
emphasised the use of hydrogen (pure or as a gas mixture) as a reactions in the collision/reaction cell have the potential to lead
reactive cell gas, as it is effective at removing polyatomic ions to increased errors in the reported semi-quantitative results.
(especially argon-based polyatomic species) but is virtually Further practical problems arise in the selection of operating
unreactive with most analytes,13 so the multi-element capability conditions for interference removal by highly reactive mole-
of the ICP-MS is maintained. In this respect, hydrogen has cular cell gases, due to the difficulty in predicting all the
been studied extensively, e.g., ref. 14, and has been proposed as possible reaction pathways and/or rates and their dependence
an ideal reaction gas for inorganic mass spectrometry.15 on the composition of the sample. While many papers that have
While the maintenance of the multi-element capability of the reported on the utilisation of reactive cell gases refer to the
ICP-MS has been the subject of a few publications on thermodynamics of the cell gas reactions with the interferent
commercial collision/reaction cell ICP-MS instruments, these and analyte, the observed reaction processes and pathways
publications have typically investigated a limited number of sometimes appear to be far from simple. As an example, the
interfered elements, where the interferences are well-defined relative ionisation potentials of interfering and analyte ions are
and relatively easily removed. For example, Iglesias et al.16 commonly quoted as a means of selecting, with a high degree of
assessed the impact of several different reactive cell gases on the certainty, between interfering and analyte ions. In such cases,
major background interferences ArO1 and Ar21 on 56Fe1 and the charge-transfer reaction of the cell gas with the interferent is
80
Ar1, respectively. They established that using a highly often said to be thermodynamically ‘‘allowed’’, while the
reactive cell gas (NH3) significantly degraded the detection reaction of the cell gas with the analyte ion is said to be
limits obtainable for half the previously uninterfered elements, ‘‘disallowed’’. In practice, however, individual potential
while a mixture of H2 and He (as a buffer gas) yielded a reaction paths (such as the charge transfer example), taken
significant improvement in the LOD for Fe and Se, without in isolation, do not provide enough information to ascertain
degrading the analytical performance for the previously non- whether a given reaction gas will provide a suitable interference
interfered analytes. However, in many cases the reduction or removal process. Olesik and Jones28 discussed the case of the
removal of a particular polyatomic species has been based on removal of the 35Cl16O1 interference on 51V1 in a matrix
the use of a highly reactive cell gas in combination with cell containing a moderately high concentration of Cl and noted
multipole voltage settings for the selective removal of a specific that, although the primary charge transfer reaction of V1 with
target interfering ion from a specific interfered analyte mass. In their selected reaction gas, NH3 (V1 1 NH3 A V 1 NH31),
other words, the ICP-MS has frequently been operated as a should not occur, the reaction being endothermic by 3.45 eV,
single element or even single isotope analyser.17–26 loss of V1 from the mass spectrum was in fact observed, at a
The reported benefits of using a reactive gas for interference rate about 10 times faster than the loss of Ca1. The authors
removal in a collision/reaction cell ICP-MS include a high also noted that high intensities of new, cell-formed interfering
degree of specificity, such that a particular cell gas may be species based on clusters containing V and NH3 [V(NHx)n1,
selected to provide conditions that reduce the signal for the where n ~ 1–6] appeared at high NH3 cell gas flow rates,
target interference with minimal effect on the interfered confirming that V1 does indeed react with the NH3 cell gas, but

608 J. Anal. At. Spectrom., 2004, 19, 607–615

via a clustering or condensation reaction, rather than the simple (KED) can be applied to the removal of cell-formed polyatomic
charge transfer reaction, which the thermodynamics suggested species (which have inherently low energy), but was also found
would not occur. This type of discrepancy between the to be suitable for the removal of polyatomic species that exist
theoretical thermodynamics of one particular reaction process before the cell, i.e., those arising from the plasma and/or
and the observed appearance of new interfering ions and interface region.37 The mechanism for the removal of these pre-
disappearance of analyte ions as a result of a different reaction existing polyatomic ions was found to be the selective loss of
process, suggests that the use of highly reactive cell gases energy exhibited by polyatomic ions, compared with mona-
requires a great deal of care in the selection of suitable tomic analyte ions, following collision with an inert cell gas
conditions for each analyte/interference pair and a high degree (He). The interfering species, being polyatomic or molecular
of awareness of the other factors (sample matrix type, ions, have a higher collisional cross section than the analyte
combinations of interferences and even the presence of other ions and so collide more frequently with the He cell gas. The
analyte elements) that may affect the analysis in each individual resulting collisional deceleration causes preferential loss of
sample. energy to near thermal energies for the polyatomic ions, which
The combination of reported potential problems with the use thereby permits their removal from the ion beam, along with
of highly reactive cell gases has led to the conclusion that no the cell-formed species, by the application of a small, fixed bias
single reaction gas or gas mixture provides a reliable method of voltage at the cell exit. By contrast, the monatomic ions
interference removal across a wide range of analytes and maintain several eV of energy at the cell exit and can therefore
sample types. This stimulated our interest in developing a set of overcome the KED step of the bias voltage and so be selectively
collision/reaction cell ICP-MS operating conditions that allowed to exit the cell, passing into the analyser quadrupole
functioned in conjunction with an inert cell gas and interference for separation and detection. This means that the KED method
removal by non-specific, physical processes, rather than using a of interference reduction is not suitable for the removal of
reactive cell gas and interference removal by specific reaction monatomic ion isobaric overlaps such as 40Ar1 on 40Ca1, but
pathways. Yamada et al.37 found that KED after collision with He cell gas
The use of an inert cell gas to promote collisional dissociation was effective in providing 4–5 orders of magnitude reduction in
of potential interfering polyatomic ions was reported in early the intensities of ArH1, ArO1 and Ar21, none of which was
collision cell ICP-MS publications32,33 and the efficiency of this reduced significantly by pressurisation of the cell with He alone
collisional fragmentation was even reported to be suitable for (without KED), although a small reduction in ArO1 intensity
removal of some polyatomic ion overlaps from the mass was observed without the application of KED, probably due to
spectrum.34 It was subsequently suggested35 that the main collisional dissociation of this weakly-bound molecular ion.
contributor to the reported interference removal in this case was The authors noted that this selective removal of polyatomic
a reaction process due to the presence of adventitious water species through KED at the cell exit required the analyte and
vapour as a contaminant in the cell gas, as reported by Eiden polyatomic ions entering the cell to cover a narrow spread of
et al,13 rather than a collision process due to the inert cell gas ion energies, which is achieved due to the use of a grounded
itself. Indeed, Tanner et al.30 have suggested that reaction plasma shield, known as the ShieldTorch, on the ICP-MS
chemistry resulting from gas impurities accounts for most of the system used in that work.
observed interference removal reported from cells operating As a point of interest, it should be noted that the use of a
with inert cell gases. In contrast to other reports, this review grounded shield plate to reduce plasma potential and so
even suggested that gas impurities account for much of the decrease the mean ion energy and ion energy spread has
interference reduction observed when using H2 as a reactive cell another consequence when used in conjunction with a collision/
gas. However, the authors acknowledged that the efficiency of a reaction cell, namely that it reduces the capacity of the cell to
specific cell gas will depend on the experimental conditions used perform collisional focussing as a result of ion energy
and clearly the design of the collision/reaction cells of different reduction. Clearly if the ion energies of ions entering the cell
ICP-MS instruments and the cell gas pressure at which they are are already low and consistent, there is less potential to gain
operated will have a major influence. Consequently, the limited analyte sensitivity through further reduction/focusing of ion
success of a particular cell gas on a particular instrument should energy.
not be taken as a reliable indication of the general efficacy of In the context of this work, however, it is the function of the
that gas on other instruments. non-reactive cell gas to remove non-specific polyatomic
Dexter et al.36 investigated the effect of deliberately adding interfering ions in favour of non-specific monatomic analyte
reactive species to an inert cell gas and found that the presence ions, that led us to apply this mode of interference removal to
of water vapour did indeed induce interference removal by multi-element analysis in variable sample matrices, including
reaction in a cell pressurised with He. In contrast, the presence synthetic, multi-component matrices and clinical round-robin
of water vapour had almost no effect in a cell pressurised with reference materials.
H2, where reactions already dominate the interference removal
processes. In a subsequent paper using a different instrument,
Yamada et al.37 demonstrated that gas purity has some effect Experimental
on the reduction of argon-based polyatomic species that are
Instrumentation
reactive with cell impurities but not with H2 cell gas, but most
of the argon-based polyatomic species react efficiently with H2, An Agilent Technologies 7500c ICP-MS system (Agilent
so the additional effect of reaction with cell gas contaminants is Technologies, Palo Alto, California, USA), which utilises an
limited. These authors also investigated the formation of new octopole ion guide enclosed in a collision/reaction cell, was
interfering species by reaction with cell gas impurities and used throughout. This instrument has been described else-
concluded that some species (including water cluster ions and where38 and was of standard configuration. Operating condi-
some metal oxide interferences) increased when low quality cell tions, which are summarised in Table 1, were normal for
gas was used. However, they noted that the instrument used general, high matrix analysis, i.e., the instrument was optimised
was able to distinguish between low-energy, cell-formed species with consideration for the CeO1/Ce1 ratio, which was about
and the higher energy analyte species from the plasma, with the 0.8% in standard mode (no cell gas).
result that the new interferences could be excluded from the The CeO1/Ce1 ratio was further reduced in He cell gas mode
mass spectrum by the selection of a suitable energy discrimina- (to about 0.4%), due to the preferential attenuation of the
tion (ED) voltage at the cell exit. CeO1 polyatomic ion through energy discrimination at the cell
Energy discrimination or kinetic energy discrimination exit (as reported by others38–40). The sample introduction

J. Anal. At. Spectrom., 2004, 19, 607–615 609

 Table 1 Typical operating conditions used for the Agilent 7500c instrument for multi-element analysis with He cell gas

Sample introduction and plasma

Rf power 1500 W
Plasma gas flow 15 l min21
Auxiliary gas flow 1 l min21
Carrier gas flow 1.25 l min21
Sampling depth 7 mm
Torch injector internal diameter 2.5 mm
Interface Ni (1.0 mm sampler; 0.4 mm skimmer)
Ion lens voltages Optimised for sensitivity in 10 ng ml21 tune solution (Li, Y, Ce, Tl)
Collision/reaction cell parameters
Cell gas flow He, 5.5 ml min21
Octopole bias 217 V
Quadrupole bias 213.5 V

system used was the Agilent low-flow concentric nebuliser,

fitted to a standard quartz double-pass spray chamber,
operated at 12 uC to ensure temperature stability and reduce
the water vapour present in the carrier gas/aerosol stream. The
torch used for all measurements was the standard one-piece
quartz torch, which has an injector internal diameter of 2.5 mm
and is fitted with the Agilent ShieldTorch. The interface cones
were the standard items, made of Ni. The analytes measured
and the isotopes used were dependent on the matrix and
interferences being investigated but, generally, an effort was
made to measure several isotopes, including the most
abundant, for each analyte, even where the most abundant
isotope would not typically be preferred, due to its suscept-
ibility to polyatomic overlap. Examples include the measure-
ment of Cr at mass 52 (which is overlapped by 40Ar12C1 in a
high carbon matrix), Fe at mass 56 (which is overlapped by
40
Ar16O1 in an aqueous matrix) and Cu at mass 63 (which is
overlapped by 40Ar23Na1 in a high sodium matrix.
The cell conditions used for interference removal were based
on the use of He as the cell gas, with only a single, fixed flow
rate being used for all analytes and all matrices. High purity He
(Premier Quality, 99.9992% He, Air Products, Crewe, UK) was
used, in order to minimise the potential problems caused by
unidentified reactive contaminant species in the cell.
Optimisation of the cell gas flow rate (and therefore cell gas
pressure) was not carried out specifically for each sample type
or analyte, but the general optimisation procedure used for the
identification of an appropriate He cell gas flow is illustrated in
Fig. 1.
This illustration shows the effect of increasing He cell gas
flow rate on the raw signal intensities at mass 51 and 52 u in
a blank matrix containing 1% of chloride and carbon, and Fig. 1 He cell gas flow rate optimisation plots showing attenuation of
the same matrix spiked with V and Cr at a concentration of matrix polyatomic ions in blank and spiked matrix containing 1% Cl
40 ng ml21. The plots illustrate the selective attenuation of the and 1% C: (a) ClO1 overlap on V at mass 51 and (b) ClOH1 and ArC1
interfering species (principally ClO1 at mass 51 and ClOH1 overlaps on Cr at mass 52.
and ArC1 at mass 52) present in the blank matrix, relative to
the additional analyte ion signal for 51V1 and 52Cr1 present in
the spiked matrix. The net improvement in background optimise the cell conditions for individual analytes or different
equivalent concentration (BEC) of about 3 orders of magnitude sample matrices. Where widely differing sample matrices were
for both analytes is shown, with optimum BEC levels of about analysed against a single calibration, the major components of
0.02 ng ml21 and 0.06 ng ml21 being obtained for V and Cr, the samples (e.g., acid type and concentration) were approxi-
respectively, in this matrix and these non-cleanroom laboratory mated for the calibration standards, but no major element or
conditions. Note that, in addition to the same cell voltage sample matrix matching was carried out. No blank subtraction
conditions and cell gas type being used for both analytes, was used for any of the results and no interference or inter-
the same He flow rate optimum of 5.5 ml min21 was identified element corrections were used for any of the results reported in
for the 2 analytes, illustrating the simplicity of selecting this work.
appropriate operating conditions for the removal of multiple
interferences on multiple analyte masses, when an inert
Reagents
collision gas is used.
Ion lens voltages were adjusted as required, as part of the Standard solutions were prepared from single element stock
normal system optimisation. No changes to the cell voltage or solutions (Spex Certiprep, Assurance or Claritas PPT grade) at
interference discrimination settings were made. In our experi- either 1000 mg ml21 or 10 mg ml21. Mixed working calibration
ence, the majority of sample types and analytes that suffer from standards were prepared at concentration ranges suitable for
matrix-based polyatomic interferences can be measured under the analytes being investigated, by dilution in 18 MV deionised
a single set of cell conditions, therefore no attempt was made to water with the addition of an appropriate stabiliser. For the

610 J. Anal. At. Spectrom., 2004, 19, 607–615

analysis of samples in acidic matrices, standards and blanks use of He cell gas, the method specified that each analyte
were stabilised with the addition of 2% v/v nitric acid (67–69% isotope was measured in He mode (cell pressurised with
assay, Romil, UpA, ultrapure grade). Sulfuric acid samples for 5.5 ml min21 of He) and then in standard mode (no cell gas),
the scan analysis were prepared from sulfuric acid (98% assay, the data being combined in a single sample result file.
Romil, UpA, ultrapure grade), diluted 1 : 50 in deionised water. For the clinical sample analysis, multi-element calibration
Mixed synthetic sample matrices were prepared from solutions standards were prepared in the clinical sample solubilisation
of sulfuric acid (98% assay, Romil, UpA, ultrapure grade), matrix and the samples (serum, whole blood and urine) were
hydrochloric acid (33–36% assay, Romil, UpA, ultrapure diluted 1 : 10 in the same matrix. After solubilisation for 20 min
grade) and 1-butanol (Romil, SpS, superpure grade), diluted by at room temperature, samples were analysed using the Agilent
volume in deionised water, to give several different matrix I-AS autosampler and an acquisition method that specified the
concentration levels. For the analysis of clinical samples preferred isotope and cell mode for each analyte. The measured
stabilised in a basic sample solubilisation matrix, standards and concentration for each trace element was quantified against the
blanks were prepared in the same basic matrix, which consisted external calibration in the basic solubilisation matrix, using Ge
of 1% ammonium hydroxide solution (20–22% assay, Romil, and Bi as the internal standards. Results were corrected for the
UpA, ultrapure grade), 2% 1-butanol (Romil, SpS, superpure 10 times dilution and compared with the target or certified
grade), 0.05% Triton-X (Romil, SpS, superpure grade) and concentrations. The rinse solution was the same matrix as used
0.05% ethylenediaminetetraacetic acid (Aldrich, 99.999% pure) for sample solubilisation.
prepared by weight/volume in deionised water.

Results and discussion

Procedure
Sulfuric acid analysis
Using the optimum He cell gas flow rate, identified in Fig. 1,
the ability of the consistent He cell gas mode to remove The main sulfur-based polyatomic species in the mass region
multiple polyatomic species in a sulfuric acid matrix was from m/z 50 to m/z 84 are those arising from combinations of S,
investigated, to determine whether acceptable trace element O, Ar and H, particularly 34S16O1, 32S21, 32S16O21,
32 16
backgrounds could be obtained for multiple interfered elements S O21H1, 32S34S1, 34S16O21, 34S16O21H1, 32S16O18O1,
34 1 40
in this single-component matrix. For the interference removal S2 , Ar32S1, 32S16O31 and 32S216O1. These interferences
test in sulfuric acid, two samples were prepared in a matrix of can give rise to simple (single interfering ion) or compound
2% (nominal) sulfuric acid, diluted in deionised water. One (multiple interfering ions at the same nominal mass) inter-
sample was analysed as a matrix blank (sulfuric acid only) ferences at every mass between m/z 64 and m/z 68, which
while the second sample was spiked with 5 ng ml21 of the includes all the major isotopes of Zn and the second largest
elements V, Cr, Mn, Fe, Ni, Co, Cu, Zn, As and Se, to assess (but usually preferred) isotope of Cu, at mass 65. Other major
the efficiency of removal of the potential sulfur-based S-based polyatomic ions can interfere with the measurement of
polyatomic ions and other background polyatomic ions in V at mass 51 (e.g., 34S16O1H1), Cr at mass 52 (e.g., 34S18O1),
the m/z 50 to m/z 84 region of the mass spectrum. Analysis was Ge at mass 72 (40Ar32S1) and Se at mass 80 (32S16O31 and
32 16 1
carried out using the scanning mode of data acquisition, with S2 O ).
masses between 50 and 84 being measured using 20 points per The ability of the 7500c octopole-based cell to remove all
peak and with an integration time of 0.5 s per point (giving a these interferences under a single set of operating conditions
mass spectrum with an intensity scale in counts per 0.5 s). was assessed, through comparison of the mass spectra for
For the synthetic, variable level, multi-component matrix the blank and spiked sulfuric acid samples. It should be
studies, matrix solutions containing several different potential emphasised that the aim of this evaluation was to investigate
sources of polyatomic species (sulfuric acid, hydrochloric acid the capability of the He cell gas mode to remove multiple,
and 1-butanol) were prepared at 4 different concentrations unidentified interferences in samples containing a high and
(0%, 0.1%, 0.5% and 1.0% of each matrix component). complex matrix. Consequently, the laboratory environment,
Additional matrix components were considered (NaCl and sample processing, sample introduction hardware and spike
CaCO3, for example), but were not used in this work because of levels were appropriate to these sample types, rather than the
the difficulties in obtaining such materials of a purity level free much lower concentrations that are typically measured in
from trace contamination for the elements of interest. Since semiconductor grade chemicals. Fig. 2 shows the mass spectra
acids and organic solvents are typically available at very high measured for the sulfuric acid blank both without cell gas
purity, the issue of trace element contamination was not (Fig. 2(a)) and with 5.5 ml min21 He cell gas (Fig. 2(b)),
considered to be a major problem, although several transition illustrating the effective removal of all the sulfur-based
metals are commonly found at trace levels in sulfuric acid, even interferences outlined above, as well as the ArO1 and Ar21
when high purity acid is used. For these multi-component interferences at mass 56 and 80. Fig. 2(c) shows the 5 ng ml21
matrix tests, the analytes of interest were spiked (in a mixed, spiked sulfuric acid sample under the same He cell gas
multi-element spike) into each matrix sample at a concentra- conditions as for the blank spectrum in Fig. 2(b).
tion of 5 ng ml21 of each analyte element. The resultant signal Two things are apparent from these comparison spectra.
for each analyte (spike plus any matrix-derived interfering Firstly, the spectrum for the blank 2% sulfuric acid matrix with
species) in each matrix sample was quantified by comparison He cell gas (Fig. 2(b)) illustrates that the polyatomic inter-
against a multi-element calibration consisting of calibration ferences were attenuated to a degree sufficient to allow
levels at 0, 1, 2, 5 and 10 ng ml21, prepared in 2% nitric acid. Ge quantification of the elements of interest at sub ng ml21
was added to each sample using on-line internal standard levels in the sulfuric acid matrix. Note that the intensity scale
addition, at a concentration in the sample of approximately (counts in 0.5 s) of the blank sulfuric acid spectrum in He mode
20 ng ml21, to compensate for any sample introduction effects was reduced by a factor of 200 (1E4 versus 2E6) compared with
that might arise from the different viscosities of the different the scale of the spectrum of the same blank matrix in standard
sample matrices. The rinse solution used contained 5% nitric mode. Secondly, the mass spectrum for the spiked sample
acid. (Fig. 2(c)) shows that the isotopic patterns for the spiked
The ICP-MS was configured to operate in a multi-mode analytes matched the expected natural isotopic abundances,
acquisition whereby measurements are made under different indicating that all interferences were removed effectively,
instrument settings in a single visit to each sample vial. For the without any appearance of new polyatomic species in this
purpose of comparing the reported results with and without the mass region and without evident loss of signal through reaction

J. Anal. At. Spectrom., 2004, 19, 607–615 611

 (scattering of lighter ions) that exists when the cell is
pressurised, and the fact that the reference isotope selected
for the template fit was the lowest mass isotope for each
element. This effect is also the cause of the decrease in light
element sensitivity under He cell gas conditions, compared with
standard (no cell gas conditions), although the net gain in
reduced BEC levels is significant, as shown in the optimisation
plots in Fig. 1. It should also be noted that, in the blank 2%
H2SO4 in He mode, the measured peaks for Ni, Fe and Cu (the
latter two too low to be observed on this scale) matched the
theoretical isotopic abundances, suggesting that the signal was
due to the presence of traces of these analytes in the unspiked
sample, rather than residual sulfur-based interferences. The
relatively high signal for Ni in the blank H2SO4 (w 10 6 higher
than any other trace element peak) may have been due to trace
contamination or the use of the standard Ni interface cones,
rather than the Pt-tipped cones that are commonly used when
high concentrations of sulfuric acid are analysed or when low
concentrations of trace contaminant elements are quantified.
The demonstration of effective removal of sulfur-based
interferences, to allow a range of transition metals to be
measured directly at their principal masses, contrasts with the
method of using a reactive cell gas (NH3) to promote the
formation of remote cluster ions for indirect measurement.
Since Zn, in common with many other analyte ions, reacts with
NH3 to form M(NHx)n1 cluster ions,28,30 one reported method
for the analysis of Zn in a sulfuric acid matrix is the indirect
measurement of the cell-formed cluster ion Zn(NH3)31, derived
from 64Zn, at mass 115.41 This approach has the merit of
removing the analyte ion to a remote part of the mass
spectrum, where the original interference is not present (based
on the thermodynamics of the formation of cluster ions derived
from the analyte and the original, sulfur-based interfering ion).
However, the analysis of cluster ions is clearly unsuitable for
measurements in samples which contain a high or variable
matrix, where the remote cluster ion mass may already be
occupied by another analyte or matrix ion, or another cluster
ion derived from a different sample component. For example,
even in the relatively simple case of the determination of Zn in
high-purity sulfuric acid, measuring the Zn(NH3)31 cluster ion
will give erroneous results if the sample contains In (which
also has an isotope at mass 115), Cu or Ti (which might give
rise to other cluster ions, such as CuNH2(NH3)21 and
TiNH2(NH3)31, at the same mass). Use of an inert cell gas,
together with efficient removal of new (cell-formed) and
existing (plasma-formed) interfering polyatomic ions by
KED, avoids the twin problems of analyte loss by reaction
and formation of new cluster ions.

Synthetic sample matrices

Having successfully established that a single set of cell
conditions could be used effectively to attenuate both the
Ar-based background species (particularly ArO1 and Ar21)
and the various sulfur-based polyatomic species that might
overlap with the analytes in the mass region 50–84 u, the same
multi-element conditions, again using He as an inert cell gas,
were applied to the analysis of multiple analyte elements in a
variable, multi-component sample matrix containing several
Fig. 2 Measured mass spectra for the mass range from m/z 50 to m/z 84 in
a 2% H2SO4 solution: (a) blank matrix with standard (no cell gas) matrix elements that are commonly found in environmental,
conditions; (b) same blank matrix with He cell gas (5.5 ml min21 He flow); clinical and food sample matrices. The most prominent of the
and (c) blank matrix plus a 5 ppb spike containing V, Cr, Mn, Fe, Ni, Co, range of possible matrix-based interferences derived from the
Cu, Zn, As and Se, also with 5.5 ml min21 He cell gas flow. Note, the plasma gases and solvent (Ar, O, H and N) and the sample
intensity scale is 2E6 counts in 0.5 s for (a) and 2006 lower at 1E4 counts in matrix (S, Cl and C) are shown in Table 2.
0.5 s in (b) and (c) (inset scale 500 counts). Theoretical isotopic templates are
The spiked matrix samples at the various matrix levels (0%,
shown for the spiked elements in (c) and for Ni and Br backgrounds in (b)
0.1%, 0.5% and 1.0% of each matrix component) were analysed
as unknown samples, with the spike concentrations being
of the analyte elements. There was a small discrepancy between determined against a nitric acid calibration. The results for
the theoretical and measured abundances for the higher mass each of the analytes in the variable synthetic sample matrices
isotopes of each element, due to the increased mass bias are presented in Figs. 3 and 4 as comparison plots showing the

612 J. Anal. At. Spectrom., 2004, 19, 607–615

Table 2 Principal theoretical polyatomic ion interferences derived
from the mixed matrix standard used for the variable synthetic matrix
test

Isotope Principal interfering species (S, Cl, C matrix)

47 32 14
Ti S NH, 12C35Cl
51 35
V Cl16O, 37Cl14N
52 36
Cr Ar16O, 40Ar12C, 35Cl16OH, 37
Cl14NH
53 36
Cr Ar16OH, 40Ar13C, 37Cl16O, 35
Cl18O, 40Ar12CH
55 37 18
Mn Cl O
56 40
Fe Ar16O
63 12 16 35
Cu C O Cl, 12C14N37Cl
64 32 16
Zn S O2, 32S2, 36Ar12C16O
65 32 16
Cu S O2H, 32S2H
66 34 16
Zn S O2, 32S34S, 33S2
72 40
Ge Ar32S, 35Cl37Cl

5 ng ml21 spike recoveries measured in each of the the two cell

modes. In each case, the measured concentration in ng ml21
(ppb) is shown for each analyte of interest, both without cell
gas (Std mode) and with cell gas (He mode). Consistent
instrument conditions were used for each of the two different
cell gas modes, for all analyte masses and all sample matrices,
throughout the experiment. From the comparison of the
measured concentration results for each of the analytes, it is
clear that the He cell gas mode gave consistent recovery of the
5 ng ml21 spike for all analytes in the different sample matrices.
This indicates that the He cell gas mode was able to reduce the
matrix-based interferences on all the target analytes, without
requiring any adjustment of the instrument operating condi-
tions to minimise specific, identified interfering species or,
indeed, even the prior identification of the interfering species
being removed. This clearly has implications for the operation
of the collision/reaction cell ICP-MS instrument for applica-
tions where the sample matrix (and therefore the potential
interfering species) are either not known, variable or both.
In contrast with the consistent results obtained using He as
the cell gas, the concentrations measured in the standard mode
(Std-no cell gas) showed significant increases in the reported
concentrations (or less accurate spike recovery) of many of
the analytes, due to the contribution of the matrix-based
polyatomic species at the analyte mass. This additional,
erroneous contribution to the reported analyte concentration
was, as would be expected, in proportion to the increasing
concentration of the matrix components. Some of the
interference levels observed in standard mode may have been
significantly higher had the plasma not been optimised for low
levels of polyatomic ion formation, using the low CeO1/Ce1
ratio conditions (around 0.5%) that are typical of the
instrument used.12
As mentioned in connection with the scan data for the
sulfuric acid matrix, Ni cones were used throughout these
experiments and it is likely that the increase in reported Ni
concentration with both modes of operation (Std and He cell
gas) was the result of increased sample cone erosion with the
increasing sulfuric acid concentration.
It is interesting to note that, in the case of 55Mn, 59Co and
63
Cu, which did not suffer from significant polyatomic overlap
in these matrices, the reported results in standard mode
actually decreased slightly with the increasing matrix level. The
He cell gas mode results did not exhibit the same pattern,
suggesting that the decrease was not due to any problems of
sample transport or nebulisation, or errors in internal standard
compatibility. More likely, the lower measured results in the
standard mode analysis of the higher matrix samples were due Fig. 3 Measured concentrations (ng ml21) for 5 ng ml21 spikes
for trace elements in variable matrices, using He cell gas (open
to the inability of the standard mode of operation to resolve the circles) and Std (no cell gas—closed squares) modes. Matrix level
matrix based polyatomic ions 40Ar32S1 and 35Cl37Cl1 on the refers to the concentration of each matrix component—H2SO4, HCl
internal standard, Ge, measured at mass 72. The He mode of and 1-butanol.
operation was able to remove these interferences effectively,
leading to better accuracy of internal standard correction in the

J. Anal. At. Spectrom., 2004, 19, 607–615 613

 changing matrix. The potential errors associated with the
presence of incompletely removed or newly formed interfer-
ences on the isotopes of internal standards is a factor that is
often overlooked, but should be considered carefully when
selecting suitable conditions for the analysis of unknown
sample matrices.

Clinical reference samples

The clinical samples supplied were provided as part of a round-
robin exercise and were analysed using the Agilent 7500c
operating in a multi-mode acquisition. The analytes presented
in Table 3 were all measured using a single, consistent He cell
gas mode to illustrate the capability of this mode of operation
to provide accurate results for a range of elements, which may
suffer from different individual or compound interferences, in a
range of different sample types. Although the clinical sample
matrices were all prepared by 10-fold dilution in the same basic
solubilisation matrix, the major element compositions of
serum, whole blood and urine are significantly different, so
the matrix-derived interfering species would also be expected to
vary. As with the multiple synthetic matrix test, no attempt was
made to optimise the cell conditions for specific analyte/
interference pairs or for specific sample matrices; i.e., the
clinical samples were approached as unknown samples.
The results presented in Table 3 indicate that accurate
analysis was possible for the analytes measured in the range of
clinical sample matrices analysed. All results for the trace
elements were within the expected range of concentrations and
close to the target values, with the exception of the Cd results,
which were marginally lower than the expected range. While
relatively few elements had target or reference values, all the
elements with known concentrations were analysed in all of
the samples, using a single multi-element calibration. This
illustrates the efficient removal, using a single set of non-
specific cell operating conditions, of several matrix-derived
polyatomic interferences, which might have affected the
different analyte elements in one or more of the clincal

Table 3 Comparison of target and range concentrations for clinical

reference samples and concentrations measured using ICP-MS with
octopole collision/reaction cell (all values in ng ml21, corrected for the
106 dilution factor)

Specimen Target valuea Rangea 7500c resulta

Serum-1 Quebec E-01-11

Cu(63) 320 270–370 288
Zn(66) 640 560–720 645
Serum-2 Quebec E-01-13
Cu(63) 1000 890–1110 920
Zn(66) 1140 1030–1260 1224
Blood-1 SAS IQC-1
Cd(111) 2.4 2.0–2.8 2.2
Pb(208) 126 110–140 128
Blood-2 SAS IQC-2
Cd(111) 5.5 5.1–6.0 4.9
Pb(208) 320 300–340 313
Blood-3 SAS IQC-3
Cd(111) 11.8 11.1–12.4 11.0
Pb(208) 680 650–710 701
Urine-1 Quebec S-02-05
As(75) 135 107–164 141
Urine-2 Quebec S-02-02
As(75) 24 4–44 26
Urine-3 Quebec B-01-08
Cr(52) 1.04 0.31–1.77 1.17
Fig. 4 Measured concentrations (ng ml21) for 5 ng ml21 spikes for Urine-3 Quebec B-01-11
trace elements in variable matrices, using He cell gas (open circles) and Cr(52) 46 43–50 46
Std (no cell gas—closed squares) modes. Matrix level refers to the a
Target and range data courtesy of the Interlaboratory Comparison
concentration of each matrix component—H2SO4, HCl and 1-butanol.
Program, The Centre for Toxicology, Quebec National Institute of
Public Health, Canada, and the Regional Laboratory for Toxico-
logy, City Hospital, Birmingham, UK. 7500c result is the mean of 2
sequences.

614 J. Anal. At. Spectrom., 2004, 19, 607–615

sample matrices, including ArNa1, MoO1, ClO1, Cl21, 7 J. L. M. De Boer, J. Anal. At. Spectrom., 2000, 15, 1157.
ClOH1, ArC1, ArSH1 and ArCl1. 8 R. C. Hutton and A. N. Eaton, J. Anal. At. Spectrom., 1987, 2,
595.
9 J. C. Ely, C. R. Neal, J. A. O’Neill, Jr. and J. C. Jain, Chem. Geol.,
Conclusions 1999, 157, 219–234.
10 K. Sakata and K. Kawabata, Spectrochim. Acta, 1994, 49B, 1027–
For the variable matrices evaluated and the range of analytes 1038.
measured, an ICP-MS instrument, fitted with an octopole 11 P. Marriott, R. Fletcher, A. Cole, I. Beaumont, J. Lofthouse,
collision/reaction cell pressurised with He gas, gave effective S. Bloomfield and P. Miller, J. Anal. At. Spectrom., 1998, 13, 1021.
12 E. J. McCurdy, S. Wilbur, G. D. Woods and D. Potter, in Plasma
removal of the various polyatomic ion overlaps derived from Source Mass Spectrometry: The New Millennium, eds. G. Holland
the matrix components. No optimisation of the cell conditions and S. D. Tanner, The Royal Society of Chemistry, Cambridge,
was required for individual analyte elements (or isotopes) or for 2001, p. 45.
individual sample matrices. The use of an inert cell gas (He) 13 G. C. Eiden, C. J. Barinaga and D. W. Koppenaal, J. Anal. At.
was found to eliminate any problems of analyte loss by reaction Spectrom., 1996, 11, 317.
and no newly formed interfering species were observed in any 14 C. P. Ingle, P. K. Appleblad, H. J. Reid and B. L. Sharp, J. Anal.
At. Spectrom., 2001, 16, 1076.
of the sample types analysed. Consequently, no analyte-specific
15 G. C. Eiden, C. J. Barinaga and D. W. Koppenaal, Rapid
cell conditions were required to reject newly formed polyatomic Commun. Mass Spectrom., 1997, 11, 37.
ion interferences and therefore no sample matrix screening was 16 M. Iglesias, N. Gilon, E. Poussel and J.-M. Mermet, J. Anal. At.
required to identify such new potentially interfering species. Spectrom., 2002, 17, 1240.
Operation of the cell in a single He cell gas mode allowed the 17 G. K. Koyanagi, V. I. Baranov, S. D. Tanner and D. K. Bohme,
analysis of all isotopes of the normally interfered analytes Cu J. Anal. At. Spectrom., 2000, 15, 1207.
and Zn in sulfuric acid, directly on the analyte mass, so 18 P. R. D. Mason, K. Kaspers and M. J. van Bergen, J. Anal. At.
Spectrom., 1999, 14, 1067.
avoiding the inherent unreliability of analysis based on the 19 J. M. Marchante Gayon, I. Feldmann, C. Thomas and
measurement of cell-formed cluster ions formed from sequen- N. Jakubowski, J. Anal. At. Spectrom., 2000, 16, 457.
tial reaction with a highly reactive cell gas. 20 Y.-L. Chang and S.-J. Jiang, J. Anal. At. Spectrom., 2001, 16,
While the use of He cell gas and energy discrimination gave 1434.
effective attenuation of all the interfering species under a single 21 T. Makazato, H. Tao, T. Taniguchi and K. Isshiki, Talanta, 2002,
set of operating conditions, it should be noted that several 58, 121.
22 E. H. Larsen, J. Sloth, M. Hansen and S. Moesgaard, J. Anal. At.
other factors must be taken into account when analysing highly
Spectrom., 2003, 18, 310.
variable sample matrices against a simple calibration standard. 23 H.-T. Liu and S.-J. Jiang, Anal. Bioanal. Chem., 2003, 375, 306.
Firstly, the changing matrix composition of the samples may 24 L. Hinojosa Reyes, J. M. Marchante Gayon, J. I. Garcia Alonso
have a significant effect on the sample transport efficiency, such and A. Sanz-Medel, J. Anal. At. Spectrom., 2003, 18, 11.
that the analyte signal may increase or decrease due to changes 25 D. R. Bandura, S. D. Tanner, V. I. Baranov, G. K. Koyanagi,
in sample viscosity, aerosol generation effects (e.g., due to V. V. Lavrov and D. K. Bohme, in Plasma Source Mass
sample surface tension), droplet evaporation rates (e.g., Spectrometry: The New Millennium, eds. G. Holland and
S. D. Tanner, The Royal Society of Chemistry, Cambridge,
resulting from different acid concentrations) and plasma 2001, p. 130.
ionisation effects (e.g., from high levels of easily ionised 26 C. C. Chery, K. DeCremer, R. Cornelis, F. Vanhaecke and
elements or ionisation enhancers such as carbon). These effects L. Moens, J. Anal. At. Spectrom., 2003, 18, 1113.
cannot be compensated for by the polyatomic ion interference 27 E. J. McCurdy and D. Potter, in Atomic Spectroscopy in Elemental
removal processes in the collision/reaction cell, but they may be Analysis, ed. M. Cullen, Blackwell Publishing, Sheffield, 2003,
alleviated through the use of internal standards and, where ch. 2, p. 23.
28 J. W. Olesik and D. Jones, in Plasma Source Mass Spectrometry:
possible, some effort to match the chemistry (such as the acid
Applications and Emerging Technologies, eds. J. G. Holland and
type and approximate concentration) of the standards to the S. D. Tanner, The Royal Society of Chemistry, Cambridge, 2003,
known chemistry of the samples. p. 261.
Nevertheless, the fact that polyatomic ion levels could be 29 B. Hattendorf and D. Gunther, J. Anal. At. Spectrom., 2000, 15,
reduced effectively using this instrument with He cell gas, with 1125.
operating conditions that are independent of both the analyte 30 S. D. Tanner, V. I. Baranov and D. R. Bandura, Spectrochim.
being measured and the sample matrix, means that this mode Acta, Part B, 2002, 57, 1361.
31 H. Louie, M. Wu, P. Di, P. Snitch and G. Chapple, J. Anal. At.
can be applied to multi-element analysis in a range of sample Spectrom., 2002, 17, 587.
types, where the sample matrix is unknown or variable. The use 32 D. J. Douglas, Can. J. Spectrosc., 1989, 34, 38.
of an inert cell gas eliminates the extensive method develop- 33 J. T. Rowan and R. S. Houk, Appl. Spectrosc., 1989, 46, 976–980.
ment that is required to identify and address the existing 34 P. Turner, T. Merren, J. Speakman and C. Haines, in Plasma
(matrix-derived) and new (cell-formed) interfering species, Source Mass Spectrometry: Developments and Applications, eds.
which are characteristic of the sequential chemistry that occurs G. Holland and S. D. Tanner, The Royal Society of Chemistry,
Cambridge, 1997, p. 28.
when a highly reactive cell gas is used. 35 S. D. Tanner and V. I. Baranov, in Plasma Source Mass
Spectrometry: New Developments and Applications, eds.
References G. Holland and S. D. Tanner, The Royal Society of Chemistry,
Cambridge, 1999, p. 46.
1 A. L. Gray, Fresenius’ Z. Anal. Chem., 1986, 324, 561–570. 36 M. A. Dexter, P. K. Appleblad, C. P. Ingle, J. H. Batey, H. J. Reid
2 J. A. Olivares and R. S. Houk, Anal. Chem., 1986, 58, 20–25. and B. L. Sharp, J. Anal. At. Spectrom., 2002, 17, 183.
3 A. L. Gray, in Inorganic Mass Spectrometry, eds. F. Adams, 37 N. Yamada, J. Takahashi and K. Sakata, J. Anal. At. Spectrom.,
R. Gijbels and R. Van Grieken, John Wiley and Sons, New York, 2002, 17, 1213.
1988, ch. 6, p. 257. 38 P. Leonhard, R. Pepelnik, A. Prange, N. Yamada and T. Yamada,
4 D. J. Douglas and S. D. Tanner, in Inductively Coupled Plasma J. Anal. At. Spectrom., 2002, 17, 189.
Mass Spectrometry, ed. A. Montaser, Wiley-VCH, New York, 39 M. A. Dexter, H. J. Reid and B. L. Sharp, J. Anal. At. Spectrom.,
1998, ch. 8, p. 615. 2002, 17, 676.
5 A. L. Gray, Spectrochim. Acta, 1985, 40B, 1525–1537. 40 Z. Du and R. S. Houk, J. Anal. At. Spectrom., 2000, 15, 383.
6 N. M. Reed, R. O. Cairns, R. C. Hutton and Y. Takaku, J. Anal. 41 K. Kawabata, Y. Kishi and R. Thomas, Spectroscopy, 2003, 18(1),
At. Spectrom., 1994, 9, 881. 16.

J. Anal. At. Spectrom., 2004, 19, 607–615 615