JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1993 VOL. 8 25 Preliminary Assessment of Laser Ablation Inductively Coupled Plasma Mass Spectrometry for Quantitative Multi-element Determination in Silicates John G. Williams and Kym E. Jarvis ICP-MS Facility Department of Geology Royal Holloway University of London Egham Surrey UK TW20 OEX A Nd:YAG (yttrium aluminium garnet) laser was used to ablate pressed powder pellets of seven silicate rock reference materials for sample introduction into an inductively coupled plasma mass spectrometer. Laser operating parameters such as mode (fixed-Q or Q-switch) energy and number of shots per site were optimized to meet the criterion of maximum analyte signal without excessive loading of the plasma with ablated material. To compensate for differential laser sampling (ie.variable amounts of material being removed during each analysis) 55Mn was used as an internal standard for multi-element determinations and l3'Ba for the rare earth elements (REE) Hf Ta and W. Relative responses for the major elements indicate that the chemistry and mineralogy of individual rock samples influence the ablation behaviour and that samples with very similar chemical and mineralogical compositions exhibit similar elemental sensitivities. Alkali alkaline earth and hydride-forming elements also show similar behaviour to the major components. Multi-element detection limits were typically less than a few hundred ng g-l. The accuracy of major element determinations for the materials studied was generally better than +5% relative with a precision of 10% relative standard deviation (RSD).Trace elements in Groups I II Ill REE volatile elements or those which exhibit refractory characteristic displayed good accuracy with precision of generally <lo% RSD. The quantitative determination of major and trace elements in silicate rocks is therefore possible providing that standards and samples are closely matched both in terms of bulk chemistry and physical (mineralogical) composition. Keywords Laser ablation; inductively coupled plasma mass spectrometry; quantitative analysis; silicate rocks; reference materials The initial development of inductively coupled plasma mass spectrometry (ICP-MS) was carried out almost entirely using liquid samples.' Although there were several considerations the main reason was that the existing simple technology for liquid sample introduction into a flame or plasma for atomic absorption or emission spectrometry could be transferred to ICP-MS with little modification.Introducing solutions into an ICP is very convenient with simple switching between samples and straightforward calibration with synthetic solutions. In addition pro- duction of a solution from initially solid materials can eliminate sample inhomogeneity problems and usually reduces matrix effects. However during the development of the technique it was soon realized that the rapid data collection capability of ICP-MS could be used to good effect in the direct analysis of solids using laser ablation (LA) for sample introduction.' In addition the advantages that ICP- MS possesses over other atomic analytical techniques such as simple spectra high sensitivity low background and sub- ng ml-' limits of detection could in principle make LA- ICP-MS a very potent technique.Lasers in Atomic Spectrometry The potential use of lasers as excitation devices for spectrometric analysis was recognized soon after the first report of laser action in ruby in 1960.2 The subsequent development of different forms of laser microanalysis have been reviewed by Moenke-Blankenburg3 and the advan- tages of laser vaporization for micro-sampling in a variety of materials using analytical atomic spectroscopy have been discussed by Dittrich and Wennri~h.~ The role of lasers in MS falls into two distinctive categories. In the first the laser is used both to ablate and ionize the sample while.in the second laser energy is used only to introduce material into an ionizing device. Plasmas generated by laser radiation contain a high concentration of ions offering the possibility of spatially resolved mass spectrometric investigations. The first instru- ments combining a high-energy laser and a mass spectro- meter were commercialized in 1977. In laser microprobe mass spectrometry sample vaporization and ionization is a single-step process brought about by a pulse of laser radiation focused onto the surface of a sample. The ions produced are subsequently separated using a time of flight (TOF) mass spectrometer. This technique with a spatial resolution of several micrometers has been used in surface and bulk analysis of organic and inorganic material^.^ Two-step analytical processes where the sampling and ionization are separate lead to an increase in the ion yield.One example of a two-step analytical process is resonant ionization mass spectrometry. In geological applications sample vaporization is generally achieved using a conven- tional thermal ionization mass spectrometry source but (photo-)ionization is subsequently carried out using a laser which is tuned to a frequency relevant to the selected analyte elements. Mass analysis can be carried out using a quadrupole TOF or sector mass spectrometer. In the work described below a highly practical two-step analytical process is used where a pulsed laser source is used to ablate locally the sample material.The ablated sample is transported to an atmospheric pressure ICP where ions of all elements present are generated with subsequent detection by MS. This approach has the advantage that each step in the process can be optimized separately and therefore more readily controlled. There is considerable interest in utilizing LA with plasma spectrometry for the direct analysis of solid samples. Attractive features of the technique include the capability for in situ micro-analysis and the ability to analyse both conducting and non-conducting materials. Considerable scope exists for application of the technique and to date most studies have been directed to metallurgical industrial or geological samples using either ICP atomic emission spectrometry (AES),697*8 direct current plasma AES9 or ICP- GrayIS reported the first investigations on LA-ICP-MS with a system that was analogous to that of earlier work reported on LA-ICP-AES.' In the former work was carried out using a 1 J ruby laser (JK Type 2000) operating at a MS.10-1426 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1993 VOL.8 wavelength of 694 nm with a maximum repetition rate of 1 shot per second (1 Hz) and in either fixed-Q or Q-switched mode (see under Laser Operation Mode). The applicability of the technique to the direct analysis of solid geological materials was demonstrated both for trace element deter- minations and isotope ratio measurements. Later work by Arrowsmith16 showed that a pulsed Nd:YAG laser operat- ing at a wavelength of 1064 nm with high repetition rates (10-20 Hz) and a maximum energy of 160 mJ offered improved precision and duty cycle.This type of laser is currently proving to be the most popular for LA-ICP-MS although other designs such as continuous wave Nd:YAG13 lasers are being studied. Recent developments in the application of high resolution LA systems has also been reported by Pearce et a1.l' The majority of published work on LA-ICP-MS has concentrated on the diagnostics of the system e.g. the generation of calibration elucidation of relative response factors19 and parameter optimization.20 Despite the potential of LA for solid sample analysis of geological materials to date virtually no fully quantitative data have been reported with the notable exception of the work by Imailo and a recent study by Perkins et aL2' which showed that for the analysis of carbonate samples Ba Mg Mn and Sr could be determined with an accuracy of better than & 1 O% when the samples were prepared as fused glass discs. Laser Operation Mode For most commercially available systems the laser may be operated in either the fixed-Q (or normal) and Q-switched mode.In fixed-Q mode laser radiation emerges when the threshold conditions for laser operation are reached follow- ing a pulse from the flashtube. Pulse widths are typically 100 to 1000 ps. In Q-switched mode the excited atoms are held in the optical medium using a light seal. Upon release of this seal either one single pulse or a few giant pulses of maximum power are obtained. In this mode pulses of 10-1000 ns are obtained.The nature of the interaction between the laser beam and the sample surface is highly dependent on the mode of laser operation used. In general crater depth is greater in fixed-Q mode. In Q-switched mode the crater is shallow but the width is greater than that obtained in fixed-Q mode. Further discussion of these phenomena are given in ref. 3. Experimental Sample Preparation The preparation of geological samples for LA-ICP-MS may be carried out in a number of different ways. Simple cut slabs have been used1° and particularly small samples such as single crystals may be mounted in resin blocks with the surface ground down to provide a flat substrate for analysis. In general the surface does not need to be level for more than about a 200 pm distance across the area to be analysed providing that the range of the laser focusing is sufficient.I6 In this work reference materials (RMs) were prepared as pressed powder pellets.Preparation of pressed pellets Certified RMs were prepared as provided by the supplier. The maximum grain size in these materials is expected to be between 60 and 70 pm with an average of about 40 pm. Powders were oven dried at 105 "C for 24 h prior to preparation. Sub-samples of 2-3 g depending on the density of the material were weighed into disposable plastic weighing boats. To each was added 250-300 p1 of 1 Oh m/v of a poly(viny1 alcohol) binder (Mowiol 8-88 Hoechst) the exact volume being dependent on the nature of the sample. Some samples are particularly absorbent while others form strong coherent pellets with only 200 pl of liquid binder.In this study a liquid binder is preferred since it is relatively easy to mix of high purity with respect to elemental contamination and does not result in a dilution of the sample. Solid poly(viny1 chloride) binders have been used for geological sample preparation15 but typically about 20% by volume of binder is added resulting in a significant dilution factor and problems of inhomogeneity may arise. The powder and binder were thoroughly mixed (2-3 min) using a spatula transferred into a 1 cm die and pressed to 10 tonnes. The resulting pellets were oven dried at 105 "C for 12 h and subsequently stored in a drying oven to prevent water absorption. Care was taken to prevent contamination of either side of the pellet such that both surfaces could be used for analysis.Reference Materials The major element composition and colour of the seven RMs used in this work are shown in Table 1. Of these well characterized materials three are basic rocks of similar major element composition [US Geological Survey (USGS) BIR-1 W-2 DNC-I] and of similar initial grain size. Also included are two evolved rocks (USGS AGV-1 G-2) of similar chemical compositions but with very different 'initial' grain sizes (i.e. the original geological crystallinity of the sample before it was ground) and two sediments [USGS SCo-1 and National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 2704 Buffalo River Sediment]. The influence of particle size mineralogy and pellet colour on the ablation process are not well understood.For example dark coloured materials usually couple well to the laser radiation resulting in an efficient ablation. By contrast white surfaces may reflect a majority of the laser radiation resulting in little Table 1 Major element composition (O/o m/m) of the seven RMs used in this study; reference values from Govindaraju2* AGV- 1 G-2 BIR-1 DNC-1 w-2 sco- 1 NIST SRM 2704 Compound Andesite Granite Basalt Diabase Diabase Sediment Sediment 58.79 17.14 6.76 1.53 4.94 1.05 4.26 2.9 1 0.092 0.49 69.08 15.38 2.66 0.75 1.96 0.48 4.08 4.48 0.032 0.14 47.77 15.35 11.26 9.68 13.24 0.96 1.75 0.027 0.171 0.046 47.04 18.30 9.93 10.05 11.27 0.48 1.87 0.229 0.149 0.085 52.44 15.35 10.74 6.37 10.87 1.06 2.14 0.627 0.163 0.131 62.78 13.67 5.14 2.72 2.62 0.63 0.90 2.77 0.053 0.206 62.2 1 11.54 5.88 1.99 3.64 0.76 0.74 2.4 1 0.072 0.229 Colour Light grey Light grey Grey Grey Grey Taupe Dark brownJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1993 VOL.8 1500 27 ( a ) - Table 2 ICP-MS operating conditions for LA-ICP-MS Instrument VG Elemental Forward power/kW 1.5 Plasma gas Argon Outer gas flow rate/] min-' Intermediate gas flow rate/l min-' PlasmaQuad PQ2 + Reflected power/W <5 Carrier gas flow rate/] min-' 14 1 .o 0.5 Scan conditions Dwell time/ps Channels Sweeps Scan width multi-element Skipped regions multi-element Scan width REE Hf Ta W Skipped regions REE Hf Ta W 80 4096 115 mlz 6-240 m/z 11-22 32-41 and 79.5-80.5 m/z 88-1 84 m/z 90- 130 material being ablated. Thus materials chosen reflect this range of properties.System Optimization ICP-MS For accurate analysis by ICP-MS correct system optimiza- tion is essential such that the conditions used give maximum analyte sensitivity whilst minimizing potential interferences from polyatomic and refractory oxide ions.23 During analysis by LA no water vapour is introduced into the ICP and the optimum conditions are therefore some- what different to those used for conventional solution analysis. The conditions shown in Table 2 were determined experimentally and give maximum sensitivity whilst min- imizing interference effects. Tuning of the ion lenses was initially carried out with the mass spectrometer set to rest at 12C with the carrier gas entering the plasma via the ablation chamber. No ablations were carried out at this stage. The measured signal from I2C (resulting from air entrained in the plasma) was relatively high (equivalent to x 1 pg g-l of a mono-isotopic fully ionized element such as Co) and of a steady state.In practice these operating conditions are very close to those which are appropriate for any analyte element. Fine tuning of the lenses was subsequently carried out at 59C0 and 140Ce during continuous ablation of a sample containing several hundred pg g-l of these elements. Laser Optimization of the LA system itself is more complex and a number of parameters must be established before quantita- tive measurements can be made. These include the mode of laser operation amount of energy delivered to the laser and hence output the rate at which shots are fired number of shots per site pattern of ablation sites and the number of replicate analyses.These variables have been rigorously assessed for the pressed powder pellets. The LA system used in this work has no device for the direct detection of energy in each shot. Therefore the laser output energy can only be inferred from the voltage input to the laser flashlamps. The effect of laser mode (fixed-Q or Q-switched) is critical to the sensitivity and precision of analyses and should be evaluated in each matrix under consideration. In addition the amount of laser energy required to ablate is highly dependent on the physical nature of the sample. The effect of increasing laser energy in each of the two modes of operation is shown in Fig. 1 for RM W-2.The relative behaviour of Mn and Pb is rather different. In the Q- 7 1000 i500 5 700 750 800 850 900 950 1000 0 0 0 700 750 800 850 900 950 1000 Laser flashlamp voltageN Fig. 1 Relative behaviour of (a) Mn and (b) Pb in RM W-2 using A fixed-Q (free running) and B Q-switched mode of laser operat ion switched mode Mn shows a clear increase in sensitivity up to 850 V while beyond this point an increase in energy results in a gradual loss of sensitivity. By contrast Pb shows a steady increase in sensitivity from low to high power. Using fixed-Q mode however the behaviour of these elements is very similar. This feature coupled with the generation of neat steep-sided ablation craters (a feature that would be advantageous for discrete profiling of a sample) has led to the preferred choice of fixed-Q mode in this work. In practice the choice of laser energy is limited by the range of elemental concentrations required.Throughout this work 800 V were used which gave good trace element sensitivity whilst ensuring that major element peaks remained within the linear measurement range. Repeated shots were fired at a rate of one every 0.05 s (20 Hz). To determine the optimum number of shots required at each site depth profiles were constructed for Mn and Pb in W-2 RMs. A single shot was fired at the sample surface and the maximum sensitivity recorded. This process was re- peated ten times on the same spot. By the fifth shot sensitivity was reduced in all cases by about 50% and therefore little advantage is gained by firing more than this number of shots at a single site.The reason for this rapid decrease in signal is not entirely clear and may be a result of the laser becoming de-focused with depth or simply that ablated particulate material is unable to escape from the pit as increasing depths are reached. The total number of sites that are required for a single analysis is dependent on several factors. Firstly if a bulk analysis is required then sufficient surface and substrate area should be sampled to ensure that a representative proportion of material is analysed. Using the fixed-Q mode under the conditions listed in Table 3 a pit is produced which is approximately equal in diameter and depth typically 50- I00 pm. The sensitivity obtained is also partly Table 3 Laser operating conditions for quantitative elemental determination in silicate pressed powder pellets Laser type Wavelengthhm 1064 Laser mode Laser input energy/V 800 Shot repetition rate/Hz 20 Shots per site 5 Sampling scheme Nd:YAG (Spectron Laser Systems) fixed Q (free running) maximum output 600 mJ 6 x 4 raster i.e. 24 sites per analysis28 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1993 VOL.8 determined by the amount of sample carried into the ICP during the time period of the analysis. However exceeding the load that the plasma is able to vaporize and ionize leads to blocking of the sampling cone and plasma torch injector in a matter of minutes resulting in a rapid loss of analyte signal. There is therefore no advantage in ablating excessive amounts of material. Secondly if only a few sites are taken ( < 5 ) precision is generally poorer than when larger num- bers are sampled.The physical arrangement of the sampling sites may be constrained by the size and shape of the pellet. For normal mode analysis it is important that each site is completely separated from its neighbour in order to ensure good repeatability. The data given here were obtained using a sampling scheme of 5 shots per site with a raster pattern of 6 x 4 sites. The total mass of material ablated during a 30 s analysis is approximately 35 pg assuming 100% transport efficiency. During an equivalent analysis using conven- tional solution nebulization the amount of sample intro- duced into the ICP was about 20 pg. The sampling of relatively small masses of material during LA may be an important limiting factor for the accurate determination of certain trace elements.Sample heterogeneity is overcome in solution analysis by sub-sampling of >O. 1 g of material. In the present application however the relatively small mass ( ~ 2 0 pg) analysed during a 30 s data acquisition may not always be representative of the bulk sample. These factors should therefore be borne in mind when assessing analytical accuracy. Internal Correction Internal standards are widely used in many analytical techniques where signal variability occurs on both a short or long timescale. A single element is usually chosen whose behaviour matches that of the analytes of interest. The reasons for the signal fluctuation may be many fold but are often due to instability of electrical components or in the case of ICP spectrometry plasma noise.For conventional solution nebulization ICP-MS analysis internal standardi- zation may result in improved precision compared with uncorrected data although other forms of data manipula- tion such as drift monitoring may further improve data quality if the fluctuation in signal response is progressive with time.24 For sample introduction by LA repeated analysis of the same sample may show significant differ- ences in sensitivity from one analysis to another. Indeed the sensitivity from one sample to another will also vary and this is thought to result from the removal of different masses of material during ablation. To compensate for this differential sampling it is necessary to employ an internal standard.For multi-element determinations SSMn was chosen. The concentration of this element from sample to sample varies over a relatively narrow range it displays good sensitivity but is not sufficiently high in concentration to cause peak saturation. Manganese is in addition well characterized in the RMs studied. Narrow range scan determinations of rare earth elements (REE) were made using 137Ba as internal standard. Barium is well character- ized in the RMs studied is relatively abundant and lies close in mass to the REE. Results and Discussion Multi-element Data Major elements When laser light is absorbed by solids a variety of heating phenomena occur. These include surface heating vaporiza- tion dissociation and excitation of the surface materials and a phase change inside the sample.The interaction of laser radiation with the sample is a complex process and may be affected by a number of criteria. A part of the laser light is not only absorbed at the sample surface but also to a depth of some micrometers. The temperature is raised above the boiling-point of an individual element com- pound or mineral in the case of geological samples and evaporation begins. This process is followed by a change in state of the material at the ablation site which results in a high speed eruption and the formation of a crater in the surface of the sample. The volume heated depends on the thermal conductivity of the material.4 During the ablation of a rock sample in this case of silicate composition it is the major mineral components that will most strongly influence the ablation behaviour.Although in the RMs considered here Si02 is the major component its content varies from 47.04 (DNC-1) to 69.08% (G-2). There is a corresponding difference in the composition of the other major elements (Table 1). These RMs may be sub-divided depending on their physicochemical nature. Basic igneous rocks W-2 BIR-1 and DNC-1 are characterized by relatively low silica content and fine initial grain size SCo- 1 and NIST SRM 2704 are sediments with a similar major element compositon to that of AGV-1 but with a different geological origin and AGV- 1 and G-2 are intermediate/acid igneous rocks with silica contents up to about 70% m/m. The relative responses for some major elements in six RMs are shown in Table 4.Although there are a range of sensitivities within the basic rocks between one element and another (e.g. A1 displays about 25% of the sensitivity of Na) all three rocks exhibit similar responses for any individual element. The two sediments display a similar pattern and for some elements (e.g. Al) relative sensitivity is uniform throughout both the sediment and basic rock groups. The sediments can however be clearly separated into a distinctive group based on their over-all relative Table 4 Relative responses* for some major elements in six standard silicate RMs by LA-ICP-MS Basic rocks Sediments Acid rock Element W-2 BIR-1 DNC-1 G-2 NIST 2704 SCO-1 Si A1 Fe Mg Ca Na K Ti P ndt nd 10 11 35 36 30 29 20 20 40 44 nd nd 35 29 1.7 1 .o nd 9 37 30 15 39 nd 36 1.6 nd 4 30 25 6 10 nd 26 0.5 nd 10 33 20 18 12 nd 21 1.8 nd 10 42 19 13 18 nd 27 1.6 *Results given in counts g pg-I corrected to s5Mn as an internal standard and for isotopic abundance. tnd = Not determined.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1993 VOL.8 29 1600 rn + .- 5 1200 2 2 & 800 w .- a - a3 C 0 2 400 a 0 1 2 3 4 5 6 co nce n t ra t i o n/l o4 pg g-' Fig. 2 Offset in relative sensitivity for Mg in A the basic rocks compared with B the sedimentary rocks and G-2 elemental responses. The highest silica sample G-2. usually displays very low sensitivity particularly when compared with the basic rocks. The distinction in relative sensitivity is illustrated in Fig. 2 for Mg where it can be seen that the basic rocks lie on a significantly different slope to that of the sediment RMs.Differences in relative sensitivity may be expected from samples of differing chemical composition (e.g. (3-2 and W-2). However significant differences are also observed between samples of similar chemical composition but with different geological origins and histories (broadly expressed by degree of crystallinity initial grain size and mineralogy). Sub-division of geological rock types is made taking into account not only similiarities in chemistry but more importantly in mineralogy. The observed differences in relative elemental behaviour is likely to be closely con- trolled by the physical properties of the individual minerals in which the elements occur rather than by the bulk rock chemistry.Hence although broadly similar in chemical composition G-2 and SCo- 1 are different in mineralogical composition. Simple chemical matching of standards and samples is clearly not sufficient if quantitative data are required. Trace elements In the RMs analysed some trace (< 1000 pg g-') elements are contained within silicate mineral phases often substi- tuted for the major elements. However many trace ele- ments are located in minor non-silicate phases where they may in fact be present at the mass per cent. level. Evaluating the relative behaviour of these elements has proved complex and only general statements may be made concern- ing the prediction of elemental behaviour in these matrices. The relative responses for a number of trace (and major) elements are shown in Table 5 with various sub-divisions highlighting different chemical properties and groupings in the Periodic Table.In a number of cases concentration data are not available for all of the RMs and in individual cases elements are below the detection limit. The alkali metals display a wide range of sensitivity both within a single sample and between sample type groups with G-2 displaying particularly poor sensitivity. The alkaline earths by contrast display rather similar sensitivi- ties for each element in all sample types with the notable exception of G-2. It should therefore be possible to make quantitative measurements certainly within a rock group and in some instances between groups. It is relevant to note that extensive Sr and Ba substitution for Ca occurs in certain silicate minerals e.g.feldspar and this fact com- bined with the similar chemical behaviour of these elements may be the reason why similar sensitivities are recorded. In general the alkaline earths behave in a more predictable and coherent fashion than the alkali metals and this observation also holds true for some other geological matrices under study in this laboratory e.g. CaPO (unpublished data). Those elements that are relatively volatile and can form hydride species are also shown in Table 5. Although the data are a little sparse in general sensitivity is higher than the alkaline earth elements with the exception of As where sensitivity is poor in all of the rock groups studied. This latter point may reflect the low degree of elemental ionization for As in the ICP.In order to characterize behaviour patterns further elements have been sub-divided into arbitrary categories based on elemental boiling-point since enhanced sensitivity has already been observed for some elements with low boiling-points (Table 6). However there are no clear relationships displayed either within a single rock group or with boiling-point. Sensitivity is therefore not simply related to the physical properties of an element but seems to be more closely controlled by the physical location of that element in a particular mineral in the silicate rock matrix. Some of the elements for which data are presented in Table 6 occur in silicate rocks as the major component in minor phases such as phosphates (e.g. P in apatite) oxides (e.g. Table 5 Relative responses* for some trace elements in six silicate RMs Element W-2 BIR-1 DNC-1 G-2 SRM 2704 Alkali metals- Li 112 87 1 1 1 78 - Na 40 44 39 10 12 Rb 49 - 48 21 - Ca 66 - - 33 - Alkaline earths- Mg 30 29 30 25 20 Ca 20 20 15 6 18 Sr 17 21 17 4 Ba 19 26 19 7 16 - Volatile elements- 10 As 9 - - - Sn - 457 - 124 - Sb 131 57 131 - 102 Te Pb 137 121 235 41 170 - - - - - sco- 1 93 18 39 48 19 13 19 19 14 140 101 178 - *Results given in counts g pg-' corrected to 55Mn as internal standard and for isotopic abundance30 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1993 VOL.8 Table 6 Relative responses* for some trace elements in six silicate RMs categorized by boiling-point Element W-2 BIR- 1 DNC- 1 G-2 SRM 2704 Boiling-point < 10 000 K- 49 - 48 21 - 10 Rb c s 66 As 9 P 1.7 1 .o 1.6 0.5 1.8 - 33 - - - - - Boiling-point 1000- 1 500 K- Na 40 44 39 10 12 Mg 30 29 30 25 20 Zn 40 28 47 23 56 95 Cd - - - - Boiling-point 1 500-2000 K- - Li 112 87 11 78 Ca 20 20 15 6 18 Sr 17 21 17 4 Ba 19 26 19 7 16 T1 - - - 67 128 - Boiling-point 2000-2500 K- IS IS IS IS 121 235 41 170 Mn 1st Pb 137 Sb 131 57 131 - 102 Boiling point >2500 K (in order of increasing temperature)- Sn - (457)$ - 124 Ga 55 54 55 33 A1 10 11 9 4 c u 43 23 54 12 Cr 39 27 29 91 Ni 25 23 32 Fe 35 36 37 30 sc 18 14 13 18 c o 29 31 35 18 13 Nd 36 Ti 35 29 36 26 Y 20 20 19 13 V 38 38 44 30 Ce 35 32 31 18 La 17 - 21 8 12 Pr 20 18 B 23 - 278 U 107 - Zr 35 21 37 16 Nb 41 8 27 26 10 Th 36 33 Hf 37 Ta - 380 W - - - - - - - - - - - - - - - - - - 10 34 46 21 33 26 21 29 - - - - 43 *Results given as counts g pg-’ corrected to 5sMn as internal standard and for isotopic abundance.?IS= internal standard. $Value in parentheses represents an unreliable reference value. sco- 1 39 48 14 1.6 18 19 41 93 13 19 19 124 IS 178 101 140 95 10 33 33 22 42 16 30 25 27 14 31 23 13 24 15 56 8 35 17 55 136 - Cr in spinel) sulfides (e.g. Cu in chalcopyrite) and also Ti in sphene and Zr in zircon. The degree of release of specific elements from the solid as vapour during the ablation is difficult to predict (using fixed-Q mode part of the sample is removed as vapour and part as particulate). However it is clear that extremely close matrix matching is required in terms of both chemical and physical form and that mineralogy plays an important role in determining elemen- tal sensitivity.Further studies are currently underway to try to elucidate these problems. Multi-element Results It is clear from the above data that to make fully quantitative measurements appropriate standards are re- quired matched both in chemical and mineralogical com- position. To test the accuracy of LA-ICP-MS for a wide range of elements W-2 was used as a standard in a single- point calibration (forced through the origin) and BIR-1 and DNC- 1 analysed against that calibration. The remaining RMs G-2 SCo-1 and NIST SRM 2704 were not considered further. Detection limits were calculated for all elements for which concentration data are available in either of the three basic RMs. Measurement of an appropriate ‘blank’ for such calculations has received some attention in the literature.A number of alternative methods have been suggested including collecting spectra whilst no laser action is taking place15 or by analysing of a solid poly(tetrafluor0- ethylene) block which might represent the solid bindersJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1993 VOL. 8 31 Table 7 Multi-element limits of detection (LOD) calculated as the concentration equivalent to three times the standard deviation of the background counts at mlz 193 (n= 10) Element Mass (mlz) LOD/pg g-l Element Mass (mlz) LOD/pg g-' Li B Na Mg A1 P Ca sc Ti V Cr Mn Fe c o Ni cu Zn Ga AS Rb 7 1 1 23 24 27 31 44 45 46 51 52 55 57 59 60 63 66 69 75 85 0.05 0.31 0.14 0.24 0.58 3.20 0.30 1.96 0.15 0.17 0.12 7.37 0.19 0.87 0.19 0.50 0.17 0.62 0.16 13.3 Sr Y Zr Nb Cd Sn Sb c s Ba La Ce Pr Nd Hf Ta W TI Pb Th U 88 89 90 93 1 1 1 120 121 133 138 139 140 141 146 178 181 182 205 208 232 238 0.40 0.28 0.3 1 0.14 0.19 0.1 I 0.07 0.08 0.4 1 0.34 0.18 0.28 0.90 0.32 0.08 0.03 0.1 1 0.08 0.15 0.05 used during sample preparati~n.~~ However neither of these approaches is entirely satisfactory and no generally agreed method has been proposed.An alternative technique is measurement of background counts at a single mass position where an element is not present in the sample. This procedure is appropriate for most elements since many polyatomic peaks which would normally result in enhanced backgrounds (4-g. 40Ar1601H at mlz 57) are absent in the dry plasma. Although the method is not entirely satisfac- tory it provides a useful measure of a lower limit for quantitative measurement.Detection limits (Table 7) are calculated here as the concentration equal to three times the standard deviation of the background counts at mlz 193 for ten scans recorded during the acquisition of elemental data. Correction has been made for isotopic abundance. The scan range was from mlz 6 to 240. Concentrations are reported as pg g-l and although they range from 0.03 (W) to 13.3 (Ca) pg g-l typical values are a few hundred ng g-' or less. Major elements All of the major elements lie in the lower part of the mass range below mlz 60 and concentrations range from < 120 to 97 000 pg g-l over two orders of magnitude (Table 8). The precision on the raw integrals (n= 5) is typically better than 10% relative standard deviation (RSD) with the exception of Ti which is poorer.The accuracy of the results when compared with what are considered highly reliable refer- ence values (& t 2 % relative) is generally better than f 5% relative. Those elements that display poorer accuracy particularly P and Ti tend to occur at the lowest concentra- tion while in addition P is poorly ionized in the ICP (only about 37%) resulting in low sensitivity (Table 6). It is also worth noting that the errors reported on the reference values for P are between 25 and 50% relative. The accuracy of the data presented here is not un-typical of that obtained during conventional solution analysis by ICP-MS where sample dilution factors are typically between 10 000 and 50 000 although still not of the standard expected for the determina- tion of major elements by competitive techniques.Trace elements The results for the trace elements are more wide ranging in terms of accuracy and precision and therefore for ease of discussion have been sub-divided into five categories. Uncertainties in the trace element reference values should be taken into account when assessing the accuracy of the measured data. Several factors may affect the quality of the results including the uncertainty on the values used for W-2 as the calibration standard and the precision of individual measurements. However taking into account these factors some general statements may be made. The term 'good accuracy' is used below to identify the measured data which are statistically indistinguishable from the recommended values taking into account the errors reported on these values.To evaluate fully the accuracy of the technique a broad range of sample types and greater number of RMs would be required. Alkaline earths and alkali metals Of the two RMs analysed results for DNC-1 display a higher degree of accuracy within this elemental group with excellent results for Ba Li and Sr (Table 9). The precision of the raw integrals is generally better than 10% RSD. ~ ~~ Table 8 LA-ICP-MS data for two basic rocks using a single-point calibration for major elements; n= 5 Relative Measured Reference* error RSD Oxide (% m/m) (% m/m) (%) (Oh) A 1 2 0 3 18.13 15.35 + 18 4 Fe2°3 11.63 11.26 + 3 9 9.28 9.68 -4 8 CaO 13.24 +0.4 4 Na20 1.95 1.75 + I 1 5 Mgo 13.19 Ti02 0.77 0.96 - 19 17 PZOS 0.027 0.046 - 40 1 1 BIR- I - DNC- 1 - A1203 17.66 18.30 -3 4 Fe203 10.35 9.93 +4 7 MgO 9.94 10.05 - 1 5 CaO 8.57 11.27 - 24 4 Na,O 1.83 1.87 -2 3 Ti02 0.49 0.48 + 2 23 PZOS 0.078 0.085 + 8 8 *Reference values from Govindaraju.2232 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1993 VOL.8 Table 9 LA-ICP-MS data for two basic rocks using a single-point calibration for alkali and alkaline earth elements Table 11 calibration for refractory elements LA-ICP-MS data for two basic rocks using a single-point Relative Measured/ Reference*/ error RSD Element Pg g-' Pg g-' (Ole) (O/O) Li 2.64 3.4 -1- 0.4 - 22 8 Na2Ot 1.95 1.75 + 1 1 5 Sr 134 108+ 14 + 24 4 Ba 10.7 7.7 t- 2.2 + 39 5 CaOt 13.19 13.24 + 0.4 4 Li 5.03 5.1 + 0.5 - 1 16 Na2Ot 1.83 1.87 - 2 3 Sr 146 14526 +0.7 6 Ba 113 114k 16 -0.6 7 CaOt 8.57 11.27 - 24 4 BIR- I- DNC- I- *Reference values from Govindaraju22 with errors from Gladney toh m/m.et al. 25 Measured Reference*/ Error RSD Element Pg g-' Pug g-' ( OiO ) (O/O) Y 16.0 16-1-2 0 1 1 Zr 13.5 2 2 k 7 - 39 10 Nb 0.37 2-1-0.5 - 82 41 Ce 2.26 2.55 1.1 - 9 12 BIR- I - DNC-I- Y 16.7 18-1-3 - 7 9 Zr 43.4 41 -t7 +6 25 Nb I .98 3t-0.7 - 34 34 Ce 9.19 10.6-1- 2.4 - 13 18 *Reference values from GovindarajuZ2 with errors from Gladney et al. 25 Table 12 LA-ICP-MS data for two basic rocks using a single-point calibration for Group I11 metals and volatile and hydride forming elements Table 10 LA-ICP-MS measured for two basic rocks using a single- point calibration for first row transition elements Element BIR- 1- s c V Cr c o Ni c u Zn s c V Cr c o Ni c u Zn DNC- I - Measuredl Pg g-' 34 316 265 55 158 68 50 22 170 217 66 325 123 77 Reference*/ PI2 g-' 44*4 313+23 382 -1- 38 5 1.4 -1- 3.4 166-1- 16 126-1-5 71 -1-9 31 -t 1.4 148+9 285 -1- 32 54.7 + 3.7 247+ 18 96+9 66-1-5 Relative error (Yo) - 24 + 1 -31 +7 - 5 - 46 - 30 - 30 + 15 - 24 +21 +31 + 28 + 16 RSD (O/o) 2 19 24 7 7 8 13 6 25 13 10 1 1 24 5 Relative Measuredl Reference*/ error RSD Element Pg g-' Pg g-' (O/O) (Yo) A1203t 18.13 15.35 + 18 4 BIR- 1 - Ga 15.7 16-1-2 - 2 6 DNC- I- A1203t 17.66 18.30 -3 4 Ga 15.1 15+2 +0.8 8 BIR- I - Sb 0.34 0.58k0.16 -41 18 Pb 2.82 3.2 -1- 0.8 - 12 23 DNC- I- Sb 0.96 0.96 -1- 0.15 0 33 Pb 10.8 + 72 28 6.3+ 1 *Reference values from Govindaraju22 with errors from Gladney to/o m/m.et al. 25 *Reference values from Govindaraju22 with errors from Gladney et al.25 First row transition metals The precision of measurements within this group are very variable from 2 to 25% RSD and are not directly related to concentration (Table 10). Of the 20 results reported only five have an accuracy of better than 20% relative. Refactories The elements within this group include those which may be difficult to determine accurately in solution owing to a number of dissolution/stability problems26 and therefore their accurate determination by LA-ICP-MS could offer a viable alternative method of analysis. The precision ranges from 9 to 40% RSD reflecting the rather low counts and low concentrations (Table 1 1).Counts for Nb in BIR-1 are only just above the detection limit for this element and therefore poor accuracy and precision is not unexpected. However measured values for Ce Y and Zr (and Nb in DNC-1) all fall within the reference values reported for both samples and therefore display a high degree of accuracy within the limitations noted above. Tantalum Hf and W also fall within this category but there are insufficient data available here to comment further. Group 111 metals Only two elements are available from this group either because reference data are unavailable (e.g. B) or elemental concentrations fall below the detection limit (e.g. In). Gallium substitution for A1 can occur in silicate minerals and hence they are reported together. Both elements display good precision (< 10% RSD) and the measured data for Ga are very accurate when compared with the recommended values (Table 12).Volatile and hydride forming elements The relative sensitivity for both Sb and Pb is high with detection limits of about 80 ng g-' (Table 12). The accuracy for Sb even at the sub-ppm level is good measurements on both RMs being within the range of the reference values. Certain elemental groups therefore display good accu- racy e.g. refractories Group I11 metals hydrides and to some extent the alkali metals and alkaline earths when determined by LA-ICP-MS in basic rock matrices. The transition metals present a more complex picture and atJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1993 VOL. 8 33 Table 13 Limits of detection (LOD) for REE Hf Ta and W calculated as the concentration equivalent to three times the standard deviation of the background count at m/z 185; n= 10 Mass LOD/ Element (m/z) Pg g-’ La Ce Pr Nd Sm Eu Gd Tb DY Ho Er Tm Yb Lu Hf Ta W 139 140 141 146 147 151 157 159 163 165 167 169 172 175 178 181 182 0.01 3 0.0 12 0.01 1 0.072 0.084 0.032 0.083 0.0 14 0.056 0.015 0.057 0.0 18 0.063 0.0 15 0.059 0.016 0.0 19 this stage little more can be concluded from these data without further experimental work.To evaluate the possi- bility that error had been introduced by the use of a single- point standard calibration a new calibration was generated using both W-2 and BIR- 1. The remaining basic RM DNC- 1 was then analysed as the ‘unknown’ against this calibra- tion. Accuracy of these new data showed no improvement (and was poorer in a few cases) over the original single- point calibration.Determination of REE Hf Ta and W In addition to the multi-element determinations discussed above analyses were carried out by scanning only the higher part of the mass range from rnlz 137 to 184. Under these conditions improved sensitivity can be obtained for the REE Hf Ta and W elements which typically occur in silicate rocks at concentrations from 0.1 to 100 pg g-l. Detection limits (see under Multi-element Results) have been calculated for these elements using rnlz 185 as a background point. Values reported (Table 13) are typically 1000 1 . . iTj 200 = m . = m m La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Element Fig. 3 Relative responses (counts gpg-I corrected to I3’Ba) for the REE in AGV-1 a few tens of ng g-l and are almost an order of magnitude lower than those available using a wide range scan.There are clear advantages therefore to the selection of elemental groups which lie in discrete parts of the mass range if particularly good sensitivity is required. The relative responses for the REE in AGV-1 are shown in Fig. 3 corrected for isotopic abundance and concentra- tion. Although the relative responses are fairly uniform there is a general decrease in sensitivity from light to heavy REE. Replicate analyses of G-2 have been used to calibrate and measure concentrations in AGV-1. Two runs each consisting of five replicate analyses of G-2 were averaged and a calibration graph constructed from the mean inte- grals.Five replicates of AGV-1 were averaged giving a precision of 5% RSD for the higher abundance light and middle REE (La-Gd) and better than 10% RSD for the lower concentration heavier elements. The accuracy for all elements determined was very good and all measured concentrations with the exception of Eu and Hf lie within the error of the reference values (Table 14). The Hf figure is high by nearly loo% and this may reflect its rather inhomogeneous distribution a feature common to many silicate samples. Con c 1 us i o n System optimization experiments have been carried out to establish those conditions which are best suited for quanti- Table 14 Quantitative REE determination in USGS AGV-1 by LA-ICP-MS Element La Ce Pr Nd Sm Eu Gd Tb DY Ho Er Tm Yb Lu Hf Ta W Isotope w.4 139 140 141 143/ 145/ 146$ 147 151 157 159 163 165 167 169 172 175 178 181 182/ 184$ Measured/ Pg g-‘ 35.9 63.5 33.1 8.82 5.98 1.32 5.10 0.698 4.27 0.7 10 1.80 0.31 1 1.86 0.217 8.35 0.843 0.376 RSD (%I 3 4 3 3 4 5 5 6 6 8 8 5 7 13 4 5 9 Working value*/ ,Ug g-’ 38 67 7.6 33 5.9 1.64 5.0 0.7 3.6 0.67 1.7 0.34 1.72 0.27 5.1 0.90 0.55 Consensus value*/ Pg g-‘ 38k3 66k6 6.5k0.9 34k5 5.9 k 0.5 1.66k0.1 1 5.2 k 0.6 3.8 k 0.4 0.73 +0.08 1.61 k0.22 0.32 k 0.05 1.67 +.0.17 0.28 k 0.03 5.1 k 0.4 0.92 k 0.12 0.53 k0.09 0.71 kO.1 *Working values from Govindaraju.22 ?Consensus values from Gladney et al.25 $Mean concentration used.34 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1993 VOL. 8 tative elemental determination at major minor and trace levels in silicate rock matrices.The ICP-MS instrument was operated at a higher forward power (1.5 kW) and carrier gas flow rate (1.0 min-I) than that used for conventional solution nebulization. A number of laser operating para- meters were also optimized resulting in the use of fixed-Q mode 800 V laser energy a shot repetition rate of 20 Hz with five shots per site and a total of 24 sites per analysis. Under these conditions sensitivity was good with detection limits better than 50 ng g-I for many elements whilst major element concentrations (e.g. 97 000 pg g-’ for Al) were still within the linear range of the ion detector. Internal standardization was necessary to compensate for differential ablation within a single sample and between samples 55Mn being used for wide scans and 13’Ba for the narrow mass range scans for the REE and heavier trace elements.As a result of this study further work will be carried out to evaluate the role of multiple internal standardization using elements from each of the chemical groups identified. Relative elemental responses varied from one element to another by over an order of magnitude and in some cases were found to be sample dependent. Quantitative measurements could be made under these optimum conditions providing that samples and standards were closely matched both in terms of bulk chemistry and more importantly mineralogy. Accuracy and precision were therefore assessed for the basic rocks. Major elements showed good accuracy and precision. Trace element data were more mixed but certain elemental groups i.e.refrac- tories Group I11 metals hydride forming elements and to some extent the alkaline earths showed good accuracy when compared with working values. Transition metal behaviour was more difficult to predict with many ele- ments showing poor accuracy and precision. A narrow mass range scan across the upper part of the mass range gave an improvement in detection limit for the REE Hf Ta and W by nearly an order of magnitude. Accurate REE determinations were made down to a few hundred ng g-’. The ICP-MS Facility Royal Holloway University of Lon- don is supported by the Natural Environment Research Council and the Ministry of Defence and this continued support is gratefully acknowledged. Thanks are also due to Dr W. Perkins (Aberystwyth) for providing us with laser time during the early part of this work.1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 References Gray A. L. and Date A. R. Analyst 1983 108 1033. Mainman T. H. Nature (London) 1960 187 493. Moenke-Blankenburg L. Laser Micro Analysis Wiley-Inter- science New York 1989. Dittrich K. and Wennich R. Prog. Anal. ,4t. Spectrosc. 1984 7 139. Verbueken A. H. Bruynseels F. J. Van Grieken R. and Adams F. Laser Microprobe Mass Spectrometry eds. Adams F. Gijbels R. and Van Grieken R. Wiley New York vol. 95 in Chemical Analysis 1988 pp. 173-256. Carr J. W. and Horlick G. Spectrochim. Acta Part B 1982 37 1. Thompson M. Goulter J. E. and Sieper F. Analyst 1981 106 32. Thompson M. Chenery S. and Brett L. J. Anal. At. Spectrom. 1989 4 1 1 . Mitchell P. G. Ruggles J. A. Sneddon J. and Radziemski L. J. Anal. Lett. 1985 18 1723. Imai N. Anal. Chim. Acta 1990 235 381. Marshall J. Franks J. Abell I. and Tye C. J. Anal. At. Spectrom. I99 1 6 145. Mochizuki T. Sakashita A. Iwata H. Kagaya T. Shima- mura T. and Blair P. Anal. Sci. 1988 4 403. Mochizuki T. Sakashita A. Tsuji T. Iwata H. Ishibashi Y. and Gunji N. Anal. Sci. 1991 7 479. Mochizuki T. Sakashita A. Iwata H. Ishibashi Y. and Gunji N. Anal. Sci. 1991 7 151. Gray A. L. Analyst 1985 110 551. Arrowsmith P. Anal. Chem. 1987 59 1437. Pearce J. G. Perkins W. T. Abell I. Duller G. A. T. and Fuge R. J. Anal. At. Spectrom. 1992 7 53. Darke S. A. Long S. E. Pickford C. J. and Tyson J. F. Fresenius’ J. Anal. Chem. 1990 337 284. Hager J. W. Anal. Chem. 1989 61 1243. Arrowsmith P. and Hughes S. K. Appl. Spectrosc. 1988,42 1231. Perkins W. T. Fuge R. and Pearce N. J. G. J. Anal. At. Spectrom. 1991 6 445. Govindaraju K. Geostand. Newsl. 1989 13 1 . Gray A. L. and Williams J. G. J. Anal. At. Spectrom. 1987 2 599. Jarvis K. E. Gray A. L. and Houk R. S. Handbook of Inductively Coupled Plasma Mass Spectrometry Blackie Glas- gow 1992. Gladney E. S. Burns C. E. and Roelandts I. Geostand. Newsl. 1983 7 3. Jarvis K. E. Chem. Geol. 1990 83 89. Paper 2 /02828J Received May 29 1992 Acceuted Seutember 9. 1992