399 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 Removal of Organic Solvents by Cryogenic Desolvation in inductively Coupled Plasma Mass Spectrometry* Invited Lecture Luis C. Alvest Michael G. Minnich Daniel R. Wiederin* and R. S. Houk6 Ames Laboratory US Department of Energy Department of Chemistry low&tate University Ames IA 5001 1 USA Methanol ethanol acetone or acetonitrile were nebulized continuously with an ultrasonic nebulizer. The solvent was removed from the aerosol stream by repetitive heating at approximately 100 "C and cooling in a set of cryogenic loops at -80 "C. The resulting aerosol was then introduced into an inductively coupled plasma mass spectrometer. Ethanol was the only solvent that required a continuous dose of additional 0 ( 1 4 % ) in the aerosol gas to prevent deposition of carbon on the sampler.Oxide ratios for LaO+:La+ and UO+:U + were 0.03-0.1 YO. Cryogenic desolvation attenuated but did not eliminate the usual carbon-containing polyatomic ions (e.g. CO' C02+ Arc+ and ArCO+). Analyte sensitivities from metal nitrate salts in methanol were comparable to the sensitivities from aqueous metal solutions. Substantial memory effects were observed from several metal complexes. Keywords Organic solvents; inductively coupled plasma mass spectrometry; desolvation; polyatomic ions The introduction of organic solvents into an inductively coupled plasma (ICP) is beset with Generally volatile organic solvents overload the plasma so that a high forward power is necessary to stabilize the di~charge.~ Analysis of organic solvents by ICP mass spectrometry (ICP-MS) faces several additional complications.Carbon deposits on the interface numerous polyatomic ions such as CO+ C02+ and Arc' are observed and the sensitivity and detection limits for analyte ions are usually poorer than those obtained when aqueous solutions are nebulized.l0?'l These problems are so severe that many analysts prefer to simply digest organic samples and introduce them as aqueous solutions. When organic solvents cannot be avoided most analysts use only relatively involatile ones such as xylene or isobutyl methyl ketone. In ICP-MS a small dose of O2 is often added to the aerosol gas to prevent carbon deposition1*l2. As might be expected this remedy increases spectral overlap problems from metal oxide ions (MO+).Removal of the solvent after nebulization is one general solution to these problems. Maessen et d5 have examined this remedy thoroughly for ICP emission spectrometry. Cryogenic desolvation at -77 "C was used by Wiederin et ~ 1 . ' ~ for the analysis of organic solvents by ICP atomic emission spec- trometry. Essentially the ICP operated stably at forward power levels of only about 1.0 kW when cryogenic desolvation was employed. Analyte emission sensitivity was similar to that obtained from aqueous solutions and interferences from mol- ecular bands were minimal.13 Cryogenic desolvation has proven very valuable for removal of water and HCl from aqueous samples in ICP-MS.14,15 The current paper extends this work to organic solvents. Hill et di2 have recently described a condensation system based on Peltier coolers for analysis of organic solvents by ICP-MS.They did not cool the aerosol below -4O"C and they noted the probable advantages of removing more solvent by cooling the aerosol at still lower temperatures. The solvents studied in the present work are relatively volatile and are usually considered * Presented at the XXVIII Colloquium Spectroscopicum Internationale (CSI) Post-Symposium 5th Surrey Conference on Plasma Source Mass Spectrometry Durham UK July 4-6 1993. t Present address G. D. Searle Searle R&D Building P 4701 Searle Parkway Skokie IL 60077 USA. $ Present address Cetac Technologes Incorporated 5600 S. 42nd Street Omaha NE 68107 USA. 9 To whom correspondence should be addressed.among the more 'difficult' solvents used for introduction into the ICP.16 The behaviour of simple inorganic salts is also compared with that of relatively volatile metal complexes because loss of volatile species or memory effects are possible problems when aerosols are heated. Experimental A Perkin-Elmer Sciex ELAN Model 250 ICP mass spec- trometer with upgraded ion optics and software was used. Typical operating conditions are listed in Table 1. Note that an ultrasonic n e b ~ l i z e r ' ~ ~ ~ ~ was used. Operating conditions were optimized daily to maximize the signal for La+ in methanol. The ion lens voltages required to maximize the analyte ion signals were the same when either the various organic solvents or aqueous solution^'^^^^ were nebulized. The liquid flow rate was 1.5 ml min-'.The cryogenic desolvation system is shown in Fig. 1. This apparatus is similar to the standard condenser provided with the nebulizer. The bulk of the organic solvent became con- densed in the first loops. This condensed solvent was still liquid and was drained off periodically through Tygon tubing under the loops. The aerosol was then heated and cooled repeatedly in a set of copper loops similar to those used for aqueous ~olvents.'~ This process removed much of the remain- ing solvent from the aerosol. Without the heating steps solvent tended to condense back from the vapour phase onto the sample particles in the cold loops; these undesirable wet droplets were then transported readily to the plasma as noted previously by Maessen et aL5 and Wiederin et ~ 1 .' ~ Peak hopping data were acquired in the multi-element mode at low resolution setting ( 1 u width at 10% valley) with three measurements per peak 20 ms dwell time and 1 s measurement time. Spectra were acquired in the sequential mode with ten measurements per peak and a 1 s measurement time. Count rates were not corrected for isotopic overlaps. Chemicals and Standard Solutions High-performance liquid chromatography grade organic sol- vents were used for all experiments (Fisher Scientific Pittsburgh PA USA). Standard solutions were prepared by diluting aliquots from 1000 mg 1-1 aqueous standards (Plasma Chem Farmingdale NJ USA) with methanol. The aqueous standards were supplied as nitrate salts. For comparative purposes stock solutions of metal complexes400 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL.9 Table 1 Typical operating conditions _____~ ~ Ultrasonic Nebulizer Current setting (arbitrary units) Desolvation heater temperature Desolvation condenser temperature/"C Cryocooler ICP torch Plasma forward power/kW Argon flow ratefl min-' Outer Intermediate Aerosol Sampling position Sampler Skimmer Ion lens settingsp Bessel box stop Bessel box barrel Bessel box plate Einzel 1 and 3 Einzel 2 Electron multiplier voltagem Cetac Technologies (Omaha NE USA) Model U-5000 6 Dependent on solvent - 10 Cryocool CC-10011 at - 80 "C; Neslab (Portsmouth NH USA) Ames Laboratory designI7; outer tube extended 30 mm from inner tubes 1.25-1.5 12 0.3 1.6 20 mm above load coil on centre Copper 1.1 mm diameter orifice Nickel 0.9 mm diameter orifice - 5.9 + 5.4 - 11.0 - 19.8 - 130.0 - 4000 Heating/cooling coils (CU loops) Kept at \fl? heater temperature Dry -7 + Y Drain Fig.1 Diagram of cryogenic desolvation device -80 "C .(absolute ethanol) with various ligands were also prepared. These analyte com- pounds were selected to represent a range of melting- and boiling-points. Metal acetate stock solutions were prepared by dissolving analytical-reagent grade acetate salts of Zn Co and Mn (Fisher Scientific) at a concentration of 1 mg 1-l in meth- anol. The Co(CO),NO Y(acac) (acac = acetylacetonate) and Nb(OC,H,) (Strem Chemicals Newburyport MA USA) were weighed and dissolved completely in methanol to give solutions that contained 1 rng1-l of the metal. Analyte solutions were then prepared by further dilution of these stock solutions.Results and Discussion General Observations The solvents and analyte compounds studied and their melting- and boiling-points are presented in Table 2. As noted pre- vi~usly'~ the ultrasonic nebulizer produces an extremely intense aerosol from these solvents. Nevertheless the plasma operates stably at moderate forward powers of 1.25-1.5 kW. After ignition the axial channel is punched by simply turning on the aerosol gas flow during nebulization of the solvent The heater temperature was selected to be 40°C above the boiling- point of the solvent as noted previ0us1y.l~ Methanol and acetonitrile could be nebulized indefinitely without noticeable green C2 emission or deposition of carbon on the sampler.Substantial carbon deposition and C emission are observed when ethanol is nebulized. In this case oxygen is added to the aerosol flow through the side arm of a small T-junction ( 5 mm id.) at the base of the torch. The argon carrying the aerosol passes straight through the T-junction into the injector tube of the torch. The additional oxygen burns off the carbon deposited on the sampling cone. Oxygen is bled in gradually until the O2 flow rate is just high enough to remove the green C2 emission. This 0 flow is typically 1-5% of the aerosol gas flow. When acetone was nebulized a small amount of carbon was deposited on the cone. This carbon deposition caused the analyte signal to drift down by roughly 10% per hour which is worse than the drift normally seen.Adding a brief burst of O2 into the aerosol gas flow for approximately 30 s every 2 h removes this deposit. After the 0 burst the analyte signal recovers to its original value within the usual relative precision interval of 1-2%. With cryogenic desolvation continuous addition of 0 was not necessary for the analysis of acetone. The different behaviour seen for these four solvents can be explained based upon their melting-points and on the ratio of carbon to oxygen in each solvent molecule. The cold loops are at - 80 "C some 34 "C below the melting-point of acetonitrile (Table 2). Thus the vapour pressure of acetonitrile at the exit of the cryocondenser is very low. The other three solvents have higher vapour pressures because the loop temperature is above their melting-points. Methanol and acetone have almost the same melting-points (about - 95 "C) but acetone causes more C emission and carbon deposition because it has a 2:l ratio of C atoms relative to 0.Ethanol has a still lower melting- point and a 2:l ratio of C:O. Thus the solvent load out of the condenser is greatest for ethanol and the stoichiometry of ethanol also favours carbon deposition and C2 emission. Background Spectra Generally the mass spectra from an ICP containing organic solvents show the usual major ions (e.g. Ar+ ArH' Ar2+ 0' and H20f) as well as substantial levels of additional polyatomic ions from the constituents of the solvent.l0?l1 The count rates observed for four of the more troublesome polya- tomic ions from each solvent are shown in Table 3.As expected the count rates for these background ions vary with the solvent used since a fixed loop temperature yields different solvent loads. No Cu' from the loops is observed when organic solvents are nebulized. Acetonitrile was unexpectedly found to give the highest level of ArO' even though there is no oxygen in acetonitrile. This was possibly due to the presence of a substantial concentration of Fe' (up to 3 pg l-') which contributes to the background at m/z=56. The background at m/z= 56 in Table 3 is also substantially higher than that seen during cryogenic desolv-JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 401 Table 2 Physical data for solvents and solutes and heater temperatures used Solvent Methanol Ethanol Acetone Acetonitrile Solute* Mn( CH3C00)2 * 4H20 Co( CH3COOj2*4H20 Zn( CH3COO) - 2H20 Y (acac) Co(CO),NO Nb(OC3H5)5 Melting-point/"C Boiling-point/"C Heater temperature/"C - 93.9 65 105 - 117.3 78.5 119 - 95.35 56.2 96 - 45.7 81.6 122 > 300 > 270 248 139 6 - ~ ~~~~ * These solutes were introduced as methanol solutions; the heater temperature was 105 "C in each case.Table 3 Count rates for polyatomic ions observed for various organic solvents Count rate/counts s-' Solvent CO + (m/z = 28) CO (m/z=44) Arc+ (m/z=52) ArO+ (m/z=56) Methanol 1.7 x lo6 11 x 103 5.4 x 103 25 x 10' Acetone 3.0 x lo6 48 103 24 x 103 30 x 10' Ethanol 1.5 x lo6 29 x 103 15 x 103 17 x 10' Ethanol + 0 3.7 x lo6 64 x 103 83 x 103 64 x lo2 Acetoni trile 0.12 x lo6 24 x 103 15 x 103 66 x lo2 ation of aqueous aerosol^.^^*^^ When ethanol is nebulized the addition of 0 enhances the signals from all four of these ions even from Arc' which contains no oxygen.Naturally CO' C02' and Arc' are much more intense when organic solvents are nebulized than is the case with aqueous solutions. Background equivalent concentration (BEC) values for '*Si' 44Ca+ 52Cr+ and 56Fe+ are presented in Table4 for methanol as solvent. For each case the BEC is better (ie. a lower value) in the present work than in two earlier attempts at ICP-MS of organic even though methanol is more volatile and thus a more 'difficult' solvent than the xylene and white spirit used in previous studies.l07l1 Despite this improvement the four polyatomic ions CO' CO,' Arc' and ArO' are still abundant enough to cause problems even with cryogenic desolvation at - 80 "C.Analyte Sensitivity and Oxide Formation Solutions containing La and U at 5OOpg1-' in different solvents were nebulized to determine if analyte sensitivities varied among solvents. These two elements form refractory oxide ions so the abundance of metal oxide ions was also measured relative to that for metal ions (M') to evaluate Table 4 Background equivalent concentrations for analytes at m/z values corresponding to CO+ CO,+ Arc' and ArO+ BEC values/pg 1-'* Reference Si + Ca + Cr + Fe+ (solvent) (m/z = 28) (m/z = 44) (m/z = 52 j (m/z = 56) Methanol 8 50 240 3.0 1.2 Xylene'O 12x 103 5oox 103 400 40 White spirit'' -? 8.4x 103 1.2x 103 20 * BEC=solution concentration of analyte required to give a net signal equal to the signal for the polyatomic ion. Concentrations refer to the total amount of the element required to provide the necessary signal at the particular m/z shown i.e.they have been corrected for isotopic abundance. t Signal for CO' was too high to measure in this case. possible problems from spectral interferences caused by MO +. A summary of the data obtained is presented in Table 5. Sensitivities for La and U vary slightly among the solvents with the exception of ethanol where the analyte sensitivities are poorer by a factor of 5. The addition of 1-2% 0 to the central channel while nebulizing ethanol boosts the signal for La' and U+ without a prohibitive increase in the abundance of L a o + and UO'. As expected acetonitrile gives the lowest MO+:M + ratios because it lacks oxygen.Metal carbides (MC') are not observed for either La (< 10 counts s-l net) or U (<20 counts s-l net) even when the methanol sample contains La and U at 100mgl-l. The analyte sensitivities obtained from organic solvents are also similar to those observed from aqueous samples with this particular ICP-MS instrument. 1 4 9 1 5 Organic versus Inorganic Metal Standards The objectives of this experiment were to determine (i) if inorganic- and organic-bound metals had different sensitivities and (ii) if volatile and low-melting organometallic complexes would be lost in the cryogenic loops. Solutions of inorganic Co Zn and Mn ions [e.g. Zn(NO,),] at 500 pgl-' were prepared in methanol and a second set of solutions of Co Zn and Mn acetates at 500 pg 1-1 were also prepared in methanol.The melting-points for the acetates were measured and are shown in Table 2. The measured sensitivities i.e. count rate per unit concen- tration are presented in Table 6. For each metal the sensitivity from the acetate complex is essentially the same as that from the inorganic nitrate. Therefore little or no analyte is lost from the acetate complexes during the desolvation process. These acetate complexes have melting-points of 248 "C or higher. They neither melt nor boil in the heaters which are at 105 "C. Thus the acetate complexes are not volatilized and pass through the desolvation device readily as solid aerosol particles. Three other metal complexes were tested in much the same fashion. Solutions of Co(C03)N0 Y (acac) and Nb(OC,H,) were prepared in methanol with each metal present at a concentration of 100 pg 1- '.Reference solutions were prepared by dissolving Co Y and Nb nitrates in methanol at concen-402 ...,,. JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 Table 5 Signals for La+ U+ and metal oxides (MO') in different solvents* Count rate (counts s-') or oxide ratio (YO) Species 1 3 9 ~ ~ + 2 3 8 ~ + La0 + Lao+ :La+ uo+ uo+ :u+ Methanol Acetonitrile Acetone Ethanol Ethanol + O,? 1.8 x lo6 1.9 x lo6 2.6 x lo6 3.8 x 105 2.1 x 106 6.1 105 8.1 x 105 1.4 x lo6 3.2 x 105 9.6 x 105 766 655 1349 203 1759 583 367 1300 250 733 0.04 0.03 0.05 0.05 0.08 0.1 0.05 0.09 0.08 0.08 * Organic solutions contained 500 pg 1-' of La and U. t Oxygen was added to the central channel at about 2% of the total aerosol gas flow.Table 6 Sensitivity for analyte elements as acetate complexes and as inorganic nitrate salts in methanol Sensitivity/I06 counts s-' per mg I-' Analyte "Mn+ 'j4Zn + 59c0 + Acetate complex Nitrate salt 1.6 1.3 1.1 1.7 1.1 1.4 trations of 100 pg1-I. As shown in Table 7 the sensitivity fo-r Nb in the low-melting Nb(OC2H5) is similar to that for the Nb nitrate salt. The sensitivities for Co and Y in the complex Co(CO),NO and Y(acac) are somewhat higher than the sensitivities for Co and Y as nitrate salts. The Co(C0,)NO boils at 50 "C (Table 2) some 55 "C below the temperature of the heated loops. Likewise Nb(OC,H,) should melt in the heated loops yet the Nb+ signal shown in Table 6 for this complex indicates that it passes through the loops readily.Apparently boiling or melting of the species in these aerosols does not cause a drastic loss of analyte during the desolvation process. Detection limits for Co Y and Nb in either complex or nitrate form are 80 20 and 30 ng l-' respectively. Memory Effects Substantial memory effects are seen from some of the com- plexes. The worst such problem is illustrated by the rinse-out curves in Fig. 2. The Nb' signal when the analyte is present as Nb(NO,) decays to 0.1% of the steady-state level about 60s after the sample is removed as noted previously for various elemenfs.l4 In contrast the Nb+ signal from Nb(OC,H,) never decays to the 0.1% level but stops at a level corresponding to about 2% of the steady-state signal. Similar though less severe memory effects are seen for Y (acac) and Co(CO),NO whose rinse-out curves level off at 0.2 and 0.3% of the steady-state signals respectively.Such memory effects are occasionally seen when solutions containing volatile species are dispersed into finely-divided aerosols. Memory problems can also be exacerbated when the aerosols are heated. For example inorganic forms of Hg B Table 7 Sensitivity for analyte elements as complexes and as inorganic nitrate salts in methanol Sensitivity/106 counts s-' per mg 1-' Analyte 5 9 c ~ + 89y + 93Nb+ Complex 2.8* 3.8t 1.8$ Nitrate salt 1.7 2.4 1.7 * Co present as Co(CO),NO. -f Y present as Y(acac),. $ Nb present as Nb(OC,H,),. I I I I 0 120 240 360 480 Time/s Fig. 2 Rinse out curves for Nb' in methanol. A Nb at 100 pg 1-' as Nb(N03),; and B Nb at 300 pg 1-' as Nb(OC2H5)5.In each case the Nb sample was removed at the time indicated by the vertical broken line and 0 s cause similar problems in conventional desolvation The precise causes of memory effects are often obscure or not easily attributable to simple chemical reasoning and the present work is no exception. For example Co(CO),NO boils at 50°C and should be readily volatilized in the heaters which are at 105 "C. The Nb(OC,H,) (b.p. 142 "C) should melt in the heaters but not boil yet Nb(OC,H,) shows a much worse memory problem than Co(CO),NO. At any rate the chemical form of the analyte element does influence the sample throughput and rinse-out procedures required as discussed previously by Van Heuzen.16 Conclusion Cryogenic desolvation allows continuous analysis of difficult solvents that would otherwise plug the sampler or extinguish the plasma.Polyatomic ions are less abundant than is usually the case with organic solvents although some important analytes are still obscured particularly 28Si 44Ca 52Cr and 56Fe. This desolvation method should prove valuable for the measurement of trace amounts of inorganic ions in highly purified organic solvents used in the semiconductor industry and in materials sciences. However the analyst should investi- gate possible memory effects that depend on the chemical form and physical properties of the analyte particularly when the analyte is present as neutral or volatile complexes. Also the sensitivity may depend somewhat on the chemical form of the analyte for reasons that are unclear at this time.In the future the use of still lower temperatures for the cold loops could prove advantageous as none of the solvents were frozen into solids in the present work. Caution Since the heater temperature (100-140 "C) may be higher than the flash point of the solvent highly flammable vapours and aerosols produced from organic solvents and organometallic compounds must be kept within the inert gas stream present inside the nebulizer and cryocondensers. TheJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 403 aerosol or vapour must not come in contact with the heating elements on the exterior of the heater. Care should also be observed when handling oxygen. An arrestor valve was installed on the oxygen cylinder.The loan of an ultrasonic nebulizer from Cetac Technologies Incorporated is gratefully acknowledged. Ames Laboratory is operated by Iowa State University for the US Department of Energy under Contract No. W-7405-Eng-82. This research was supported by the Office of Basic Energy Sciences Division of Chemical Sciences. References Boorn A. W. and Browner R. F. Anal. Chem. 1982 54 1402. Hausler D. W. and Taylor L. T. Anal. Chem. 1981 53 1223. Hausler D. W. and Taylor L. T. Anal. Chem. 1981 53 1227. Kreuning G. and Maessen F. J. M. Spectrochim. Acta Part B 1989 44 367. Maessen F. J. M. J. Kreuning G. and Balke J. Spectrochim. Acta Part B 1986 41 3. Maessen F. J. M. J. Seeverens P. J. H. and Kreuning G. Spectrochim. Acta Part B 1984 39 1171. Brotherton T. Barnes B. Vela N. and Caruso J. J. Anal. At. Spectrom. 1987 2 389. 8 9 10 11 12 13 14 15 16 17 18 19 Barrett P. and Pruszkowska E. Anal. Chem. 1984 56 1927. Blades M. W. and Caughlin B. L. Spectrochim. Acta Part B 1985 40 579. Hausler D. Spectrochim. Acta Part B 1987 42 63. Hutton R. C. J. Anal. At. Spectrom. 1986 1 259. Hill S. J. Hartley J. and Ebdon L. J Anal. At. Spectrom. 1992 7 23 895. Wiederin D. R. Houk R. S. Winge R. K. and D’Silva A. P. Anal. Chem. 1990 62 1155. Alves L. C. Wiederin D. R. and Houk R. S. Anal. Chem. 1992 64 1 164. Alves L. C. Allen L. A. and Houk R. S. Anal. Chem. 1993 65 2468. Van Heuzen A. A. 4th Surrey Conference on Plasma Source Mass Spectrometry Guildford Surrey UK July 1991. Scott R. H. Fassel V. A. Kniseley R. N. and Nixon D. E. Anal. Chem. 1974 46 75. Fassel V. A. and Bear B. R. Spectrochim. Acta Part B. 1986 41 1089. Olson K. W. Haas W. J. Jr. and Fassel V. A. Anal. Chem. 1977 46 632. Paper 310465 1 F Received August 3 1993 Accepted November 29 1993