Anal. Chem. 1995, 67,4557-4564
Quantitation of Metal Isotope Ratios by Laser Desorption Time-of-Flight Mass Spectrometry lphigenia L. Koumenis,tl* Marvin L. Vestal,+Atfred L. Yergey,*ss Steven Abramsi," Stanley N. Deming,l and T. William Hutchens*lf-+
Vestec Corporation, Houston, Texas 77054,Laboratory of Theoretical and Physical Biology, NICHD, Bethesda, Maryland 20892, USDNARS Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texas 77030,Department of Chemistry, University of Houston, Houston, Texas 77004,and Protein Structure Laboratory, USDNARS Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texas 77030
Laser desorption time-of-flightmass spectrometry (LD/ TOF-MS) is evaluated for the determination of stable metal isotope ratios. The isotope ratios of f i e metal ions (Cu, Ca, Mg, Fe, Zn) in atomic absorption standard solutions and two metal ions (Ca, Mg) in human serum samples are determined. With an existing LDflDF-MS instrument we show that the technique can overcome the ditticulties of the most commonly used methods for measuring metal isotope ratios: (1) all metals are ionizable without surface treatment, thus overcoming the major drawback of thermal ionization mass spectrometry (TIMS); (2) there is no matrix involved to interfere with the metal ion detection, thus overcoming the major disadvantage of inductively coupled plasma mass spectrometry (ICPMS ); (3) there is no interference from hydride ions, a major disadvantage of fast atom bombardment secondary ionization mass spectrometry; (4) a mixture of metals can be detected simultaneously using a single laser wavelength, overcoming the major disadvantage of resonance ionization mass spectrometry; (5) accuracy and precision comparable to ICPMS can be achieved with the current instrumentation; (6) precision comparable to TIMS is feasible; and most importantly (7) high precision can be achieved on very small quantities of material because the LDflDF-MS instrument permits all masses to be monitored simultaneously and very small differences in isotope ratio can be detected. There are a number of important metallic nutrients which, when deficient in the body, may lead to a variety of serious disorders.'S2 It is generally recognized that stable isotopes of the metals have advantages over the use of their radioactive equivalents in nutrition and metabolic research in human subjects. It +
Vestec Corp.
of Anesthesia, Stanford Medical School, Stanford University, Palo Alto, CA 94305. 8 NICHD. For reprints, write to: Dr. Alfred L. Yergey, NIH, 10 Center Drive, MSC 1580,Room 6C208, Bethesda, MD 20892. l 1 Baylor College of Medicine. Wniversity of Houston. Protein Structure Laboratory, Baylor College of Medicine. Current address: Department of Food Science and Technology, University of California, Davis, CA 95616. (1) Anthony, A. Current Topics in Nutrition and Disease; A R Liss: New York, 1977; p 1. (2) Dobbing, J., Ed. Springer-Verlag: London; New York, 1990. :Current address: Department
'
+
0003-2700/95/0367-4557$9.00/0 0 1995 American Chemical Society
is also recognized that the most general method of determining these isotopes is by mass spectrometric isotope ratiometry. Thermal ionization mass spectrometry VIMS) can be used to determine virtually all of the nutritionally important metals with well-recognized high levels of precision and acc~racyP-~The major disadvantages of TIMS are the preparation of sample filaments, and the sometimes slow rate of data acquisition,20 min/ sample. Part of the sample preparation requirement is the need for ionization enhancement techniques to obtain adequate signal intensities when measuring transition metals, Fe, Zn, and Cu being nutritionally important example^.^ Other techniques currently in use for measurement of metal isotope ratios are inductively coupled plasma mass spectrometry (ICPMS) fast atom bombardment secondary ion mass spectrometry (FAEVSIMS) ,l4-I5 and resonance ionization mass spectrometry (RIMS).16-17While ICPMS is a very sensitive multielement technique, it suffers from interference ions generated by plasma matrix ion/molecule reactions that restrict its use in Ca isotope ratio measurements and limit its precision and accuracy for certain elements, Fe and Se in particular.'O The principle problem with the use of FAEVSIMS arises from perturbation of minor isotope intensities by hydride ions of adjacent major isotopes. Recent results show that the method can account satisfactorily for fractionation but not for the hydride a d d i t i ~ n . ~ J ~ (3)Huemann, IC G. Biomed. Mass Specfrom. 1985,12,477-488. (4)Eastell, R;Viera, N. E; Yergey, A L; Riggs, B. L. J Bone Miner. Res. 1989, 4,463-468. (5) Yergey, A. L.; Vieira, N. E Hansen, J. W. Anal. Chem. 1980,52, 18111814. (6)Montes, J. G.; Sjodin, R. A; Yergey, A L.; Keira, N. E. Biophys. J. 1989, 56,437-446. (7)Tumlund, J. R Crit. Reu. Food Sci. Nufr. 1991,30.387-396. (8) Tumlund, J. R J. Nutr. 1989,119,7-14. (9)Eagles, J.; Fairweather-Tait, S. J.; Portwood, D. E.; Self, R; Gotz, A; Heumann, IC G. Anal. Chem. 1989,61,1023-1025. (10)Fomon, S.J.; Janghorbani, M.; Ting, B. T.; Ziegler, E. E.; Rogers, R R.; Nelson S. E.; Ostedgard, L. S.; Edwards, B. B. Pediatr. Res. 1988,24,2024. (11) Janghorbani, M.; Ting, B. T. Anal. Chem. 1989,61,701-708. (12) Egan, C. B.; Smith, F. G.; Houk, R S.; Serfass, R E.Am. J. Clin. Nutr. 1991, 53,547-553. (13) Schuette, S.A.; Ziegler, E. E.; Nelson, S. E.; Janghorbani, M. Pediatr. Res. 1990,27,36-40. (14)Ramseyer, G. 0.; Brenna, J. T.; Morrison, G. H.; Schwartz, R. Anal. Chem. 1984,56,402-407. (15) Miller, L. V.; Hambridge, K M.; Fennessey, P. V. Anal. Chim. Acta 1990, 241,249-254. (16)Fassett, J. D.;Powell, L. J.; Moore, L. J. Anal. Chem. 1984,56,2228-2233. (17)Smith, D. H.; Young, J. P.; Shaw, R. W. Mass Spectrom. Rev. 1989,8,345378.
Analytical Chemistry, Vol. 67,No. 24, December 15, 1995 4557
The disadvantages of RIMS are that different wavelengths are Parts Inc, Miami Lakes, FL) and platinum probe tips 2 mm in needed for each element (because the technique depends on the diameter (Alpha Chemicals, Ward Hill, MA) were used as sample desorption of sample with the specific wavelength of each metal substrates. ion at its absorption maximum), and depending on the element, Sample Preparation. Serum samples (0.5 mL) from preterm a one-photon or two-photon desorption/ionization is needed. In infants (some samples enriched in 25Mgand some enriched in some cases, one laser is needed for desorption and another for 42Ca) were centrifuged, and the plasma was digested in a ionization, which can make RIMS complicated and expensive. microwave oven with concentrated HNO3 (Ultrex grade, J. T. Time-of-flight mass spectrometry (TOF-MS) has been comBaker, Phillipsburg, NJ) and then neutralized with concentrated mercially available since the mid-l950s, but recent advances in NHdOH (Ultrex grade, J. T. Baker). Each was then loaded onto timing circuitry and high-speed data acquisition systems have led a cation exchange membrane filter, AG 50 x 8, hydrogen form to major improvements in resolution. Coupling of these TOF(BioRad, Richmond CA). The membrane was washed with 10 MS instruments to the laser desorptidionization t e c h n i q ~ e * ~ ~ mL ~ ~ of methanol and 10 mL of deionized water before the sample has made this method of mass spectrometryan extremely useful, was loaded. Mg was eluted with 2 mL of 6 M HCl (Ultrex grade, and still evolving, tool to address a variety of problems in BioRad). The sample was dried under a Nz purging evaporator molecular recognition, structural biology, and b i o p h y s i ~ s . ~ ~ - ~ ~ and then redissolved in 10 pL of 0.3% HN03. The early time-of-flight instruments were equipped with Serum samples (0.5 mL) for Ca determination were prepared electronics that did not permit the recording of all ions produced by adding 10 mL of a saturated solution of ammonium oxalate to during each time-of-flightcycle. Instead, the mass spectrum was the plasma which was then centrifuged. The remaining solid was constructed by recording the ion signal measured during a gated baked in a 200 “C oven. The calcium oxalate salt was redissolved time window in each cycle and advancing the gate in successive in 10 p L of 0.3%HN03 for anal~sis.~ cycles, a method termed “time slice detection”.33 This method The sample, usually 1pL (1pg/pL) of the AA standards or 5 was limited to very low mass/charge detection with low resolution p L (-0.6 pg/pL of metal ion total) of the serum samples, was and sensitivity. Modem time-of-flight instruments are regarded airdried on the center of the probe tip prior to insertion in the as highly sensitive, with essentially unlimited mass range. instrument for analysis. Because of their linear geometry and absence of beam-defining Instnunentation. A modded Vestec Model VT2OOO laser slits, they have very high ion transmission. In contrast to the desorption time-of-flight mass spectrometer (Vestec Corp., Houssector instruments, the time-of-flight mass spectrometer has ton TX) was used. The instrument was designed primarily for become the simplest and most reliable way to record ion signals the determination of molecular weights of proteins using the distributed over a very broad mass range. matrix-assisted laser desorption technique pioneered by Karas and The purpose of this work was to explore the potential of laser Hillencamp18 and Koichi et al.19 The laser is neodymiumdesorption time-of-flight mass spectrometry (LD/TOF-MS) for yttrium-aluminum-garnet (Nd-YAG) Q-switched pulsed (Model accurate and precise quantification of metal isotope ratios; more HY-400, Lumonics Inc, Livonia MI). The optics include a variable specifically, to determine the limitations and define the problems attenuator, a Bin. focal length fused silica lens, and a phototransof existing LD/TOF-MS instrumentationso new instruments can istor for generating the “start” signal. The system produces 90 be developed for application to metal ion bioavailability studies. mJ of UV radiation/lO ns pulse at frequencies of 20 Hz. Power density is -lo6 W/cm2 for a laser spot size of 0.03 mm2. The EXPERIMENTAL SECTION wavelength used is the third harmonic, 355 nm. The mass Reagents. Metal salt samples [Cu(NO&, Zn(NO&, Mg(C1)2, spectrometer has a 2.1-m flight tube and two voltage stages (10 Fe(C1)2, Ca(N03)2]were all atomic absorption (AA) standards and 4 kV used for these experiments) for ion acceleration. The from Fisher Scientific. The serum samples analyzed were from detector is the combination of a 1-in. channel plate (Model 1330human infants. Stainless steel probe tips 2 mm in diameter (Small 2500, Galileo Electrooptics Corp., Sturbridge, MA) and a discrete dynode electron multiplier (EM) (Model R2362, Hamamatsu (18) b a s , M,; Hillenkamp. F.Anal. Chem. 1988,60,2299-2301. (19)Koichi. T.; Hiroaki, W.; Yutaka, I.; Satoshi, A; Yoshikazu, Y.; Tamio, Y. Rapid Corp., Bridgewater, NJ) withvariable gain (-102-108), which can Commun. Mass Spectrom. 1988.2,151-153. be adjusted by varying the multiplier high voltage. For data (20) Karas, M.; Bahr, U.; Hillenkamp, F. Anal. Chem. 1988,60,2299-2301. acquisition, a LeCroy modular transient digitizer (Model TR8828D, (21)Cotter, R J. Biomed. Enuiron. Muss Spectrom. 1989, 18,513-532. (22) Hutchens, T.W.; Nelson, R W.; Yip, T. T. j . Mol. Recognit. 1991,4,151Le Croy, Chestnut Ridge, NY) with onboard averaging and GPIB 153. interface to a 46/33 computer was used. The transient spectrum (23)Hillenkamp, F.;Karas, M.; Beavis, R C.; Chait, B. T. Anal. Chem. 1991, is digitized using an &bit digitizer. The digitizer can acquire 255 63,1193-1203. (24)Hutchens, T. W.;Nelson, R W.; Yip, T. T. FEBS Lett 1992,296,99-102. transient spectra with an averaging rate of 10 spectra/s for a wide (25) Nelson, R W.; Hutchens, T. W. Rapid Commun Mass Spectrom. 1992,6, mass range and at higher rates over limited mass ranges. 4-8. For this work, a second digitizer with higher acquisition rate (26) Cotter, R J. Anal. Chem. 1992,64,1027A-1039A (27) Hutchens, T.W.; Nelson, R W.; Allen; M. H.; Li; C . M.; Yip, T. T. Biol. and higher number of channels was evaluated. The Tektronk Mass Spectrom. 1992,21,151-159. transient digitizer (Model TDS520, Tektronk, Beaverton, OR) (28) Yip, T. T.; Hutchens, T. W. FEES Lett. 1992,308,149-153. uses an &bit flash and can average at rates up to 500 mega(29) Yip, T. T.; Hutchens, T. W. Tech. Protein Chem. 1992,4 , 201-210. (30)Eckerskorn, C.; Strupat, IC; b a s , M.; Hillenkamp, F.; Lottspeich, F. sampleds, up to 50 OOO channels at 2 ns/channel. As many as Electrophoresis 1992,13,664-665. 9999 spectra can be acquired and averaged at rates up to 500 (31)Hutchens, T.W.; Yip, T. T. Rapid Commun Mass Spectrom. 1993,7,576580. megasampleds over a limited mass range. (32)Nelson, R W.; McLean, M. A: Hutchens, T. W. Anal Chem. 1994,66,1408Integration of the Peaks. Mass spectral peaks were inte1415. grated over the upper 90% of the peak height. This minimizes (33) Holland, J. F.: Enke, C . G.;Alfison, J.; Stults, J.T.; Pinkston, J. D.; Newcombe, B.;Watson, J. T. Anal. Chem. 1983,55, 497A the error provided by random noise baseline noise, and it gave 4558
Analytical Chemistry, Vol. 67, No. 24, December 15, 1995
24
24 5
26
265
m
*5
I
m r u a r g e (MI
Figure 1. (Upper) LDROF mass spectra of Mg AA standard; (lower) serum enriched in 25Mg: 1 pg, 250 laser shots averaged, both acquired on Pt probe tips.
MssslCharge
Figure 2. (Upper)LDROF mass spectra of Ca AA standard; (lower) serum enriched in 42Ca: 1 pg, 250 laser shots averaged, both
acquired on Pt probe tips.
much more reproducible results than integrating the whole peak from the baseline or from the upper 80%or 50%of the peak. Software. LabCalc software (Galactic Industries Corp., Salem, NH) was used for data analysis. A user-written program was used to calculate percent RSDs. RESULTS AND DISCUSSION A major part of this research project was devoted to the study of the factors that might affect the precision and accuracy of the quantitation of the metal isotope ratios of Cu, Ca, Mg, Fe, and Zn. Emphasis was given to the metal ions with relatively high ionization potentials (IP) such as Mg (IP = 7.646 eV), Fe (IP = 7.870 eV), Cu (IP = 7.726 eV), and Zn (IP = 9.394 eV). These elements present the highest degree of dif6culty during analysis by the TIMS instrument because the signal is very short lived, and amounts of 50-60 pg (50-60 p L of 1mg/mL) of sample are required. All the metals mentioned above can be detected by LD/TOFMS with very good signal/noise ratios and excellent resolution. This allowed baseline separation of all the isotopes studied. Figure 1,upper spectrum, shows a representative LD/TOF spectrum of a Mg atomic absorption standard solution. The sensitivity of the instrument is very high as illustrated in Figure 1, lower panel, where a Mg spectrum can be obtained, and isotopic ratios can be quantified with RSD of 0.55%,from only 5 pL (-0.6 pg/pL metal ions) of a 25Mg-enrichedserum sample from a preterm infant. TIMS requires 50 pmol of the same sample for analysis. Spectra of a Ca atomic absorption standard solution and a 42Caenriched serum extract are shown in Figure 2. The 46Ca cannot be quantified in the present configuration of the instrument because of insufficient signal (0.004%abundance), Similar spectral pairs are shown in Figures 3 and 4 for Zn and Cu, respectively. The 70Zn signal is too low to give a response. In Figure 5 the full spectrum of a 25Mg-enrichedserum sample is shown. A mixture
U
Is
W
67
Is
m
m
(W
Figure 3. LDROF mass spectra of Zn AA standard: 1 pg, 250 laser
shots averaged, acquired on Pt probe tip.
of metals present naturally in the serum can be detected simultaneously with high signal/noise ratio. Finally, Figure 6 further illustrates the potential of this technique for multielemental analysis by showing the spectrum of a Cu/Zn 1:3 mixture. Because Cu and Zn have relatively high ionization potentials and because Zn has 1.668 eV higher IP than Cu, the simultaneous ionization of both metals was not expected. Fundamental Considerations of the LDATIF-MS Technique. All ions generated, including the high ionization potential Analytical Chemistw, Vol. 67, No. 24, December 15, 1995
4559
/;
{
I
l i ' I
--
I P
Q
ea
M
ed
n
ea
~ . . . ~ c h(rmgh .)
Figure 4. LDFOF mass spectra of Cu AA standard: 1 pg, 250 laser shots averaged, acquired on Pt probe tip.
Figure 6. LD/TOF mass spectra of a mixture of Cu and Zn AA standards in a ratio of 1:3, respectively: 250 laser shots averaged, both on Pt probe tips.
scale signal is 10 mA = 10 x total number of ions is given by
N,
= [(30 x 10-9 s)(10 x
C/s out of the EM and the
C/s)(6.023 x
iondmol) (mo1/96500 C)I/G = 1.87109/G= -109/G where N,, is the number of ions in the full-scale peak and G is the gain of the multiplier, which can be adjusted from 100 to 108 by changing the EM high voltage. The minimum number of ions that can be detected at signal/noise = 2 is 10 pA = 10 x C/s so Nminis approximately
N,
= [(30 x
(10 x
C/s) (6.023 x
ions/mol) (mo1/96500 C)I/G = 1.87 x 106/G =
106/G I 2s
x)
4
34 W
f
D
O
46
50
I
(mh)
Figure 5. LDFOF mass spectra of NaC, Mg+, AI+, K+, Ca+, and Fe+, which exist naturally in serum sample: 4 pmol, 250 laser shots averaged, acquired on Pt probe tip.
metal ions (e.g., Mg, Fe, Cu, and Zn), seem to be efficiently ionized and transmitted to the detection system without affect on the quantitation of isotopic ratios (see Table 2, and section below Accuracy of the Method). Consequently, the limitations on the precision of the metal ion isotope quantitation seem to be determined primarily by the ion detection and the recording system. The typical time width of a 63Cupeak, for example, is -30 ns (fwhm) with the 2.1-m TOF analyzer used in this work. A full4560 Analytical Chemistry, Vol. 67, No. 24, December 15, 1995
If D is the dynamic range of the digitizer that is used for a particular peak (for a full-scale peak D = loo), then the maximum number of ions in that peak is given by
N = N,,/D
= 109/GD
(1)
The number of ions collected in the peak is
NI = NFT= 10gFT/GD
(2)
where F is the frequency of the laser and Tis the total acquisition time. Following Poisson statistics, the standard deviation (% SD) is given by % SD = 100(NI)-"2
(3)
The above represents the ideal situation where the collected ions are accurately counted, but this is not always possible. In practical terms, the electrical signal is digitized and those numbers are recorded. By using the same triangular approximation to the peak as above the number of counts recorded for a particular peak is given by
N,, = nm
5 ,
(4)
where n is the number assigned to a full-scale peak and m is the width (fwhm) of the peak in channels. The total number of counts recorded, N,, is given by Nc = FTN,,/D
=
(nm)FT/D
and the overall % SD is given by % SD = 100(NI-’/z N,-’/’)
+
(5)
I
(6)
I
I
I
I
I
I
1800 2000 2200 2400 2600 2800 3000 3200 3400
Elactron Multiplier Voitage (volts)
+
% S D = 1 0 0 ( D / m > ” 2 [ ( G / 1 0 ~ 1 ~(l/nm)’’’I 2
(7)
where (G/lOg)is the number of ions collected in a full-scale peak and l/nm is the number of counts recorded from the digitizer for the same peak. We attempted in this study to determine the contribution of these numbers to the overall precision and to find which of the two numbers (the number of ions collected in a fullscale peak or the number of ions recorded) is the limiting factor for the precision. Effect of “Electron Multiplier Gain” on the Precision. Laser power determines the number of ions initially produced; EM gain determines the number of ions finally detected. The reproducibility of the measurements can depend on both the number of ions that can be produced and the number of ions that can be detected during each time-of-fight cycle. Because these two factors are highly interactive, they were varied simultaneously and the signal was monitored. CuII was used for this study because it has only two isotopes and it was easy to evaluate over a broad range of instrumental settings. The EM high voltage, which controls the gain, was varied from 2000 to 3200 V, and the relative standard deviation was calculated at each point for the 63Cu/65Curatio. The results are illustrated in Figure 7A Ten runs, each an average of 250 laser shots, were used at each EM voltage level to calculate % RSDs. The signal on the oscilloscope was kept at 70%of the full scale (70%of total number of bits) to avoid inadvertent saturation of the signal. As shown in Figure 7 4 the precision improves as the EM voltage decreases. As the EM voltage was decreased, the laser intensity was increased to keep the signal at the 70%of the full scale. By increasing the laser intensity, the number of ions desorbed by the surface was increased. Consequently, the number of ions counted by the digitizer is the determining factor for the precision at the low area of EM voltage. As the EM voltage increases; the % RSD for the Cu ratio increases, as is observed in Figure 7R The laser intensity is decreased as the voltage on the EM increases and the number of ions desorbed by the surface are decreased. In this case, the precision is limited by the number of ions desorbed by the surface. The data from this study suggest that the precision should improve by keeping the EM in the low-voltage region, using a relatively high laser power, and employing a larger dynamic range digitizer.
2.7
1
1.5
4
1.2
1 1800 2000 2200 2400 2600 2800 3OOO 3200 2400
Electron Multiplier Voltage (volts)
Figure 7 . (A, top) Increasing RSD as a function of EM voltage (gain). Precision is ion limited at the low end of the EM gain. (6, bottom) decreased 63Cu/65Curatio as a function of EM gain because of inadvertent saturation of signal.
Effect of “Laser Power Density” on the Cu Isotopic Ratio. Figure 7B shows the 63Cu/65Curatio as a function of the EM voltage over the range 2000- 3200 V. Three measurements consisting of an average of 10 runs of 250 laser shots of three replicated data sets are plotted at each EM voltage to get the isotopic ratio. The 63Cu/65Curatio decreases with increasing EM voltage. This result was unexpected because there was no limitation in the system that could have caused an increase in the W u isotope in the higher EM gain. This result is due to inadvertent saturation of the signal, which leads to a decrease in the peak area for 63Cu. Even though extra care has been taken to avoid saturating the detector (use of 70%of the total digitizer scale), some of the spectra were out of scale as a result of saturating the detector. This effect was more prominent with the higher abundance peaks such as 63Cu,24Mg,“Zn, 56Fe,and 40Ca. This is a result of the variabilitl in laser power density from laser shot to laser shot which leads to a variable number of total ions desorbed and ionized from the sample surface. Analytical Chemistty, Vol. 67, No. 24,December 15, 1995
4561
Mechanisms for improving accuracy and precision due to inadvertent saturation of signal are discussed in the section Improvements Required below. Evaluation of Substrate for Sample Deposition. Different substrates have been evaluated with the five metals under study to find a substrate that would give the best signal/noise and best % RSD values, especially for the high ionization potential metals such as Cu and Zn. Stainless steel substrates, which are the most commonly used in LD/TOF, had interfering signals with the two lower abundant isotopes of iron: "Fe by 54Cr(2.36%abundant) and 5sFeby 58Ni (68.3%dbundant). Gold-plated stainless steel probe tips were also evaluated because the IP of gold is relatively high and it should not interfere with the ionization of the studied metals. The Mg and Ca isotope ratios and % RSD values were the same with both tip surfaces. The Cu signal was slightly weaker on the Au-surface probe tips. Stainless steel has Fe and Zn, which interfered with the actual measurements of the isotopes of these two metals. High-purity graphite (Unocal-Poco), with an even higher IP (11.260 eV) than Au, was evaluated. The ionization and detection of carbon clusters, C3 and above, interfered with the ionization and detection of the metals of interest. Vitreous graphite (Sigradur K, Sigri Corp.) is a special brand of graphite that is fused at very high temperatures to achieve glassy carbon properties (harder but more fragile than ordinary graphite). There were no carbon clusters formed, but the metal signals were very weak. Platinum probe tips (Alpha Chemicals) proved to be the best sample deposition substrate, and useful data were obtained for all metals. All metals seemed to be ionizable on Pt (IP = 9.OOO eV), even Zn, which has a higher IP than platinum (IP = 9.394 eV). The Pt+ was detected but it was not interfering in the m/z region of the Zn or any of the metal ions. In the case of Ca, the 43Caand W a isotopes were detected but the signal/noise was not high enough to obtain an accurate estimate of the area of the peaks, and these ratios are not presented here. Effect of the Digitizer Statistics on the Precision. The mean ratios and RSD values for the five metals of interest are shown in Table 1. The EM was operated at 2300 V. For the LeCroy %bit digitizer (n = 28 = 256), 5-ns channels (m = 30/5), the maximum number of counts per peak (nm) is 1500. Because only 70%of the full scale was used, only 7.4 from the 8 bits, were used. The maximum number of counts for a peak is thus 27.4= 168. The results reported here correspond to 1000 laser shots, so % SD for 63Cu= 1.95%(see equation (7)) is expected when multiplier gain is around lo6. Considering the 65Cupeak, which has a natural abundance of 30.83%,D is 31.31 because it is 0.4472 times as abundant as 63Cu. The number of bits used is 31.31%of the &bit scale, so n = 5.676 and a % SD (65Cu)= 3.0%is expected. The % SD of the ratio (r) of the two components 63Cu/65Cuis estimated by34 % SD(r) = [(%SDt3Cu))'
+ (% S D ~ 5 C ~ ) ) 2 ] 1 ' 2(8)
and the calculated RSD(r) = 3.577%/2.236 = 1.599%. The observed RSD for the 63Cu/65Curatio is 1.32%,which agrees very well with the expected value. (34) Daniel, H. C. Quantitative Chemical Analysis, 3rd ed.; Wiley: New York, 1991; p 43.
4562 Analytical Chemisty, Vol. 67, No. 24, December 15, 1995
Table I. Mean Ratlos and RSD Values.
metal
m/z
% abund
ratio
Cu
63 65
63/65
2.236
Zn
64
69.17 30.83 48.69 27.90 4.100 18.80 0.600 78.99 10.00 11.01 5.800 91.72 2.200 0.280 96.94 0.647 0.135 2.086 0.004 0.187
64/66
1.742
67/66 68/66 70/66 24/25
ME -
Fe
Ca
66 67 68 70 24 25 26 54 56 57 58 40 42 43 44 46 48
exptd
obsd
%RSD
2.234
1.32
0.147 0.674 0.022 7.899
0.204 0.637
8.00 3.00
6.841
4.00
26/25 54/56
1.101 0.063
1.147 0.076
1.50 8.00
57/56 58/56 40/42
0.024 0.003 149.8
0.028 0.003
12.00 15.00
43/42 44/42 46/42 48/42
0.209 3.224 0.006 0.289
3.472
15.00
0.397
14.00
Average values from 10 runs of 1000 laser shots. The digitizer used is a LeCroy TR8828D, 8-bit, 5 n s channel. Metal salts are atomic absorption standards. A 1-pL aliquot of a 1 pg/pL sample was deposited on a 2-mm-diameterprobe and air-dried. The Tektronix &bit, 2-ns channel digitizer was used to acquire 4000, 8000, or 16 000 laser shots of 10 runs in each case. The results from this study are shown in Table 2. The expected % SD for 63Cu,for example, where 1250 laser shots at F = 20 Hz require T = 62.5 s and D is 70 as with the LeCroy (both &bit), is 1.62%. The gain is 106,and because this is a 2-11s channel digitizer, m = 30/2 = 15, and n = 168. Following the same logic, the % SD for 65Cuwas calculated to be 2.216%. From eq 8, the expected % SD for the ratio is 2.74%and the observed value for the RSD of the ratio is 1.20%. Following equation (7), by increasing the total number of laser shots from 1250 to 8000, the RSD of the ratio should improve by a factor of (8000/1250)1/2= 2.52, and it should become 1.20/2.52 = 0.47%. Indeed, the observed value is 0.45%. For the 16 000 laser shots the RSD should improve by a factor of (16 000/1250)1/2 = 3.57, and the RSD should become 1.20/3.57 = 0.33%. The observed value is 0.19%. From above data we can calculate the number of laser shots need to be averaged to achieve 0.01% precision, which is 1.8 x lo7laser shots. The same trend, where % RSD decreases as the number of laser shots increase, was observed for all the metals analyzed. For the metals that are in lower abundance, a higher number of laser shots is needed to reach the precision levels required for the nutritional studies. In the case of 26Mg,which is 11.01%abundant, D = 9.76, n = 2.17, and RSD is calculated to be 0.13%for the 4000 laser shots and 0.97%is observed. At 16 000 laser shots RSD is 0.55%. For RSD of 0.01%,4.8 x lo7 laser shots need to be averaged. Mass Depletion of Sample as a Function of Analysis Time (Number of Laser Shots Averaged). To investigate the depletion of sample, five different spots on the Pt probe tip with deposited Cu (1yg) were analyzed for 20 min. EM voltage was kept constant at 2300 V and the laser 0 frequency was fixed at 20 Hz. It required 12.5 s to digitize 250 laser shots, and 20 min to digitize 24 000 laser shots. Data were analyzed for the 1st (12.5 s), 4th (50 s), 10th (2 min), 24th (5 min), 48th (10 min), 72nd (15 min), and 96th (20 min) runs (of 250 laser shots) from each of the five spots. The mean ratios as a function of time are shown
Table 2. Mean Ratios and RSD Values for Cu, Mg, and Ca.
metal m/z %abund ratio Cu Cu Cu Mg
Ca
63 65 63 65 63 65 24 24 25 26 26 26 40 42 43 44 44 46 48 48
69.17 30.90 69.17 30.90 69.17 30.90 78.99 78.99 10.00 11.01 11.01 11.01 96.94 0.647 0.135 2.086 2.086 0.004 0.187 0.187
40 42 43 44 44 46 48 48
96.94 0.647 0.135 2.086 2.086 0.004 0.187 0.187
24 24 26 25 26
78.99 78.99 11.01 10.00 11.01
exptd
no. of obsd lasershots %RSD
63/65
2.236 2.239
1250
1.20
63/65
2.236 2.239
8000
0.45
63/65
2.236 2.238
16000
0.19
24/25 24/25
7.899 6.281 7.899 6.402
8000 16000
7.39 4.26
26/25 1.101 1.078 26/25 1.101 1.076 1.101 1.077 26/25 40/42 149.8
4000 8000 16000
0.97 0.62 0.55
8000 16000
0.91 0.57
43/42 44/42 44/42 46/42 48/42 48/42
0.209 3.224 3.224 0.006 0.289 0.289 Ca Enriched in 40/42 149.8
2.508 2.503
2.4
4
2'o 1 .8
11
1.6
5
0
15
10
25
20
Time (min)
0.225 0.225
8000 16000
5.00 2.72
8000 16000
1.68 0.82
42Ca
43/42 0.209 44/42 3.224 1.573 44/42 3.224 1.583 46/42 0.006 48/42 0.289 0.170 48/42 0.289 0.171 Mg Enriched in z5Mg 24/25 7.899 3.638 24/25 7.899 3.635 26/25 1.101 1.480
8000 16000
4.13 2.15
8000 16000 8000
5.65 4.97 2.06
26/25
16000
0.67
1.101 0.479
Figure 8. Constant 63CuP5Curatio with increasing analysis time (number of averaged laser shots). No mass depletion of sample is observed.
5 4
3
3 4
The digitizer used is a Tektronix TDS 520, &bit,2-ns channel. Cu, Mg, and Ca salts are atomic. absorption standards. The Ca and Mg are serum samples from pahents under an isotope-enriched diet. A 1-pLaliquot of a lpg/pL meta! salt AA standard and 5 pL of a 0.6 pug/ pL serum sample were deposited on a 2-mm diameter probe and airdried in each case. Serum sample, subject EC S1. Serum sample, subject NEVS1.
*
0
I
I
I
I
I
I
I
10
20
30
40
50
80
70
80
Time (min)
in Figure 8. The Cu ratio is independent of the analysis time and consequently the number of laser shots; thus, there appears to be no mass depletion of the sample as a function of time. The same study was performed with the enriched biological samples of Mg and Ca on the Pt probe tip with 10 p L of deposited sample. In Figures 9 and 10 are plotted respectively the Mg and Ca ratios as a function of analysis time. The ratios were measured every 8000 laser shots, which corresponds to 400 s (6.7 min) between measurements. The last ratio was measured after 80 OOO laser shots, which corresponds to a total analysis time of 4000 s (66.7 min). As with the Cu study, no mass depletion is observed as a function of time for Ca and Mg in the serum samples. This finding is of fundamental importance because it demonstrates that precision can be improved by averaging a higher number of laser shots without a detrimental effect on the isotopic ratio. Accuracy of the Method. By accuracy we refer to the agreement between the isotopic ratio observed, and the natural abundance ratio. The percent relative error for the Cu 63/65 ratio is 0.06%, for Mg 24/25 is 3.2% , for Mg 26/25 is 2.02%, for Ca 44/42 is 8.84%,and for Ca46/42 is 0.42%. For Mg and Ca, the
Figure 9. Constant 24Mg/25Mgand 26MgP5Mgratios in the 25Mgenriched serum sample with increasing analysis time (number of averaged laser shots). No mass depletion of sample is observed.
relative error is higher than expected and as seen from Table 2 the observed ratios are lower than the calculated values. This might be due to inadvertent saturation of signal of the more abundant peak as observed and discussed before. Several potential sources of inaccuracy besides the inadvertent saturation of signal, were considered mass discrimination of the ion optics, mass selective depletion of the sample, inadequate mass resolution, interferences, and chemical noise. The high-transmission timeof-flight mass analyzer used in this study eliminates mass discrimination in the ion optics. Mass-selectivedepletion is not a signiticant issue with LD (see Figures 7-10). In thermal ionization it appears that a microscopic bulk removal of the sample occurs instead of thermal vaporization. This causes the lighter isotope to deplete faster than the heavier one and leads to large inaccuracies in the isotope ratios. The mass resolution of the existing instrument is at least 3 times that required to separate adjacent peaks to 10%valley in Analytical Chemistry, Vol. 67, No. 24, December 15, 1995
4563
31
I
E l
I
“‘1
i
Ca48lCa42
0
I
I
I
I
I
0
10
20
30
40
50
I
I
60
70
I 80
Time (min)
Figure 10. Constant 48Ca/42Caand 44Ca/42Ca ratios in the enriched serum sample with increasing analysis time (number of averaged laser shots). No mass depletion of sample is observed.
the 20-70-Da range required for this work. Therefore, additional means for improving mass resolution (e.g., reflectrons) are not necessary. There are no hydride ions observed or any other chemical noise that might perturb the intensities of the metal isotopes. “Improvements” Required. The above experiments demonstrate the need for a digitizer with a larger number of bits. If the experiment is carried out at low EM voltage and the number of ions recorded by the digitizer is the limiting factor for the precision, then by increasing the bits on the digitizer it should improve the precision by a factor of 2”-*,where x is the number of bits on the new digitizer. So for a 12-bit digitizer, the precision should improve by (Z4 = 1/16) a factor of 16. Such a digitizer is commercially available. To reach the 0.01%precision required for the metal isotopic ratio, it is not practical to use the 20-Hz laser because in the case of Mg, for example, a total number of 576 h of acquisition time would be required. An alternative higher frequency laser that can be used for this type of work is a variable frequency of 1-5 Wz, CW, Q-switched Nd-YAG laser, which is commercially available. With the full use of the 12-bit digitizer coupled with the automatic laser control mentioned above, to achieve a 0.01% precision for the Cu isotopic ratio, for example, a total acquisition time of 42 min is required at 5 kHz and a total of 21 min at 10 kHz. For a 0.1%precision, 25.5 min is required at 5 kHz and 12.7 min at 10 kHz. Irreproducibility in the signal intensity can be caused by fluctuations in the output of the laser or changes in the condition of the target area. Inadvertent saturation of the signal can be
4564 Analytical Chemistry, Vol. 67, No. 24, December 15, 1995
eliminated from the total average of laser shots by incorporating an “automatic laser control” system which monitors the signal intensity. This technique takes advantage of the fact that the transient digitizer uses a “stop” signal to trigger the system to transfer the most recently digitized signal to the averager. Two discrimination levels are set: one corresponding to nearly fullscale signal and the other to the m i n i u m acceptable signal. If the peak is higher than the upper level, the first discriminator is triggered, inhibiting the “stop” trigger signal. That spectrum is not included in the average, and the laser intensity is decreased by a small amount. In a similar way, failure to trigger the lower level discriminator causes the spectrum to be rejected and the laser intensity is increased by a small amount. This way any spectra outside the desired window are rejected. Thus the intensity of the major peak is not distorted from any saturated signal, but it averages all smaller signals in the window and fluctuations of the laser output are eliminated. CONCLUSlONS The data presented here outline the most important significant opportunities as well as the limitations of LD/TOF technique for the determination of metal isotopic ratios. We show that the technique is capable of detecting and quantifying metal isotopic ratios with precision comparable to ICPMS . We also demonstrate the potential to achieve precision comparable to TXMS with LD/TOF, a technique much easier to use, with much higher sensitivity, and where all the significant metal isotope ratios can be detected and quantified. Furthermore, with LD/TOF-MS, two or more metals can be analyzed simultaneously with one laser and one kind of substrate, overcoming the disadvantages of ICPMS , and RIMS. We believe LD/MF is a promising technique for metal isotope determination with significant advantages over the existing techniques. ACKNOWLEDGMENT We thank Dr Randall W. Nelson (Arizona State University) for his technical support and constructive comments throughout this project. This research project was funded in part, with an NIH SBiR Grant R43-HD29334-01 (Vestec), with NIH Grant RO1DK 438 50-01 and with federal funds from the U S . Department of Agriculture, Agricultural Research Service under Cooperative Agreement Number 58-6250-1-003 (T.W.H.) . The contents of this publication do not necessarily reflect the views of the U.S. Department of Agriculture, nor does mention of trade names, commercial products, or organizations imply the endorsement by the US. goverment.
m),
Received for review November 28, 1994. Accepted May
IO, 1995.@ AC9411381 @Abstractpublished in Advance ACS Abstracts, July 1, 1995
