Ion Composition Elucidation (ICE)
Determination of Elemental Compositions from Mass Peak Profiles of the Molecular Ion (M) and the M+1 and M+2 Ions
Andrew H. Grange,1 Joseph
R. Donnelly,1 G. Wayne Sovocool2
and William C. Brumley2
Systems & Technologies Co., 980 Kelly Johnson Drive, Las Vegas, Nevada
2U.S. Environmental Protection Agency, National Exposure Research Laboratory, Characterization Research Division, P.O. Box 93478, Las Vegas, Nevada 89193.
The relative abundances of M+1 and M+2 ions help to identify the elemental composition of the molecular ion (M). But scan speed, sensitivity, and resolution limitations of mass spectrometers have impeded determination of these abundances. Mass peak profiling from selected ion recording data (MPPSIRD) provided faster sampling and enhanced sensitivity, which permitted use of higher resolution. M+2 profiles having only a few percent of the ion abundance of M were monitored at 20,000 resolution. The relative abundances, exact masses, and shapes of M, M+1, and M+2 mass peak profiles were determined. By applying five criteria based on these quantities, elemental compositions were determined even for ions too large (up to 766 Da) to be uniquely assigned from their exact mass and accuracy limits alone. A profile generation model (PGM) was written to predict these resolution-dependent quantities by considering all M+1 and M+2 ions for each candidate composition. The model also provided assurance that no other compositions were possible. Characterization of the M+1 and M+2 profiles by MPPSIRD and the PGM greatly expanded the practical ability of high-resolution mass spectrometry to determine elemental compositions.
The U.S. Environmental Protection Agency (EPA), National Exposure Research Laboratory, Characterization Research Division, in Las Vegas, NV, conducts research in separation and identification of unknown compounds in hazardous wastes and other environmental samples. Although analysis by GC/MS is convenient and provides valuable data, it is limited in that unknowns can be difficult to identify from low-resolution mass spectra alone, especially when partial coelution of large amounts of other compounds degrades the quality of background-subtracted mass spectra. Misidentifications based on low-quality library matches can result.
Recent articles describe exact mass measurements that were made to determine compositions of ions. Accuracies of 3 ppm were obtained using a Fourier transform mass spectrometer1 and 5-60 ppm using a high-resolution mass spectrometer at 2500-9000 resolution with conventional scanning.2 A previous article described how the advantages of enhanced sensitivity and reduced measurement time, realized by applying the selected ion recording (SIR) mode of operation to exact mass measurements of single ions, provided exact masses for single ions, formed from compounds that eluted from chromatographic columns as narrow peaks.3 This report demonstrates adaptation of mass peak profiling with SIR data (MPPSIRD) to determine the elemental composition for ions with masses up to 766 Da when multiple compositions have masses within the error range of an exact mass determination.
The number of compositions possible for an exact mass increases with the error limits of the measurement and with the mass of the ion. If more elements and larger numbers of heteroatoms are considered, even more compositions are possible. The number of compositions is related to the mass defect (non-integer mass). For 399.8, 399.9, 400.0, 400.1, 400.2, 400.3, and 400.4 Da, 2, 14, 22, 17, 8, and 3 compositions, respectively, are possible within an error limit of 2.5 ppm (1.0 mDa) when up to 33 C, 70 H, 6 N, 6 0, 4 P, and 4 S atoms are considered. The abundances of M+1 and M+2 ions relative to M have long been used to help determine elemental composition. M+1 and M+2 ions arise from naturally occurring isotopes of elements in the earth's crust.4 The isotopic abundances of elements with multiple isotopes vary considerably. Consequently, the relative abundances of M+1 and M+2 ions can often distinguish between elemental compositions. In addition, differences in mass defects between the isotopes of each element provide multiple M+1 and M+2 ions. Multiple M+1 ions result from different isotopes, such as 2H, 13C, 15N, 17O, 29Si, and 33S. Multiple M+2 ions arise from combinations of two +1 isotopes, such as 13C and 33S or two 13C, as well as from one +2 isotope, such as 18O, 30Si, or 34S. High mass resolution allows observation of three additional features useful for identifying compositions: the exact masses of the M+1 and M+2 profiles and the shape of the M+2 profile. This paper describes data acquisition and plotting procedures that are used to acquire these quantities and presents a model for their prediction. Two examples of composition identification based on data for M+1 and M+2 partial profiles are illustrated. In addition, exact masses and relative abundances were determined at 20,000 resolution for four standards having molecular ions with masses between 536 and 766 Da to estimate the accuracy of the measurements and to demonstrate that correct compositions were determinable for higher mass ions.
Instruments and Conditions. Analyses were performed with a Hewlett-Packard 5890 GC interfaced to a VG 70-250SE high-resolution mass spectrometer equipped with the VG 11-250 data system. A Gateway 486-66 personal computer (PC) was used as an auxiliary data system. A 30-m x 0.25-mm-i.d. Restek RTx-5 GC column was interfaced to the ion source for GC/HRMS analyses. GC/MS analyses used a GC temperature program of 50 oC for 3 min, 12 oC/min to 230 oC, 25 oC/min to 310 oC, final hold of 10 min. Positive ion SIR data were acquired with 70-eV electron ionization, perfluorokerosene (PFK) calibrant, an accelerating potential of 8 kV, mass resolutions between 10,000 and 40,000, an ion source temperature of 250 oC, and a GC transfer line temperature of 280 oC. A heated probe inlet was used to introduce each of four standards into the source with 20,000 resolution. To determine accuracy limits, 15 data acquisitions that monitored the full M profile and 15 data acquisitions that monitored M, M+1, and M+2 partial profiles simultaneously were made for simulated chromatographic peaks 2 min wide. Peaks were simulated to permit automated integration of peak areas by turning a switch located between the preamplifier and flash box off and on to provide a baseline at the start and end of each peak. The switch replaces the practice of closing and opening the valve between the source and analyzer regions.3 Each data set was collected and processed in about 3 h. A new mass calibration preceded each acquisition. The photomultiplier tube voltage was 390 V when 10,000 resolution was used and 500 V when greater resolutions were employed.
Sample Extract. The environmental extracts were obtained through the EPA Contract Laboratory Program and were prepared according to the EPA Statement of Work.5
Standards. Four standards, dibenzo-30-crown-10 (C28H40O10, molecular mass = 536.2621 Da, purity = 98%), methylpyrroporphyrin XXI ethyl ester (C34H40N4O2, 536.3151 Da), diacetyldeuteroporphyrin IX dimethyl ester (C36H38N4O6, 622.2791 Da), and coproporphyrin III tetraethyl ester (C44H54N4O8, 766.3942 Da, 98%) were purchased from Aldrich (Milwaukee, WI) and used without further purification.
Mass Peak Profiles from SIR Data. The ion abundances for several m/z ratios were monitored across the mass peak profile. When a compound with the target mass entered the ion source, a chromatographic peak appeared in each ion chromatogram. The areas of these peaks were calculated by the data system, transmitted to the personal computer (PC), and plotted as a function of mass to provide the mass peak profile.3
When a full mass peak profile was observed at up to 20,000 resolution to obtain an exact mass, measurements were made at 17 m/z ratios across and beyond the profile to check for interferences on either side of the apex and to observe any profile broadening. Unless peak broadening occurred, 10 m/z ratios were sufficient to delineate the full profile, and seven m/z ratios provided a baseline. Five m/z ratios monitored the top portion of the reference mass profile for calibration. For resolutions up to 20,000, the mass increment between the m/z ratios was M/(R x N), where M was the mass of the ion, R was the mass resolution, and N was the number of mass increments across an unbroadened profile. Mass increments of 5 ppm were used for resolutions >20,000.
To acquire a full mass peak profile for an ion, a good estimate
of its exact mass is required. In Figure 1 the process for acquiring
the estimated exact mass is diagrammed.
Figure 1. Full-scan, low-resolution mass spectrum (a), the mass peak profile of m/z 266 obtained with 3,000 resolution over a mass range of 1,600 ppm from which the exact mass was estimated (b), and the same mass peak profile obtained at 22,500 resolution from which the exact mass was determined. (c) The partial profiles of the calibration ion are shown to the right of profiles b and c.
The nominal mass was obtained from the low-resolution data in Figure la for the ion of interest. A mass range of 1,600 ppm that included the most saturated hydrocarbon possible was examined using a mass resolution of 3,000 to provide a survey of all ions having the nominal mass of the analyte ion. The exact mass estimated from the mass peak profile of the analyte ion was then used as the center mass in the SIR descriptor used to acquire data at about 20,000 resolution (Figure 1c). This example is discussed in detail in reference 6. An example for which interferences with the analyte ion at 3,000 resolution were resolved at 20,000 resolution is shown in reference 3.
Partial Mass Peak Profiles. To determine accurately the relative abundances of M+1 and M+2 ions, the M, M+1, and M+2 profiles must be monitored simultaneously by a "split" SIR descriptor, i.e., one that monitors multiple analyte ions. Twenty-two m/z ratios were monitored by the SIR descriptor in this work, which is the number of real-number global variables not assigned for software parameters and therefore available for storage of areas. Too few m/z ratios are available for a single SIR descriptor (25) with the VG11-250 data system to monitor simultaneously three entire profiles while maintaining a sampling resolution of 10 points across each profile. It was therefore necessary to monitor only six m/z ratios across the top portion of each profile (partial profiles). Only four points were measured across the reference mass. To prepare SIR descriptors, the user entered the calculated mass for the M profile or the calculated masses for the M, M+1, and M+2 partial profiles and the resolution into a LOTUS 1-2-3 procedure, which calculated the m/z ratios and prepared an ASCII file. The ASCII file was then executed by Reflection 4, with the PC serving as a terminal of the VG data system. Twenty-two seven-digit masses, the resolution, and other parameters were automatically entered into the SIR descriptor. Other procedures automated data processing and plotted profiles. Details of interfacing a spreadsheet and word processing software to the VG 250 data system are discussed elsewhere.7-9
RESULTS AND DISCUSSION
Often, when a complex mixture is analyzed by low-resolution mass spectrometry, interferences from other compounds with the signals of M, M+1, or M+2 ions from the compound of interest distort relative abundances. Generally, high mass resolution can then be used to separate the desired ion signals from the interferences to make the relative abundances of the M+1 and M+2 ions useful for composition identification. The relative abundances of M+1 and M+2 ions observed in full-scan mass spectra are the sums of the relative abundances of all ions containing heavier isotopes. At mass resolutions of 10,000-40,000, for masses where multiple compositions are possible within the error limit of the exact mass determination, the exact mass differences between the M+2 ions can cause the M+2 profile to broaden substantially and become visibly asymmetric. With sufficient resolution, multiple maxima were observed. For example, the M+2 profile for C27H56S, plotted by the profile generation model described later, features a 90% valley (relative to the smaller peak) between a double maximum due to C27H5634S+ and 13C2C25H56S+ ions at a resolution of 18,000. The mass difference between the two ions arises from the positive mass defect difference between 12C and 13C and the negative mass defect difference between 32S and 34S (2 x 0.00335- (-0.00420) = 0.01090 Da). Mass resolutions between 20,000 and 25,000 were used routinely in our studies. Hence, a valley would be observed experimentally for this compound.
Profile Generation Model. Criteria were based on comparisons of experimental and theoretical values. The theoretical values of both masses and relative abundances are resolution dependent when partial profiles are monitored, since multiple M+1 or M+2 ions are partially resolved at higher resolution. A profile generation model (PGM) was written in QuickBASIC and executed on the PC to compile a list of possible compositions; to predict masses of M, M+1, and M+2 ions; to estimate the relative abundances of M+1 and M+2 profiles when only six points across the top of the M, M+1, and M+2 profiles were monitored; and to display graphical representations of M+1 and M+2 profiles. All possible compositions corresponding to the exact mass of M and the error limits of its determination, chemical valences, and the elements and number of atoms of each element entered by the user were calculated. Five important features used to retain or discard compositions were calculated by the PGM for each composition: (1) the calculated mass of the M+1 partial profile; (2) the calculated mass of the M+2 partial profile; (3) the abundance of the M+1 partial profile relative to the M partial profile; (4) the abundance of the M+2 partial profile relative to the M partial profile; and (5) the shape of the M+2 profile.
For each composition and mass resolution entered by the user, the model calculates the mass and relative abundance of each M+1 and M+2 ion and the sum of their relative abundances for each profile. An example list is shown in Table 1. The PGM currently considers 10 elements: C, Si, N, P, O, S, H, F, Cl, and Br.
Table 1. M, M+1, and M+2 Ion Compositions and
The relative abundances were calculated by the PGMa
a %M+1 and %M+2 values calculated by the PGM.
To predict the appearance of full or partial profiles, a graphical approach using a Gaussian distribution10 to model the mass peak profile for each ion was used. A Gaussian distribution approximated the rounded apex and tailing observed in profiles when the ion beam was tuned. Equation 1 was used to calculate 101 points across a mass range that included the profile. At 20,000 resolution, a mass range of 100 ppm was plotted.
Mp was the mass on the x-axis, Mc was the center mass, and R was the resolution. (Mp-Mc and R/(0.20397Mc corresponded to (x-xmean) and sigma in the Gaussian distribution equation. The width of the peak was determined by the constant (0.20397). Mass resolution was defined as (R = M/deltaM) for a 10% valley between overlapping profiles of equal height, where deltaM was the mass difference between the maxima of the two profiles. With this constant, the mass difference between points at 5% of the maximum on the two sides of either peak, deltaM5%, was equal to deltaM, and the calculated and observed mass peak profiles were very similar in appearance. Hence, the resolution was determined from M profiles as M/deltaM5%.
When multiple M+1 or M+2 ions with different masses contribute to a profile, the amplitudes due to each ion (Ai) are summed at each mass along the profile using eq 2, where y is the sum of relative abundances from all M+1 or M+2 ions at a point along the profile, n is the number of M+1 or M+2 ions, Ai is the abundance of the M+1 or M+2 ion relative to M, Mi is the mass of the M+1 or M+2 ion, Mp is the mass of the point along the profile; and the other parameters are the same as for eq 1.
M+2 profiles calculated for five compositions at a mass resolution of 20,400 are displayed in Figure 2.
Figure 2. Calculated M+2 profiles at 20,400 resolution for five of the compositions in Table 2. The composition numbers are from Table 2.
The vertical lines correspond to points + to - 15 ppm from the center of each profile. Because ten m/z ratios delineate the full profile, six m/z ratios monitor 60% of an unbroadened profile's mass range. The mass calculated for each partial profile is the weighted average of the points between the vertical lines.
The percentage on the left of each plot indicates the %M+2 expected relative to M on the basis of the areas of the full profiles. When profile broadening occurs, decreased %M+2 values result when only partial profiles are monitored. In Figure 2e, only 58% of the M+2 profile's area would be monitored compared to 86% of the area under the M profile. The calculated %M+2 shown under the profile is only 6.5%, considerably less than the 9.6% based on a full M+2 profile to full M profile comparison. The apparent resolution (13,500) is based on the width of the profile at 5% of its maximum and provides a measure of profile broadening.
Identification of an Environmental Contaminant. An average exact mass of 423.2955 Da was obtained from triplicate determinations for the suspected molecular ion from a compound found in a sample from a Superfund site.3 The PGM was used to construct Table 2, which listed possible compositions for the ion, calculated mass defects, and calculated relative abundances for each composition at 21,400 resolution, the resolution at which partial profile data were acquired.
Table 2. Elemental Compositions and Discriminating Criteria among Them
a Rings and double bonds.
b Calculated mass defect.
c Based on partial profiles centered about the calculated mass for the composition.
d Based on partial profiles centered about the calculated mass of the hypothetical composition, plus or minus 1 mass increment at plus or minus 10% of resolution, and isotopic abundance error.
e An "X" indicates application of this criterion will reject this composition if the hypothetical composition is correct.
f Double maxima were observed in the profile.
A similar table was prepared at 10,100 resolution to evaluate the utility of data acquired under conditions providing 10-fold greater sensitivity. Allowing up to 35 C, 74 H, 6 O, 6 N, 4 P, 4 S, and 4 Si atoms and an error limit of 2.5 ppm (1.1 mDa), 12 possible compositions were calculated for 20,000 resolution. At 10,000 resolution, 13 additional compositions (25 in all) were possible for the average exact mass and error limits of 5 ppm (2.1 mDa). From left to right in Table 2, the 5 columns correspond to the exact mass criteria for the M+1 and M+2 profiles, the relative abundance criteria for the M+1 and M+2 profiles, and the shape of the M+2 profile. An "X" next to a value in the table indicates that a criterion has not been met and justifies rejection of the candidate composition.
Hypothetical Composition. To determine exact masses and relative abundances of the M+1 and M+2 partial profiles, a split SIR descriptor is used to monitor the M, M+1, and M+2 partial profiles. Good estimates of their exact masses are required in order to write the descriptor if all three apexes are to be centered in the observation windows. But without foreknowledge of the true composition, a hypothetical composition must be chosen in order to write the SIR descriptor.
The best choice for the hypothetical composition is made after examining the full M+2 profile. Using the average of the mass limits listed in Table 2 for the M+2 profile as the center mass in the SIR descriptor ensures that in most cases, all of the M+2 profile is observed. For broad profiles centered near a mass limit, most of the profile shape and the exact mass will be obtained when a part of one side is not monitored.
The composition with the most similar exact masses of the M and the M+2 profiles and the most similar shape of the M+2 profile was chosen as the hypothetical composition. If multiple compositions seemed possible, supplemental information could be considered, such as fragmentation and isotopic abundances in the low-resolution mass spectrum and any knowledge of the sample, such as the presence of other chemically related compounds that have been identified. If the apex of the M+1 profile was not monitored or was offset relative to the M and M+2 partial profiles, the split SIR descriptor should be rewritten on the basis of another hypothetical composition until the correct composition has been used.
M+2 Profiles from SIR Data. A full M+2 profile for the unidentified ion was plotted from data acquired at 20,400 resolution. For comparison, the PGM plotted the M+2 profiles for the candidate compositions in Table 2 at the same resolution using mass increments of 1 ppm. Each of the 12 profiles had one of the five shapes shown in Figure 2. The profile shapes in parts d and e of Figure 2 for compositions 9 and 12 were unique. Thus, the somewhat less well defined shape of the M+2 profile plotted from data using a mass increment of 5 ppm would identify either composition 9 or 12, or would reduce the list to no more than four compositions if another composition were correct. At 10,100 resolution, each shoulder or valley collapses into a single maximum, and the shape of the M+2 profile is not a discriminating criterion.
Figure 3. (a,b) Experimental M+2 profiles obtained with 20,400 and 37,600 resolution. (c,d) Calculated M+2 profiles for composition 12 in Table 2. 14,300 and 18,400 were apparent resolutions calculated from the plotted profile widths at 5% of the maxima. 19,400 and 40,000 were the resolutions entered into the model to provide calculated profiles.
Experimentally obtained M+2 profiles are shown in Figures 3a and 3b. M+2 profiles calculated by the PGM are shown in Figures 3c and 3d for composition 12 in Table 2. The M+2 profile in Figure 3a was obtained at 20,400 resolution. The resolution was measured from full profiles for the m/z 431 PFK ion acquired just before and after the M+2 profile was acquired during the same data acquisition. The valley depth observed was calculated by the PGM for a resolution of 19,400 (Figure 3b). The apparent resolution observed, 14,300, was calculated when a resolution of 22,400 was entered into the model. The average resolution of 20,900 from the PGM profiles agreed to within 2.5% of the experimental value determined. Because the analyte was a major component in the extract, sensitivity was adequate to utilize a resolution of 37,600 to obtain the profile in Figure 3d. The shape of the M+2 profile at either resolution afforded confident identification of C28H41NS as the correct composition.
Weighted averages of the three or four points nearest each apex of the two ions responsible for the double peaks in the M+2 profile were used to estimate their exact masses: 425.2928 and 425.2917 Da for the first ion and 425.3022 and 425.3024 Da for the second ion in Figure 3, parts a and b, respectively. These masses agreed with the theoretical values of 425.2918 Da (C28H41N34S+) and 425.3024 Da (l3C2C26H41NS+) to within 2.5 ppm and confirmed that the last composition was that of the M ion studied. Because the ion contained 1 N atom and had an odd mass, application of the nitrogen rule indicated the ion was an odd-electron ion, probably the molecular ion. Making this assumption a priori would eliminate about half of the possible compositions, those having non-integer numbers of rings and double bonds. In this example, even-electron ion fragments were also considered, in case the molecular ion was not observed. The full mass spectrum was consistent with the proposed structure, and the nature of the sample also suggested that the compound was a dioctylphenothiazine.
Partial Profiles from SIR Data. At 20,400 resolution, the correct composition was identified by the shape of the M+2 profile, but at 10,100 resolution, this criterion was not usable. To provide confirmation at the higher resolution and to determine if identification was possible at the lower resolution, data were acquired using split SIR descriptors based on the last composition as the hypothetical composition in order to apply the other four criteria. Five data acquisitions were made at each resolution. In Figure 4 are shown partial mass peak profiles for the M, M+1, M+2, and reference ions plotted from SIR data.
Figure 4. Partial mass peak profiles for M, M+1, and M+2 ions for an environmental contaminant and the reference ion plotted using SIR data acquired with a single SIR descriptor at (a) 10,100 and (b) 21,400 resolution.
At 10,100 resolution (Figure 4a), single, well-centered maxima were observed, which indicated the masses calculated by the PGM were correct, or very nearly correct. At 21,400 resolution (Figure 4b), the M+1 apex was consistent with the hypothetical composition, and the M+2 profile displayed the expected valley.
Application of Mass Criteria. In Table 3, the mass defects of the exact masses determined from M+1 and M+2 partial profiles are listed.
Table 3. Experimental Determinations of M+1 and M+2 Mass Defects and Relative Abundancesa
a %M+1 and %M+2, with averages and
2 sigma error limits at 10,100 and 21,400 resolution.
b Double maxima were observed in the profile.
At 21,400 resolution, the experimental exact mass of the M+1 partial profile differed from the calculated masses for compositions 1, 2, 3, and 6 in Table 2 by more than the allowed 2.5 ppm error limit of 1.1 mDa (1.7, 2.1, 1.4, and 1.8 mDa, respectively). At 10,000 resolution, a 10 ppm increment provided 10 points across a profile, and an error of one-half mass increment corresponded to 5 ppm (2.1 mDa). Ten of 25 possible compositions were eliminated on the basis of the exact mass of the M+1 partial profile.
At 21,400 resolution, double maxima were observed in the M+2 profile. The exact masses determined from the full M+2 profiles in Figures 3a and 3b for the two ions corresponding to the double maxima agreed to within 2.5 ppm with the theoretical values for compositions 9 and 12, the only compositions with double maxima in the M+2 profile. At 10,100 resolution, the M+2 profile of the unidentified ion displayed only one maximum, and simple comparisons between the calculated masses and the exact mass were possible. Mass differences between the calculated masses and the exact mass exceeded 5 ppm (2.1 mDa) for 15 of the 25 candidate compositions. Only nine of the 25 compositions remained viable after application of both mass related criteria.
Application of Relative Abundance Criteria. Also listed in Table 3 are the %M+1 and %M+2 values observed. The %M+1 and %M+2 values calculated in Table 2 for each composition are based on use of the calculated masses of the M profile and M+1 and M+2 partial profiles for each composition as the center masses in the SIR descriptor. The allowed ranges calculated for %M+1 and %M+2 listed in Table 2 and plotted in Figure 5 result from considering three types of error.
Figure 5. Estimated error ranges for (a) %M+1 and (b) %M+2 at 21,400 resolution. The innermost pair of hash marks delineate the range due to center mass offsets in the split SIR descriptor, profile broadening, mass shifts of 1 mass increment, and resolution variation of 10% (error type 1). Beyond these hash marks are small ranges due to the maximum possible isotopic abundance error (error type 2). The outermost pairs of hash marks delimit half of the maximum error range observed for the four standards for triplicate determinations (error type 3).
The first type is errors associated with monitoring partial profiles, which are estimated by the PGM. These errors arise from (1) using the calculated masses of the M profile and M+1 and M+2 partial profiles of the hypothetical composition as the center masses in the split SIR descriptor, which causes error if another composition is correct, (2) errors of up to 1 mass increment in the location of the partial profiles in the observation windows, and (3) variation of up to 10% in the resolution. These errors are zero when the correct composition is used as the hypothetical composition and when no peak broadening occurs. The second type of error is the maximum possible isotopic abundance error based on the number of atoms of each element in each composition. For example, 13C from petroleum or harvested plants has different abundances relative to 12C; variation of up to 0.03% was reported, in the IUPAC reference.4 The third type of error is due to instrumental variation during data acquisition and was estimated as the maximum error limits determined for four standards that displayed negligible broadening in the M+2 profile. The sum of these error ranges for each composition should include any experimentally determined value of %M+1 or %M+2 if the composition is correct.
In Figure 5a, the experimentally determined %M+1 value was included in the calculated error ranges for three of 12 possible compositions at 21,400 resolution. At 10,100, five of 25 possible compositions remained viable. All other compositions failed the %M+1 criterion.
In Figure 5b at 21,400 resolution, five compositions were excluded because their %M+2 ranges did not include the observed %M+2 value. At 10,100 resolution, 20 compositions were rejected. The data acquired at the higher resolution were less useful for the M+2 criterion, because profile broadening reduced the fraction of the M+1, and M+2 areas monitored and because the mass differences between the M, M+1 and M+2 profiles of the hypothetical and other compositions were larger relative to the width of the profiles. Wider %M+2 error ranges resulted.
In summary, at 21,400 resolution, 11 of 12 compositions were excluded by the exact mass and relative abundance criteria. At 10,100 resolution, 24 of 25 candidate compositions were excluded on the basis of one or more exact mass and relative abundance criteria. Thus, if the target compound had been a trace component, precluding use of 20,000 resolution, the correct composition would still have been determined. Different criteria excluded different compositions at the two resolutions.
Composition Identification of Another Environmental Contaminant. Three library matches of bis(bromophenyl)diazenes were found for the background subtracted mass spectrum of a minor component in an extract from another contaminated site. Consistent with this composition, C12H8N2Br2, the isotopic pattern for m/z 338, 340, and 342 indicated the ions contained two Br atoms. Triplicate determinations of the exact masses of the M and M+2 full profiles and of %M+1 and %M+2 based on partial profiles were made at 20,000 resolution. The diazene composition was rejected based on the average experimental mass of 339.8919 Da for the M+2 profile, because it was 34 ppm (11.5 mDa) less than the theoretical mass for the diazene ion. The calculated masses for five compositions were within 2.5 ppm of the average exact mass of the M ion. Four of the five compositions were rejected on the basis of the average %M+1, and C13H8O79Br2 was identified as correct. This composition, consistent with a dibromomethyldibenzofuran, was also consistent with the fragment ions observed in the full-scan mass spectrum.
Higher Mass Ions. For higher mass ions, many more compositions are possible. To estimate the bias, accuracy, and precision of exact mass and relative abundance measurements and to determine if unique compositions could be determined for ions with larger masses than in the previous examples, four standards with molecular masses between 536 and 766 Da were introduced individually into the ion source with a heated probe. For each standard, the known composition was chosen as the hypothetical composition. At a mass resolution of 20,000, 15 data acquisitions, each preceded by a mass calibration, were made for each standard to determine the exact mass of M from a full profile. Fifteen data acquisitions were also made using a split SIR descriptor to determine masses for the M+1 and M+2 partial profiles and their relative abundances. Both accuracy and precision were improved by averaging larger numbers of determinations.
Maximum error limits were chosen to ensure that no criterion falsely rejected the correct composition for any data set of three, seven, or 15 determinations. The maximum experimental errors for three, seven, and 15 determinations were less than 2.0, 1.5, and 1.0 ppm for the exact mass of M; 1.5, 1.0, and 0.5 ppm for the exact mass of M+1; and 2.5, 1.5, and 1.0 ppm for the exact mass of M+2, respectively. Because only four standards were studied this thoroughly, the largest errors (2.5, 1.5, and 1.0 ppm) were allowed for all exact masses. The error limit for %M+1 and %M+2 were 6%, 4%, and 2% of their values for three, seven, and 15 determinations.
The numbers of compositions possible based on the calculated mass of M with an error limit of 2.5 ppm for up to (MW/12) C, (2 x no. of C) H, 10 O, 10 N, 10 P, and 10 S atoms were 39, 24, 79, and 123 for the four standards having molecular ions with masses of 536.2621, 536.3151, 622.2791, and 766.3942 Da, respectively. For all four standards and all data sets of three, seven, or 15 successive determinations, the average masses and relative abundances passed all criteria for the correct composition. For the standard with a molecular ion weighing 536.3151 Da, one other composition passed all criteria for one of 13 triplicate determinations. For the compound with a molecular ion with a mass of 536.2621 Da, one or two other compositions were not rejected for five triplicate data sets. For both molecular ions, only the correct composition passed for all nine sets of seven successive determinations. For C36H3806N4 (molecular mass = 622.2791 Da), 1-4 other compositions passed all the criteria for all 13 data sets. For the nine sets of seven determinations, one other composition passed for one data set for the highest mass standard (molecular ion 766.3942 Da), 2-13 other compositions passed all criteria for each triplicate data set. One other composition passed all criteria for two of nine sets of seven determinations. In summary, the correct composition was identified for all four standards, because the smaller error limits associated with sets of seven and 15 determinations permitted rejection of more candidate compositions than for triplicate determinations. In practice, data are acquired until only one composition passes all five criteria based on the error limits associated with at least three, seven, or 15 determinations.
The high combined discriminating power of these five criteria permitted composition identification where mass resolutions greater than 1,000,000 would be required in order to separate the molecular ion profiles of the correct compositions from the molecular ion profiles for the compositions with the most similar mass of M.
Fragment Ions. Multiple compositions might remain after application of the five criteria for molecular ions with very large masses. Because interferences are usually excluded at high mass resolution, the PGM and split SIR descriptor can also be used to identify fragment ions (F) based on F+1 and F+2 profiles, when necessary. Fragment ions have smaller masses than molecular ions, and identification of their correct compositions is usually easier. Masses of neutral loss fragments can be determined as the difference between the exact masses of the molecular ion and fragment ions. A single composition for the molecular ion can then be identified as the sum of appropriate parts.
When partial coelution is a problem, fragment ions can be correlated with molecular ions by using a split SIR descriptor at high mass resolution to monitor the suspected fragment ions and the molecular ion.
Refinements for Routine Use. At least doubling the number of m/z channels available and halving the digital-to-analog converter (DAC) steps to 2.5 ppm in SIR descriptors would permit better delineation of profile shapes, especially at very high resolution. More subtle distinctions could then be made between profile shapes plotted from SIR data. Improved accuracy and precision for exact masses, relative abundances, and mass resolution might also result. If a 20-bit DAC were available to set m/z ratios for SIR descriptors, mass increments as low as 1 ppm would be feasible.
In addition, inclusion of programs in data system software to perform the experimental and modeling tasks described in reference 3 and herein would allow the routine use of MPPSIRD, with its advantages of greater sensitivity, scan speed, and mass resolution for exact mass determinations.
Many compounds present at trace levels in complex mixtures cannot be identified from mass spectral libraries due to insufficient spectral features or chemical interferences or because their mass spectra are not in the library. In this study, determination of the elemental compositions of the molecular or fragment ions of a compound permitted rejection of incorrect library matches with different compositions and limited the list of possible compounds to isomers having the composition identified.
Increases in sensitivity, measurement speed, and mass resolution were provided by MPPSIRD when a high-resolution mass spectrometer was used and enabled routine determination of exact masses at 20,000 resolution to within 2.5 ppm (triplicate determinations} for ions produced from analytes sampled as narrow chromatographic peaks. At 20,000 resolution, M, M+1, M+2, and fragment ions were resolvable from interferences with the same nominal masses. For low-mass ions, an exact mass determination from the full mass peak profile of the ion can, in most cases, provide a unique elemental composition. Criteria based on the exact masses and relative abundances of the M+1 and M+2 partial profiles and the shape of the M+2 profile were applied to identify the compositions of several larger ions with masses between 338 and 766 Da.
A profile generation model (PGM) was developed that enabled planning and interpretation of experiments. For any resolution, the PGM calculated masses and relative abundance ranges for M+1 and M+2 partial profiles and displayed the M+1 and M+2 profiles. The calculated masses of the M, M+1, and M+2 partial profiles for a hypothetical composition were used to write a split SIR descriptor, which monitored simultaneously the top portion of these three profiles. Exact masses and relative abundances of the M+1 and M+2 partial profiles were determined from the data obtained. Comparison of the experimental and calculated values for all possible compositions allowed identification of the correct composition by excluding all others.
Acknowledgement. The authors thank Dr. Robert W. Gerlach for his review of the statistically based procedures presented in this manuscript.
Wu, J.; Fannin, S. T.; Franklin, M. A; Molinski, T. F.; Lebrilla, C. B. Anal. Chem. 1995, 67, 3788-3792.
D'Agostino, P, A; Hancock, J. R; Provost, L. R; Semchuk, P. D.; Hodges, R S. Rapid Commun. Mass Spectrom. 1995, 9 (7), 597-603.
IUPAC. Pure Appl. Chem. 1991, 63, 991-1002.
U.S. Environmental Protection Agency. Statement of Work for Low/Medium Level Organics; Invitation for Bid Contract No. OLMO1.8, September 1991.
Grange, A H. Proceedings of the 40th ASMS Conference on Mass Spectrometry and Allied Topics; Washington, DC, May 31- June 5, 1992; pp 1145-1146.
Grange, A H.; Brumley W. C. Proceedings of the 40th ASMS Conference on Mass Spectrometry and Allied Topics Washington, DC, 1992; pp 1143-1144.
Dixon. W. J.; Massey, F. J., Jr.Introduction to Statistical Analysis, 3rd ed.; McGraw-Hill, Inc.: New York, 1969.