Ion Composition Elucidation (ICE)
Determination of Elemental Compositions by High Resolution Mass Spectrometry Without Mass Calibrants
Andrew H. Grange* and G. Wayne Sovocool
U.S. EPA, NERL, Environmental Sciences Division, Environmental Chemistry Branch, PO Box 93478, Las Vegas, NV 89193-3478
Perfluorokerosene can almost always be used as the mass calibrant for ions produced through electron impact ionization of compounds introduced into a mass spectrometer in the gas phase. Unfortunately, no completely universal calibrant is available for ions created by electrospray ionization (ESI) or atmospheric pressure chemical ionization (APCI) of analytes introduced into a mass spectrometer in the liquid phase. ESI and APCI generally provide less sensitivity than electron impact ionization of compounds introduced in the gas phase and a portion of the weaker total signal must arise from calibrant ions. Solvent conditions must be found that provide ions from both the calibrant and analytes, or an alternative flow to the ionization region must be provided for the calibrant solution. These problems were avoided by developing a methodology to determine elemental compositions of ions without using mass calibrants. The methodology utilizes the ability of double focusing mass spectrometers to accurately measure relative abundances of ions and exact mass differences between ions. This approach should simplify analyses of environmental samples that contain mixtures of compounds not amenable to gas chromatography or volatilization from a probe. From one to five steps were used. First, from mass peak profiles of the molecular ion or protonated molecular ion, [M]+, and the [M+1]+ and [M+2]+ ions, abundances for the [M+1]+ and [M+2]+ profiles relative to the [M]+ profile were determined. The [M+1]+ and [M+2]+ profiles resulted from the heavier isotopes of the elements in [M]+, and the profile abundances limited the number of possible elemental compositions for [M]+. Then, to determine if [M]+ contained N atoms, the [M+1]+ profile was observed with sufficiently high mass resolution to at least partially resolve the profiles of ions containing a 15N or a 13C atom. Next, for a prominent fragment ion, [F]+, relative abundances of the [F+1]+ and [F+2]+ profiles were also determined, and the [F+1]+ profile was inspected for a profile due to 15N atoms to provide a shorter list of possible compositions for [F]+. The lists for [M]+ and [F]+ were compared, and [M]+ compositions that could not produce any possible compositions of [F]+ were rejected, as were [F]+ compositions that could not arise from any possible composition of [M]+. Fourth, exact mass differences between ions were obtained from three mass peak profiles by referencing an unknown exact mass difference against a known exact mass difference. Exact mass differences between [M]+ and [F]+ ions provided compositions of neutral losses from [M]+. Only compositions of [M]+that could lose the observed neutral loss to provide possible compositions of [F]+ remained viable. Finally, if multiple compositions of [M]+ were still possible, profiles were obtained for [M]+ and two fragment ions resulting from known neutral losses using theoretical exact masses based on each [M]+ composition as the center masses of the three profiles. When the calibration mass option was not used in the multiple ion detector (MID) descriptor, three centered profiles were obtained only for the correct composition. This methodology was demonstrated for seven compounds with molecular weights between 159 and 318 Da. For the lowest-mass compound, only the first step was required to obtain the correct composition of [M]+ while, for the other compounds, two or more steps were needed.
The Environmental Sciences Division (ESD) of the EPA's National Exposure Research Laboratory is investigating the possibility that pharmaceuticals and their metabolites are entering rivers and lakes via sewage treatment effluent. A recent review indicated that some of these compounds are present and have been found in rivers.1 Compounds developed to induce physiological responses in animals could be endocrine disruptors2 or have other effects on the biota of the receiving waters. Identification of previously undetected compounds is necessary before their aquatic effects can be evaluated.
The ESD hopes to identify any such compounds that might be entering Lake Mead in the treated sewage discharge from Las Vegas. Mass spectra for many of these compounds will not be found in mass spectral libraries, especially compounds that are not amenable to gas phase introduction into the ion source of a mass spectrometer. About one half of the 100 drugs most prescribed in 19973 were not found in the NIST mass spectral library,4 nor are their metabolites likely to be included. The elements in these drugs are C, H, O, N, S, Cl, P, Br, F, or I. The ESD has developed analytical tools, Mass Peak Profiling from Selected Ion Recording Data (MPPSIRD) and a Profile Generation Model (PGM), to determine the elemental composition of ions containing these elements. These tools have been used to successfully identify components in municipal well water5 and to characterize a complex Superfund site sample.6
The elemental composition of an ion containing C, H, N, O, P or S with a mass up to 150 u can be determined from its exact mass if the error in the mass measurement is small. To determine elemental compositions of larger ions, Wu7 determined elemental compositions for fragment ions and identified the molecular ion or protonated molecular ion, [M]+, from its parts. As described in Reference 8, MPPSIRD and the PGM are used to determine compositions of [M]+ weighing up to 600 u that contain only these elements from the exact masses of the ions weighing 1 and 2 u more than [M]+ and from their relative abundances. These heavier ions arise from the isotopic abundances of 13C, 2H, 15N, 17O, 18O, 33S and 34S and are assuredly related to [M]+, unlike fragment ions observed with infusion of analytes. The exact masses of the three mass peak profiles are obtained using a high resolution mass spectrometer to monitor points across each profile using the selected ion recording (SIR) mode. This strategy provides the cycle speed needed to obtain exact masses for compounds that enter the mass spectrometer as chromatographic peaks and the sensitivity required to use a mass resolution of 20,000 to study minor components in mixtures with excellent selectivity.
To determine the exact mass of an ion using either full scanning or SIR, two calibration ions with masses that encompass the analyte ion's mass are required. When analytes are introduced into the ion source in the gas phase by gas chromatography (GC) or from a direct insertion probe, perfluorokerosene (PFK) is generally used as the calibrant. With electron impact (El) ionization, PFK provides positive ions every 12-14 u due to extensive fragmentation. The ions have a small negative mass defect and seldom interfere with ions produced from organic compounds. Chlorinated and brominated compounds usually have more negative mass defects and interferences are again uncommon for these compounds. PFK is volatile enough to provide ions between 51 u and over 1000 u when bled into the ion source from a heated reservoir and to be almost completely removed by a few hours of pumping. PFK is an ideal calibrant for most analytes when determining exact masses of positive ions formed by EI.
Electron impact ionizes both PFK and the analytes. Because the ions are formed in a vacuum of about 10-5 Torr, collisions between ions and neutral species are rare and the distribution of ions is little influenced by chemical equilibria. Both PFK and analytes provide measurable ion abundances regardless of each other's presence in the ion source. To determine masses with accuracy sufficient for providing elemental compositions of ions, both analyte and calibrant ions must be observed during the same data acquisition.
Unfortunately, most organic compounds are too polar, ionic, thermolabile or involatile to be introduced into a mass spectrometer in the gas phase. Many drugs are sold in ionic form to provide excellent solubility in body fluids. To analyze these compounds, liquid sample introduction and alternative ionization methods are necessary. With liquid sample introduction and ESI or APCI, chemical equilibria in the solvent and during desolvation determine the ion abundances measured by the mass spectrometer.9,10 Solution conditions must be optimized to provide ions from both analytes and calibrants. Alternatively, dual flows of calibrants and analytes can be used to ensure ions are observed from both compounds. In either case, the total signal is usually less than for El and is partitioned between the calibrant and analyte ions, which further decreases sensitivity for the analyte.
No ideal calibrant such as PFK is available for experiments using ESI-MS or APCI-MS.11 Finding appropriate calibrants for each analyte is difficult when exact masses need to be determined using these ionization methods. The effort and time needed to determine elemental compositions would be increased by this requirement. To avoid these difficulties, a methodology using MPPSIRD was developed to determine elemental compositions of ions without mass calibrants.
Double focusing mass spectrometers measure ion abundances as a function of mass-to-charge (m/z) ratio. A wide linear dynamic range enables determination of relative abundances between ions. With a constant magnet current, precisely set accelerating and electrostatic analyzer voltages permit determination of very accurate exact mass differences. Only relative abundances of ions, the exact mass differences between pairs of ions, and observation of certain mass peak profiles are required to determine elemental compositions without using mass calibrants. Because this idea is novel, if not radical, and at first might stretch credulity, each of these measurements will be discussed fully and then applied to seven compounds to determine the compositions of their molecular ions.
Instrumentation and conditions
All experiments were performed with a Finnigan MAT 900S-Trap hybrid mass spectrometer. A Digital Alpha minicomputer (Maynard, MA, USA) with a Digital UNIX 4.0 Operating system executed the Finnigan
ICLTM 10.6 and ICISTM 8.2.1 software that controls the MAT 900S (Bremen, Germany) double focusing mass spectrometer, which has an EB geometry. The Finnigan LCQ ion trap (San Jose, CA, USA) that follows the exit slit was controlled by Finnigan Navigator 1.1 software residing on a Gateway G6-200 personal computer (Sioux City, SD, USA) with a Windows NT 4.0 (Microsoft Corp., Bellevue, WA, USA) operating system.
A Finnigan MAT ESI source was used. The heated capillary was maintained at 230 oC, the flow was 2 to 8 µ/min, the voltage applied to the needle tip was 2.8-4.5 kV, the observed current was 1.8-3.2 µA, and no sheath gas was used during data acquisition. These variables were adjusted to provide maximum signal with stable operation prior to groups of data acquisitions. Different voltages were applied to the four elements of the ESI source depicted in Figure 1: the heated capillary, tube lens, skimmer and octapole lens.
Figure 1. Adaptation of a schematic drawing of the Finnigan MAT 900S electrospray ion source kindly provided by Dr. Helmut Muenster.
When a collisionally induced dissociation (CID) voltage was specified, it was added to the voltages of the first two elements by enabling the CID option.
A Harvard Apparatus 22 syringe pump (South Natick, MA, USA) controlled by the MAT 900S data system delivered the analyte solution from a 2.5-mL Hamilton gas-tight syringe (Reno, NV, USA).
For one set of experiments, a Finnigan MAT microelectrospray source was used; 5 kV was applied to the capillary tip, and the analyte solution contained 2% acetic acid. Sample was delivered pneumatically through a micro-capillary from a reservoir with a pressure of 105 Pa (1 Barr) of N2 [Air Products, Phoenix, AZ, USA (99.998% excluding Ar)], which provided a flow of about 200 nL/min.
The mass spectrometer was initially calibrated in the full scan mode for m/z ratios of 20-800 u against PFK-H (PCR Research Chemicals, Gainesville, FL) using a 30 s/decade scan speed and 10 000 resolution. The ESI source was then installed and full scans were acquired with ESI using the same scan speed and resolution. The secondary electron multiplier voltage was between 1.8 and 2.5 kV, depending on the sensitivity required for each experiment.
Mass peak profiling from selected ion recording data (MPPSIRD)
MPPSIRD includes the term "selected ion recording" used by Micromass (Danvers, MA) because the technique was developed using a VG70-250SE mass spectrometer. For these experiments, MPPSIRD procedures were modified for the Finnigan MAT 900S. Finnigan prefers the term "multiple ion detection". However, we prefer to maintain the continuity of the acronym, MPPSIRD, in the literature dating from 1992,12 rather than use MPPMIDD as suggested by the ASMS nomenclature committee.13 Data was in fact "recorded" as it was "detected". SIR and MID descriptors perform the same function.
MPPSIRD was used to acquire all relative abundance and exact mass difference data used to determine elemental compositions and is discussed fully in Reference 14. Using infusion of analytes, each MID descriptor recorded ion signals for m/z ratios across two or three mass peak profiles. For 5 s before and 5 s after a 100 s integration time, the polarity of the skimmer voltage in the ESI source was reversed to deflect the ion beam and reduce the signal to zero. The ion signal for each m/z ratio was integrated across the 100 s window by the ICISTM software and plotted to provide the mass peak profiles shown in Figures 2-5. Each point is the integrated signal at that m/z ratio. The area under a profile is the sum of points used to plot it, and relative abundances are ratios of the profile areas. The exact mass of each profile is the weighted average of the points used to plot the profile, unless the profile is off center. Then the same number of points on each side of the maximum are used for the weighted average. The center mass of the first profile was chosen as the lock mass, and 20 profile widths were scanned for the first cycle to ensure the largest profile in this mass range was selected for lock-on.
Profiles at a single nominal mass were examined in the tune view provided by the ICLTM software. The tune view displayed and provided control of the voltages applied to the ion source lenses and the secondary electron multiplier. It provided an auto tune option to maximize the signal that was used routinely. The tune view also displayed a "full scan mass spectrum" over a narrow mass range called the Tune Sweep specified by the user. The Sweep Speed for this display was set to its maximum, 10 s/sweep, to maximize the signal-to-noise (S/N) ratio and provide the smoothest possible profiles for observation of analyte and interfering profiles. These displays were hard copied.
Chemicals and solvents
Seven compounds were studied: 2-phenylquinoline (99%), 1-aminopyrene (97%), chlorpromazine hydrochloride (98%), 2-hydroxy-4-methylquinoline (97%), and cetyldimethylethylammonium bromide (85% - remainder stearyl compound), purchased from Aldrich, Milwaukee, WI, USA, and benzidine and 3,3'-dimethoxybenzidine purchased from Chem Service, West Chester, PA, USA. All contained either an amino group or a heterocyclic N atom to ensure cations were formed.
Each neutral compound (1 mg) was dissolved in separate portions of 99 mL of 4:6 acetonitrile/deionized water (J.T. Baker, Phillipsburg, NJ, USA, HPLC solvent); deionized water was obtained from a Barnsted Nanopure System (DuBuque, IA, USA). For the salts, 1 mg each of chlorpromazine hydrochloride and cetyldimethylethylammonium bromide was dissolved in separate portions of 99 mL of 1:1 methanol/deionized water (Burdick and Jackson, Muskegon, MI, USA, high purity solvent, >99.9%). After dissolved air was removed from the solvents by sparging with He to avoid oxidation of analytes, the compounds were dissolved, and the pH was decreased to ensure analyte cationization and a stable ESI current by adding 1 mL of glacial acetic acid (Aldrich, Milwaukee, WI, USA, 99.7%) to provide 1% acetic acid in the solutions that were electrosprayed.
Software on both data systems was used to prepare MID descriptors, acquire data, plot profiles, and calculate relative abundances and exact masses. DOSTM based programs, Lotus 123TM v2.2, WordPerfectTM 5.1, and FtpTM, were run sequentially by DOS batch files on the personal computer. Seamless transition between these programs (no mouse points and clicks), few computer lock-ups, and no need for annual retraining for each new version of each software package were important advantages of using DOS based programs. Automated procedures to enable MPPSIRD on the new instrument were prepared in only two weeks.
Lotus procedures wrote executable Instrument Control Language (ICL) files that were automatically transferred to the Finnigan MAT data system using Ftp. These files prepared the MID descriptors, controlled the skimmer voltage in the ESI source, plotted and integrated the ion signals, and created a summary file of the m/z ratios and integrated signals. The summary file was transferred back to the PC using Ftp, and another Lotus procedure plotted the profiles, which were printed out by a macro in WordPerfect. Before each data acquisition, the user inputs only the center masses of each profile, the resolution used to calculate the mass increments, and the data file name.
To provide thick lines and large fonts, figures showing mass peak profiles in this paper were replotted using PSI-Plot 3.0 (Poly Software International, Salt Lake City, UT, USA).
RESULTS AND DISCUSSION
Information lost when mass calibrants are not available
Table 1 lists the compositions possible based on the exact mass of the molecular ion of 2-phenylquinoline and the mass error limits (+ to -6 ppm) for a single determination made with 10,000 resolution using a VG70-250SE mass spectrometer.8
Table 1. Possible compositions for an ion with m/z 205.08915 u and quantities used to distinguish among them
aRings and double bonds using RDB = x - y/2 +
z/2 + 1, where x is the number of C or Si atoms, y is the number of H
or halogen atoms, and z is the number of NIII or PIII
atoms20 and assuming all
N and P could have valences of 5, which increases RDB by 1 for each N
bCalculated mass defect.
cBased on partial profiles centered about the calculated mass for the composition.
dBased on partial profiles centered about the calculated mass of the hypothetical composition, + to -1 mass increment at + to -10% of resolution, and isotopic abundance. To shorten the table, the ion was assumed to contain at least 1/3 carbon by weight. Compositions with fewer than 5 C atoms are eliminated by the %M+1 criterion. An "X" next to an entry indicates failure of the criterion for that composition.
This hypothetical case presupposes EI ionization and the availability of two calibration ions supplied by PFK. In the table are listed calculated values for the mass defects of the [M+1]+ and [M+2]+ partial profiles and their abundances relative to the [M]+ partial profile for each possible composition. An "X" next to a value indicates that it is sufficiently different from the value for 2-phenylquinoline that the composition could be eliminated once data were acquired. Measurement of two additional exact masses and two relative abundances would provide four additional criteria to reject all but the correct elemental composition.
Using ESI, soft ionization produces a protonated molecule (m/z 206) with a non-integer number of rings and double bonds. Without mass calibrants, the mass defects in Table 1 cannot be determined directly. However, mass calibrants are not needed to determine relative abundances, and Table 2, which incorporates much narrower ranges for %M+1 and %M+2 than Table 1, reduces further the number of possible compositions from hundreds to only three.
Table 2. Experimentally determined relative abundances
Those compositions with a nominal mass of 206, but with %M+1 and %M+2 values outside the error range for these values for 2-phenylquinoline, were not listed. Multiple compositions remain possible, because the mass defects in Table 1 are not available to reject the two incorrect entries. For most organic compounds, the dominant contribution to %M+1 is the number of C atoms in an ion times 1.11%. Consequently, the %M+1 values in Table 2 correspond to a narrow range (up to three) in the number of C atoms in the compositions (e.g. C15-C18 for chlorpromazine).
Each atom of 34S, 37Cl and 8lBr contributes 4.43, 32.0 and 97.3%, respectively, to %M+2. For the compounds in Table 2, the %M+2 established the number of these atoms that were present. When M+2 was less than 4.43%, no 34S, 37Cl and 81Br atoms were present, and these elements were not considered when determining possible compositions.
Maximizing accuracy of relative abundance determinations
Using MPPSIRD for any measurement provides a 100-fold increase in sensitivity relative to full scan modes. In addition, the relative abundances for the [M+1]+ and [M+2]+ profiles associated with the m/z 319 ion from chlorpromazine had better precision and accuracy. For nine magnetic full scans, %M+1 varied from 0-43% and %M+2 from 5-59%. These ranges were much larger than those observed using MPPSIRD (19.6-20.6% and 37.6-39.1% for 15 determinations). The full scan averages were 12.8 and 36.4%, in poor agreement with 20.7 and 38.5% as calculated by the PGM, while MPPSIRD provided average %M+1 and %M+2 values of 20.2 and 38.0%.
To reject as many additional compositions as possible for Table 2, MPPSIRD procedures were modified to reduce the error limits in the determination of relative abundances.
Elimination of relative abundance errors associated with partial profiles
In earlier work, using partial profiles to simultaneously determine the exact masses and relative abundances of the [M+1]+ and [M+2]+ partial profiles halved the number of data acquisitions that would be necessary if two pairs of full profiles were monitored.
Unfortunately, as discussed in detail in Reference 8, use of partial profiles introduces three errors largely responsible for the largest %M+1 and %M+2 error ranges in parentheses in Table 1: mass offset error, mass calibration error and resolution error. If the center masses for [M+1]+ and [M+2]+ profiles are significantly different among compositions, then using the center masses of the profiles for the hypothetical composition in the SIR descriptors will result in monitoring less of the area under the profiles if the hypothetical composition is incorrect. Mass offset error is the greatest contributor to the widest ranges observed in Table 1. The second two errors are important if the [M+1]+ or, more often, the [M+2]+ profile is broadened from a Gaussian shape by partial resolution of the profiles from the multiple ions that contribute to its total area. A mass calibration error causes a greater proportion of a Gaussian profile to be missed than for a broadened profile. Increased resolution causes greater broadening and a smaller portion of broadened profiles to be monitored. Used in conjunction with exact masses for the [M]+, [M+1]+, and [M+2]+ profiles, these errors were acceptable, since unique elemental compositions were determinable.
In this work, these errors were eliminated by monitoring full [M]+, [M+1]+, and [M+2]+ profiles. Each MID descriptor for the Finnigan MAT 900S permitted use of 31 m/z ratios. Because no calibrants were used, all were available to monitor each pair of full profiles. The [M]+ profile was monitored by 13 m/z ratios, while [M+1]+ or [M+2]+ profiles were monitored by 18 m/z ratios. At 10,000 resolution an unbroadened profile is about 100 ppm wide; a mass increment of 10 ppm ensured the entire [M]+ profile was monitored by 13 points. The seventh m/z ratio was specified as the lock mass, which ensured that the [M]+ profile was always centered in the mass range monitored and displayed. The entire [M+1]+ and [M+2]+ profiles, whether Gaussian or somewhat broadened, fell within the mass range (170 ppm) provided by 18 m/z ratios. Example profiles are displayed in Figures 2(a) and 2(b).
Figure 2. (a) [M]+
and [M+1]+ mass peak profiles, (b) [M]+ and [M+2]+
(c) [M-2H]+, [M-H], and [M]+ profiles for benzidene. [M]+ represents the protonated molecular ion. Corrected relative abundances (CRA), fH and f2H are discussed in the text under "H atom losses".
Although twice as many data acquisitions were required than when partial profiles were monitored for chromatographic peaks as they eluted, less time was required to determine relative abundances because infusion of the analytes was used.
Compensation for Data System Offset
An offset (1 count) was required by the data system to acquire MID data without truncating the lower portion of profiles. The offset disregarded the same small area under each [M]+, [M+1]+, or [M+2]+ profile. To estimate this area, triplicate acquisitions of [M]+ and [M+1]+ profile data were made with ion abundances of 10,000 and 100,000 uV for the maximum in the [M+1]+ profile. A constant was added to the areas of both profiles for each ion abundance level to make the average %M+1 values equal. The constant corresponded to the area truncated by the offset. The error in %M+1 compensated by the constant was 0.10% for the smaller ion abundance and 0.01% for the larger one, while the corrected %M+1 determined was 5.90%. The error was small, but was easily compensated by adding this constant to all profile areas acquired after this experiment.
The constant was obtained for an integration time of 100 s using 10 points to plot the [M+1]+ profile. It would be reduced proportionally for shorter integration times. To account for profile broadening, the constant was multiplied by n/10, where n is the number of points at least 2% as large as the maximum of the profile.
H atom losses
Four of the seven compounds studied, benzidine, 3,3',-dimethoxybenzidine, chlorpromazine and 1-aminopyrene, displayed significant [M-H]+ profiles (3-20% relative abundance) or [M-2H]+ profiles (up to 2.6% relative abundance) corresponding to loss of one or two H atoms from the molecular ion. Equations 1-5 define the abundances of the observed [M-2H]+, [M-H]+, [M]+, [M+1]+, and [M+2]+ profiles in terms of the abundances that would be observed in the absence of H atom losses.
AM-2H = f2HAMo
AM-H = fHAMo + f2HAM+lo
AM = (1 - fH - f2H)AMo + fHAM+lo + f2HAM+2o
AM+l = (1 - fH - f2H)AM+lo + fHAM+2o
AM+2 = (1 - fH - f2H)AM+2o
where AM-2H, M-H, M, M+1, M+2 are the observed abundances of the [M-2H]+, [M-H]+, [M]+, [M+1]+, and [M+2]+ profiles; AMo, M+1o,M+2o are the abundances of the [M]+ , [M+1]+ and [M+2]+ profiles in the absence of H atom losses; and fH and f2H are the fractions of [M]+ ions that lose one and two H atoms, respectively. Rearranging Equations 1, 5, and 2 and substituting Equation 6 into 8.
AMo = AM-2H / f2H
AM+2o = AM+2 / (1 - fH - f2H)
AM+lo = (AM-H - fHAMo) / f2H = (AM-H - fHAM-2H/f2H) / f2H
Substituting the right sides of Equations 6, 7 and 8 into Equations 3 and 4:
1 = AM = (1 - fH - f2H)AM-2H / f2H + fH
(AM-H - fHAM-2H / f2H) / f2H + f2HAM+2 / (1 - fH - f2H)
AM+l = (1 - fH - f2H)(AM-H - fHAM-2H/f2H) / f2H + fHAM+2 / (1 - fH - f2H)
Starting with the observed AM-H and AM-2H as estimates of fH and f2H, respectively, fH and f2H are varied in Equations 9 and 10, until the expressions on the right of the equations provide the observed AM and AM+1 values. Equations 6-8 are then used to determine AMo, AM+1 and AM+2, from which corrected values of %M+1 and %M+2 are calculated.
For benzidine, the corrections were +0.11% for %M+1 and +0.01% for %M+2 when the observed M-H and M-2H abundances were 8.6 and 1.3%, respectively. For chlorpromazine, loss of an H atom from the more abundant [M+2]+ ions inflated the observed %M+1. The corrections were -1.94 and +0.46% for %M+1 and %M+2, for observed M-H and M-2H abundances of 5.6 and 0.1 %, respectively.
The [M-H]+ and [M-2H]+ abundances varied as data sets of 15 relative abundance determinations were made. For example, for benzidine, the observed %M-H varied between 5.1 and 8.6%, and %M-2H varied between 1.3 and 2.6%. For this reason, a data file consisted of ion abundances acquired for 100 s each for the full [M]+ and [M+1]+ profiles, the full [M]+, and [M+2]+ profiles, and for [M -2H]+, [M -H]+, and [M]+ profiles plotted using 10 m/z ratios across each profile. Figures 2(a)-(c) show a typical print out.
Isotopic abundance errors
The uncertainties in the isotopic abundances for compounds of unknown origin contribute to the overall error range. Isotopic abundance errors from Reference 15 were used in the PGM. The PGM calculates the maximum isotopic abundance error for each composition and adds it to the precision error described below.
The relatively large uncertainty in the isotopic abundance of 13C (+ to -0.03%) arises from the different 13C/12C ratios associated with different sources of carbon. Reference 16 lists eight carbon sources with ranges of the differences in this ratio between the sources. Because no prior knowledge of the source of an ion's carbon is usually available, the abundance of 13C and its error range is not limited to that associated with a particular source. For most organic compounds, this isotopic abundance error is the largest. For example, 20 C atoms in an ion contribute + to -0.6% to the isotopic abundance error.
Without ever-present calibration ions, chromatographic peaks cannot be studied, since lock-on by the data system cannot be achieved prior to elution of each peak. Hence, infusion was used to ensure that the ions of interest were present when data acquisition started. Although chromatographic peaks obtained during GC are typically 10 s wide at their base, integration times can be longer with infusion of analytes. To provide excellent precision, an integration time of 100 s was chosen. Data sets of 15 runs for 2-phenylquinoline, benzidine, 3,3',-dimethoxybenzidine and chlorpromazine provided average standard deviations in %M+1 and %M+2 for subsets of three successive runs. The average standard deviations for triplicate runs for all four standards were less than 1.3% of the observed %M+1 and less then 1.8% of the observed %M+2. Three maximum standard deviations, 3.9% of %M+1 and 5.4% of %M+2 were incorporated into the model as the maximum error expected due to instrumental variables. These are the default values used by the modified PGM.
However, the user can enter other numbers. For the m/z 128 fragment ion for 2-phenylquinoline, the [F]+ , [F+1]+, and [F+2]+ profiles were acquired with 16,300 resolution due to the presence of unresolved interferences at 10,000 resolution. The [F+2]+ profile was jagged due to a very low signal and provided a %F+2 of 0.48%. The user entered 0.1% in place of 0.026% to allow for increased error that could result from a poor S/N ratio in the integrated signals used to plot the [F+2]+ profile. An increased precision error limit of + to -0.1% was also used for the average %F+2 (0.15%) obtained from three jagged [F+2]+ profiles for the m/z 86 fragment ion from chlorpromazine.
Use of fragment ions to determine exact masses and elemental compositions of neutral losses
Fragment ions formed from the molecular ion, which initially dominated the full scan spectrum, increased in abundance as the CID voltages on the heated capillary and tube lens were increased by 20, 40, 60, 80 and l00 V. A few fragment ions that were less abundant in the full scan spectra were observed by passing only the molecular ion through the EB portion of the instrument to the ion trap where fragmentation was induced. These ions were also used to determine exact mass differences with the MAT 900S portion of the hybrid instrument.
The exact mass difference between a fragment ion and the molecular ion, provides the corresponding composite neutral loss between the ions. Nine magnetic full scans for chlorpromazine using 10,000 resolution provided average mass differences between m/z 319 and 58, m/z 319 and 86, and m/z 86 and 58 of 261.0496, 233.0186, and 28.0311 Da, respectively. The corresponding errors in the exact masses for the composite neutral losses were 45, 52, and -7 ppm. The range of values for these differences among the nine full scans was 27, 34, and 114 ppm. Mass errors and standard deviations associated with MPPSIRD were several times less, because MPPSIRD employed 1 s cycles, which limited drift during each cycle. In addition, lock-on to the lowest-mass profile (recalibration using the lock mass) was repeated for each cycle. Drift was compensated before each cycle was recorded, and the mass resolution was preserved when profiles were plotted from data acquired over long integration times.17 Therefore, MPPSIRD was used to determine exact mass differences.
For electric scans, the magnet current is held constant while the accelerating potential (Va) and electrostatic sector voltage are decreased with a constant ratio between them. The full voltages correspond to one mass unit less than the smallest mass in the MID descriptor. The equation relating the magnetic field strength (B) and Va to the m/z ratio is:18 m/z = 4.82 x 10-5B2r2/Va, where the units of B are gauss and r is the radius of the flight path within the magnetic field in cm. With B constant, m is inversely proportional to Va. For two ions with the same charge, m2/ml = Vl/V2.
Figure 3(a) shows three profiles for two fragment ions and the molecular ion with very different center masses that were plotted from data obtained with a single MID descriptor using 10,000 resolution.
Figure 3. Mass peak profiles for two fragment ions, [C12H7]+ and [C16H8]+ and the molecular ion from 1-aminopyrene, [C16H12N]+, (a) using the exact masses for the correct compositions and (b) the exact masses corresponding to replacement of a 1 C atom by 12 H atoms for each of the three compositions. The second and third profiles were missed in (b). The apparent masses (AM) are the weighted averages of an equal number of points on each side of the profile's maximum; the hypothetical masses are those entered by the user as the center masses of the profiles; and the estimated exact masses are the apparent masses corrected for the error determined from the errors between the apparent and hypothetical masses of the other two profiles. See "Calculation of mass errors and estimated exact masses" for details.
Ten m/z ratios were sufficient to monitor these profiles, since profile broadening only occurs when heavier isotopes are present. The hypothetical mass (HM) was entered as the center mass of the first profile in the MID descriptor and into the tune view, after which the voltage was adjusted by the tune view software to center the profile at this m/z ratio. Because the center mass of the first profile is always the lock mass, it is always centered in the mass range monitored. The first and third profiles were then used to calibrate the mass scale between them. Because the center masses entered into the MID descriptor for all three profiles were correct, the three profiles were centered in the monitored mass ranges. In Figure 3(b), 0.0939 u, the mass difference between 12 H atoms and a C atom, was added to all three hypothetical masses. Although the exact mass differences between each pair of profiles were still correct, the error in the first mass caused the wrong mass ranges to be monitored for the second and third profiles. The ratio m2/m1 decreased from 1.344 to 1.342 or by 152 ppm. The ratio m3/m1 decreased by 191 ppm. Consequently, the accelerating voltages set by the MID software to monitor the second and third profiles were displaced by 152 and 191 ppm from the values required to monitor the m/z ratios of these profiles. The profiles were missed because the ten m/z ratios monitored for each profile span a mass range of 90 ppm at 10,000 resolution. The centered profiles in Figure 3(a) indicate that the correct elemental composition for the molecular ion at m/z 218 and the correct neutral losses were used to establish the exact mass differences between the molecular ion and the two fragment ions.
Determination of small exact mass differences without knowing the composition of [M]+
When the calibration mass in the MID descriptor was specified by the user as the center mass of the third profile and the wrong composition was used for [M]+, accurate exact masses for small neutral losses were still determined if the error in the m3/m1 ratio was less than 70 ppm. In these cases, half or more of the third profile was initially monitored by the m/z ratios used, and the MID software adjusted a calibration factor to center the third profile in the mass range monitored. The mass difference between the first and third profiles was then correct, and exact mass differences were accurately determined. When the calibration mass option was not used for the center mass of the third profile, the third profile was not shifted and appeared off center. Examples are shown in Figure 4(a) and 4(b) for the same three hypothetical center masses with and without specifying a calibration mass, respectively.
Figure 4. Mass peak profiles for [M]+ from 3,3'-dimethoxybenzidine and fragment ions resulting from neutral losses of CH3 and C2H6. The composition of [M]+ was incorrectly assumed to be composition 13 in Table 2, [C14H33ON2]+. In (a) the calibration mass option was on and in (b) it was off. The observed displacements of 4 and more than 5 mass increments in (b) were consistent with the expected displacements of the profile maxima of 40 ppm (4 mass increments) and 74 ppm (7.4 mass increments).
When one of the two mass differences was known between profiles 1 and 2 or between profiles 2 and 3, the three-step process illustrated in Figure 5 was used to determine the other.
Figure 5. Mass peak profiles for [C13H9NCl]+ and [C17H20N2SCl]+ from chlorpromazine with (a) 2400 resolution and mass increments of 333 ppm, (b) 10,000 resolution and mass increments of 20 ppm and (c) 10,000 resolution and mass increments of 10 ppm.
First, a mass resolution of 1200 was used with mass increments of 333 ppm to provide mass ranges of 3000 ppm for each profile. Exact masses entered for two of the three profiles provided the correct exact mass difference between the two profiles, while the exact mass entered for the third profile was a coarse estimate based on the nominal mass obtained from a full scan spectrum and the value of the first two decimal places of the mass defect entered for an adjacent profile. As illustrated by Figure 5(a), two or three m/z ratios defined each profile, and the automated plotting procedures provided estimates of the exact masses for all three profiles based on the assumption that the exact masses used for the other two profiles were correct. Second, from these, the exact mass estimate for the unknown profile was used as its center mass with 10,000 resolution and 20 ppm mass increments. As in Figure 5(b), profiles defined by five m/z ratios were obtained and a more accurate exact mass estimate was provided for the unknown profile. Finally, when the ratio of the largest mass to the smallest mass was less than 1.5, this better exact mass estimate for the unknown profile was used at 10,000 resolution with 10 ppm mass increments to provide a profile defined by ten m/z ratios as in Figure 5(c). For largest/smallest mass ratios between 1.5 and 2.0, 15 ppm mass increments were used to ensure that the third profile was not missed. In only two instances were ratios greater than 2.0 (2.04 and 2.15) used, since the sensitivity for the largest mass ion often became inadequate for larger ratios.
To test the ability of MPPSIRD to determine exact masses near the maximum largest/smallest mass ratio normally used, 15 data acquisitions were made for the m/z 101.03913, 128.05002 and 206.09697 ions from 2-phenyl-quinoline (ratio of 2.04) with 10,000 resolution and 15 ppm mass increments. Assuming each profile's exact mass was to be determined from the other two, the average errors were 0.3, -0.3 and 1.0 ppm, and the largest errors obtained were 2.8, -2.1 and 8.1 ppm, respectively. This compound was used because it did not introduce errors from [M+1]+ or [F+1]+ ions that have lost an H atom. Errors in determining exact masses of this small magnitude permitted confident selection of neutral loss candidates for the compounds studied based on single exact mass difference determinations.
Calculation of mass errors and estimated exact masses
In Figures 3-5 are listed apparent masses (AM), determined as weighted averages of points used to plot the profiles, and hypothetical masses (HM), which are the center masses used in the MID descriptor. An estimated exact mass (EM) was calculated for each profile using the other two profiles for calibration.
When HM1 and HM3 are known exact masses for the first and third profiles, HM2 is the initial estimated exact mass for the second profile. The estimated exact mass for the second profile determined from the three profiles, EM2, is then:
EM2 = AM2 - Err2
where Err2 is the error of AM2 relative
to the exact mass of the second profile.
The errors in the apparent masses for the first and third profiles are:
Err1 = AM1 - HM1
Err3 = AM3 - HM3
Err2 is estimated by linear interpolation between Err1 and Err3.
Err2 = (Err3 -Err1) / (HM3 -HM1) x (HM2 -HM1) + Err1
Substituting from Equations 11-13 into Equation 14:
EM2 = AM2 - (((AM3 - HM3) - (AM1 - HM1)) / (HM3 - HM1)
x (HM2 - HM1) + (AM1 - HM1))
The disagreement in ppm between the estimated exact mass and hypothetical mass for the second profile is then (EM2 -HM2) / EM2 x 106.
- EM1 = AM1 + (((AM3 - HM3)
- (AM2 - HM2)) / (HM3 - HM2)
x (HM2 - HM1) - (AM2 - HM2))
- EM3 = AM3 - (((AM2 - HM2)
- (AM1 - HM1)) / (HM2 - HM1)
x (HM3 - HM2) + (AM2 - HM2))
Equations (15-17) were used to provide estimated exact masses for all three profiles when they were printed out so that the operator was not required to specify which hypothetical mass was merely an estimate. For example, in Figure 5, the estimated exact masses determined for the first profile from Figures 5(a) and 5(b) were used in the MID descriptor as the center masses for the first profile to acquire the data to plot the profiles in Figures 5(b) and 5(c), respectively.
The errors in Figure 5 are the differences between the hypothetical mass entered into the MID descriptor and the estimated exact mass determined using the other two profiles. For the first profile in Figure 5(c), the error between the estimated exact mass, 214.04238, and the calculated exact mass, 214.04235, was +0.1 ppm.
Exact mass differences corresponding to unique neutral losses were determined successfully using each of the profiles as the unknown profile and the successive approximation strategy described above.
Sequential determination of neutral losses
Generally, exact mass differences were determined sequentially starting with the molecular ion and moving toward lower mass fragments or by starting with a low-mass fragment ion whose composition had been determined from relative abundances and moving toward higher masses. For example, when moving toward higher masses, the mass difference between the first and second profiles was known and the exact mass difference between the second and third profiles was determined. Then, the second and third profiles became the first and second profiles for the next higher exact mass determination. The known exact mass difference between the two profiles was the mass difference between the neutral losses from the molecular ion responsible for the two profiles. The exact masses and mass errors determined for the neutral losses from the molecular ions formed from the compounds studied are listed in Table 3.
Table 3. Elemental compositions of composite neutral losses from seven compounds
Errors were eliminated after each exact mass determination by using the theoretical exact masses of the fragment ions in the MID descriptor prepared for the next exact mass determination.
Whether sequential exact mass determinations were made in ascending or descending order, the first known mass difference was between [F]+ and [F+1]+ ions. When a low mass ion of known composition was used, then the exact masses of [F]+ and [F+1]+ were calculated using the PGM. For the fragment ion resulting from the smallest neutral loss from the molecular ion, the mass difference was estimated to be that of a 13C atom, 1.00336 u. This assumption was nearly correct for the compounds studied because the relative abundance of the [F+1]+ profile was mostly due to the presence of a 13C atom. For benzidine, loss of NH3 (17 mass units) yielded a C12H10N+ ion. The possibility that anyone of the 12 carbon atoms could be the 13C isotope provided 13.4% of the total %F+1 of 14.3%, and the resulting exact mass error for the F+1 profile was 0.00028 u (1.5 ppm). This error was insignificant relative to the mass differences between the possible neutral losses of NH3, CH3+2H and OH, which have masses of 17.02655, 17.03913, and 17.00274, respectively.
Use of non-analyte ions when few fragment ions were observed
For the quaternary ammonium ion from cetyldimethylethylammonium bromide, no fragment ions were observed across a mass range for which the largest/smallest mass ratio was greater than 2. The range was 74 to 298 u. In this case, ions due to impurities or solvents within this mass range at m/z 99, 159, 183, and 219 were used to sequentially determine exact mass differences. Non-analyte ions provide compositions inconsistent with possible analyte molecular ions, appear in multiple analyte solutions, or do not increase in abundance as the CID voltage is increased. The sum of the intervening mass differences provided the total mass difference between the analyte ions at m/z 74 and 298 u. As expected, a greater mass error resulted from the sum of five measurements (-18.6 ppm for M), but the error was still small enough to permit unambiguous assignment of the neutral loss between the molecular ion and the m/z 74 ion.
Examining profiles at a single nominal mass
Determining the presence of N atoms by examining [M+1]+ profiles
The mass difference between 13C and 12C is 1.00336 u. The mass differences between 15N and 14N, 33S and 32S, 29Si and 28Si, 17O and 16O, and 2H and 1H differ from 1.00336 u by -0.00632, -0.00396, -0.00379, +0.00086, and +0.00292 u, respectively. For small ions, the large negative difference for 15N and 14N relative to 13C and 12C permits resolution of the profiles for ions containing a 15N atom from those containing a 13C atom.
Mass resolution is defined as M/deltaM where M is the average mass of two profiles of equal height with a 10% valley between them.19 Figure 6(a) shows a profile calculated by the PGM and Figure 6(b) the same profile in the tune display for the [F+1]+ ions from C3H8N+ formed from chlorpromazine with a mass resolution of 12,700.
Figure 6. [F+1]+ profiles at m/z 59 (a) calculated by the PGM and (b) observed in the tune view for the C3H8N+ fragment ion from chlorpromazine with a mass resolution of 12,700. (c) The calculated [M+1]+ profile at m/z 299 and (d) the observed [M+1]+ profile for the cetyldimethylethylammonium cation (C20H44N+) with a resolution of 37,400.
Baseline separation between the profiles for the ions containing a 15N atom or a 13C atom was achieved and the profile for ions containing a 2H atom was seen as a high mass tail. Figures 6(c) and 6(d) are [M+1]+ profiles for the cetyldimethylethylammonium cation (C20H44N+) with 37,400 resolution. For m/z 299, the profile for the ion containing a 15N atom is barely discernable for two reasons. The mass difference between ions containing a 15N atom versus a 13C atom (0.00632 u) decreased from 107 to 21 ppm relative to the mass of the [M+1]+ profile. In addition, the relative height of the smaller profile decreased from 11 to 1.7% between Figs 6(a) and 6(c) as the number of C atoms in the ions increased from 3 to 20. These factors limited the utility of inspecting the [M+1]+ profile for the presence of 15N atoms to ions with m/z ratios less than 300.
Checking profiles for interferences, resolution requirements, and optimum CID voltage
Before acquiring profiles to determine relative abundances of the [M+1]+ and [M+2]+ profiles or the exact masses between more widely separated profiles, each profile to be monitored was inspected in the tune view at the highest resolution obtainable as the CID voltage was varied over at least 60 V. Major fragment ions associated with the analyte increased in abundance, while ions formed from the solvent or contaminants declined or increased less. The PGM was used to calculate and plot profiles for comparison. Care was taken to distinguish between profiles from interferences and multiple profiles or peak broadening from analytes as illustrated in Figures 7(a)-7(c).
Figures 7(a) and 7(b) show profiles printed from the tune view with 10,800 resolution for m/z 59 with -10 and 50 V CID voltages applied to the heated capillary and tube lens.
Figure 7. Profiles printed from the tune view with 10,800 resolution for m/z 59 with (a) -10 V, and (b) 50 V CID voltages appplied to the heated capillary and tube lens. In (c) the CID voltage is 50 V, the resolution is 3500, and a wider mass range is shown.
The profile for a contaminant ion in Figure 7(a) disappeared, and the two profiles in Figure 7(b) for fragment ions from chlorpromazine grew proportionally as the CID voltage was increased. Comparison with Figure 6(a) indicated the smaller profile was from an ion containing a 15N atom. The optimum resolution is the lowest resolution for which each profile is baseline resolved from interferences, since the maximum signal is then provided. In Figure 7(c) a wider mass range is shown with 3500 resolution. The lowest-mass profile is an interference and the profiles for C3H815N+ and 13CC2H8N+ have partially merged, which ensured the area from both profiles could be monitored by 18 m/z ratios with a mass increment selected to provide 10 points across an unbroadened profile.
Before acquiring data, the ion abundance of each of the profiles to be monitored was viewed, and the CID voltage was adjusted to provide the maximum signal for the least abundant profile.
Determining compositions of ions based on the resolution needed to separate two profiles
Figure 8 shows two profiles printed from the tune view with 1200 resolution separated by a 4% valley relative to the larger profile or a 14% valley relative to the smaller one.
Figure 8. A tune view printed out with 1200 resolution showing profiles from [C2H3O]+ and [C3H7]+ ions. Listed are several other compositions with m/z 43 and the resolutions needed to resolve their profiles from the profile of [C3H7]+.
This separation is roughly comparable to a 10% valley between profiles of equal height as specified in the definition of mass resolution. The larger profile corresponded to the C3H7+ fragment ion from the cetyldimethylethylammonium cation, while the smaller ion was observed for multiple analytes and was due to the solvent or a contaminant. From the list in Figure 8 of possible compositions and resolutions required to separate their profiles from the C3H7+ profile, only HN3+ and C2H3O+ had smaller masses than C3H7+ and required resolutions close to 1200. Because the first ion is chemically improbable, C2H3O+ (probably CH3CO+ from acetic acid) is the correct composition for the smaller profile. This approach is useful for determining the composition of small ions where chemical noise is most common.
Determination of elemental compositions of desorbed cations or protonated molecules
Seven compounds provided ions weighing between 160 and 319 u. The identity of the first five compounds were known to the operator; the last two were not. The correct elemental compositions of [M]+ were determined for all seven compounds using the methodology described based on sums of isotopic abundances, exact masses, or the numbers of atoms of each element in compositions. No chemical arguments beyond consideration of valences to calculate the number of rings and double bonds were required to reject any composition.
Table 2 shows the compositions that remained after triplicate relative abundance determinations were made for the seven standards studied. The maximum errors for %M+1 and %M+2 were the isotopic abundance errors for each composition plus 3.9% of %M+1 and 5.4% of %M+2. Multiple compositions were possible in all but one case.
For five of the six compounds with multiple possible compositions for [M]+, the composition of a fragment ion was determined from the relative abundances of the [F+1]+ and [F+2]+ ions. For small fragment ions, only a single composition was possible within the error limits established above. Determination of the composition of the neutral loss between the fragment ion and the molecular ion revealed the composition of the molecular ion.
Relative abundances were determined for the protonated molecular ion (m/z 206) and for a fragment ion (m/z 128), which provided the possible compositions in Table 2 based on C, H, O, N, P and F atoms. No S, Cl, or Br atoms were present since %M+2 was less than 4.43% for both [M+2]+ profiles. Because compositions 1-3 contained no F atoms, composition A was rejected. The exact mass difference between m/z 206 and 128 corresponded to a neutral loss of C6H6. Subtraction of C6H6 from compositions 1, 2 and 3 left C8H2ON and C8H18N, which are not listed, and C9H6N, which is composition B. In addition, the [F+1]+ profile for m/z 129 revealed the presence of a 15N-containing ion. Thus, compositions 3 and B were correct.
Five compositions were possible in Table 2 for the relative abundances determined for benzidine. The [M+1]+ profile (m/z 186) displayed a profile for the C12H1315NN+ ion. Thus, only the third composition could be correct.
In Table 3, an abnormally large error was observed for a neutral loss of 17 mass units. This resulted from loss of NH2 in addition to loss of NH3. The m/z 169 profile was a composite from 13CC11H10N+ (m/z 169.08468) and C12H11N+ (m/z 169.08915) ions. Their masses differ by 0.00447 u, which corresponds to 263 ppm for a neutral loss of 17 mass units. A resolution of 38,000 would be needed to resolve the profiles of these ions. However, the error observed relative to the exact mass of NH3, 110 ppm, is much less than the errors for the other two possible compositions of OH (1510 ppm) and CH5 (-629 ppm).
The 18 possible compositions for [M]+ in Table 2 corresponding to the relative abundances determined for the [M+1]+ and [M+2]+ profiles illustrate the rapid increase in possible compositions with increasing mass. Relative abundances determined for the m/z 187 fragment ion provided six possible compositions. The neutral loss between these ions was determined to be C3H6O. Because compositions A-F contain at least 11 C atoms and one O atom, only compositions with at least 14 C atoms and two O atoms were possible for the protonated molecular ion (m/z 245). Consequently, compositions 1-4, 6, 13, 15, 17, and 18 were rejected. The [F+1]+ profile at m/z 81 revealed an ion containing 15N. Thus, compositions 7-9, 11, 12, 14, and 16 were also rejected. Only compositions 5 and 10 remained. Because the first neutral loss from m/z 245 to 230 was determined to be CH3, composition 5 was ruled out for containing only one H atom. Therefore, composition 10, C14H17N2O2, was correct.
This compound required the most data acquisition. Twenty-eight compositions were possible and only three major fragment ions were produced at m/z 43, 57, and 74. The composition of m/z 74 was determined from its relative abundances to be C4H12N+ or C4H7F+. Observation of a profile due to an ion containing 15N in the m/z 75 [F+1]+ profile indicated C4H12N+ was correct. As described earlier, the exact mass difference between m/z 74 and 298 was determined using four non-analyte ions. An exact mass of 298.34182 u for [M]+ was estimated as the exact mass of the m/z 74 ion plus the sum of the five exact masses between each pair of ions studied. Only one composition in Table 2 was possible within error limits of + to -50 ppm, the correct one, C20H44N+.
The relative abundances determined for the [M+1]+ and [M+2]+ profiles corresponded to 27 possible compositions. Two compositions were possible based on the relative abundances of the [F+1]+ and [F+2]+ profiles for the m/z 86 fragment ion. Observation of the [F+1]+ profile revealed that a 15N atom was present and composition A containing an F atom was rejected. The presence of an N atom and 12 H atoms in composition B eliminated all but compositions 2, 7, 8, 12, and 19 for the protonated molecular ion (m/z 319). The exact mass determined between m/z 319 and 214, 105.06122 mass units, corresponded to loss of C4H11NS. The exact mass difference determined between m/z 86 and 214, 127.93863 mass units, was estimated as the sum of exact mass differences between profiles for m/z 86, three non-analyte ions (m/z 99, 131, and 159) and m/z 214. This sum plus the theoretical masses for C8H12N and C4H11NS was 319.09682 u, 21 ppm less than the theoretical value for the correct composition of the protonated molecular ion, C17H20N2SCl, and at least + to -79 ppm different than the theoretical masses for compositions 2, 7, 8, and 12.
Unknown 1: 2-hydroxy-4-methylquinoline
The operator knew the identities of the previous compounds, which were used to develop the methodolgy. The identities of this compound and the one following were not known to the operator before data was acquired and the correct compositions were determined.
For this low mass compound, only one composition was possible based on the relative abundances of the [M+1]+ and [M+2]+ profiles, C10H10NO.
Unknown 2: 1-aminopyrene
Four compositions were possible for [M]+ based on the relative abundances of the [M+1]+ and [M+2]+ profiles. Interferences and low fragment ion abundances prevented determination of relative abundances for any [F+1]+ and [F+2]+ ions. The largest neutral loss in Table 3 was C4H5N. The second composition in Table 2 was unlikely, since a [C11F]+ ion would remain. Examination of the [M+1]+ profile (Figure 6(d)) indicated an ion containing 15N was present, but its abundance relative to the ion containing a 13C atom was too small for three N atoms to be present in [M]+. Thus, composition 1 was unlikely. Figures 3(a) and 3(b) display the profiles obtained without the calibration mass option using compositions 4 and 3, respectively. Failure to observe the second and third profiles in Figure 3(b) indicated composition 3 was incorrect. The errors in the m/z ratios monitored for the third profile would exceed 115 ppm if composition 1 or 2 were tested and, again, the third profile would not be observed, since the m/z range monitored is only 90 ppm. Figure 3(a) indicated C20H44N was the correct composition. As for benzidine, the loss of NH3 indicated that the N atom was part of an amino group, rather than a heterocyclic N atom as for 2-phenylquinoline.
A concentration of analyte is required that ensures its desorbed cation or protonated molecular ion dominates the full scan mass spectrum when no CID voltage is applied. Then, the ions that increase in abundance as the molecular ion is fragmented by applying CID voltages are clearly visible. In addition, the lower abundance of the [M+1]+ and [M+2]+ profiles then pose no problem, since double focusing mass spectrometers feature a wide linear dynamic range. Unlike full scans, MPPSIRD provided the sensitivity needed to determine a %M+2 value of less than 0.2%.
To this point, all experiments employed the ESI source with continuous infusion of the samples. For 10 ng/µL of chlorpromazine in 1:1 MeOH/water with 1% HOAc, a flow of 4 µL/min for 6.75 min for each data acquisition made to determine relative abundances consumed 270 ng of analyte. When trying to identify trace levels of contaminants, this represents a very large amount of analyte. To reduce analyte consumption, the micro-ESI source was tested to determine relative abundances of this analyte. The flow rate was estimated to be 0.23 µL/min with 105 Pa of pressure applied to the sample reservoir and the effluent discharging into a 10-µL syringe. The data acquisition procedure was modified to turn the pressure and tip voltage on 20 s before the MID descriptor started and to turn them off after the data was acquired to conserve sample. A pressure of 105 Pa, the maximum recommended by the manufacturer, provided the maximum signal and rapid initation of the signal when the pressure was applied. These measures reduced analyte consumption to 14.6 ng, or 5.4% of that used with the ESI source while providing similar standard deviations for triplicate determinations.
By alternating between the ESI and micro-ESI sources three times and optimizing the signal after each change, the average signal reduction for the molecular ion was estimated to be only 30% for the micro-ESI source. When sample amount is a concern, the much lower sample consumption of the micro-ESI source is preferred; when it is not, the modestly greater signal provided by the ESI source make it the preferred source.
The methodology developed in this study would help identify compounds in HPLC fractions introduced into the mass spectrometer by off-line infusion. However, when working with complex environmental samples, obtaining relative abundances and exact mass differences for compounds as they enter the mass spectrometer directly on-line from a high performance liquid chromatograph or other separation technique is more efficient. To do so, an ever-present calibrant ion will be required for lock-on before each chromatographic peak starts to enter the mass spectrometer. Use of shorter integration times corresponding to the width of the peaks would degrade precision and increase errors for relative abundance determinations. However, the presence of one calibration compound that provides several ions over a mass range that includes the masses of many analyte ions would again permit direct determination of exact masses. This work illustrates that larger mass intervals between calibration ions than the 12-14 u provided by PFK are acceptable when using MPPSIRD. We expect that the composition of the molecular ion will again be determinable without first determining the elemental compositions of its fragment ions. In addition, greater certainty in correlating molecular and fragment ions will be provided by comparing ion chromatograms, and fewer contaminants will be present as each compound elutes relative to infusion of HPLC fractions of extracts from environmental samples.
For decades, mass calibrants have been used to determine exact masses of analyte ions using high resolution mass spectrometry (HRMS). For low-mass ions, a single elemental composition corresponds to a precisely and accurately determined exact mass. For larger-mass ions, exact masses and relative abundances of mass peak profiles that arise from ions containing less common isotopes than the molecular ion or protonated molecular ion, [M]+, provide the elemental composition of each compound. Widely applicable mass calibrants, including perfluorokerosene, are available for gas phase introduction of analytes ionized by electron impact, but no all-purpose calibrants are available for recently developed liquid sample introduction techniques that use electrospray or atmospheric pressure ionization. This limitation stimulated development of' an alternative approach for determining elemental compositions of ions.
This work shows that elemental compositions of analyte ions can be determined using HRMS without using calibrant ions by determining relative abundances of ions containing heavier isotopes and by determining exact mass differences between the molecular ion and fragment ions. Examination of profiles at single nominal masses is also important. The compositions of small fragment ions, [F]+, can generally be determined from the relative abundances of the [F+1]+ and [F+2]+ profiles, the appearance of the [F+1]+ profile, and the numbers of each element that could be present in [F]+, which are obtained from the list of molecular ion compositions that are possible based on relative abundances of the [M+1]+ and [M+2]+ profiles. Determination of the neutral loss between [M]+ and [F]+ then provides the composition of [M]+. Elemental compositions of molecules introduced into an electrospray ion source in the liquid phase were determined over a mass range suitable for identifying many compounds of environmental significance, including polar or ionic pharmaceuticals in treated wastewater.
The wide linear dynamic range provided by a double focusing mass spectrometer allowed relative abundances of less than 0.5% to be determined accurately. The 100-fold enhancement in sensitivity over full scanning afforded by Mass Peak Profiling from Selected Ion Recording Data (MPPSIRD) permitted observation of fragment ions with low abundances that were produced by applying voltages to induce collisional dissociation.
Even when sums of exact mass differences between ions from contaminants were used to bridge wide mass gaps between small-mass fragment ions and the molecular ion, the total exact mass differences were determined with accuracy sufficient to establish the correct neutral losses.
The methodology developed is useful for analyzing fractions from HPLC or other separation techniques.
- B. Halling-Sorensen, S. N. Nielsen, P. F. Lanzky, F. Ingerslev, H. C. H. Liitzhoft and S. E. Jorgensen, Chemosphere 36, 357 (1998).
- T. Colborn, D. Dumanoski and J. P. Myers, Our Stolen Future, Penguin Books USA Inc., New York (1996).
- NIST Standard Reference Database lA, NIST Mass Spectral Search Program and NIST/EPA/NIH Mass Spectral Library, Ver. 1.0, For Use with Microsoft Windows, U.S. Dept. of Commerce, Technology Administration, National Institute of Standards and Technology, Standards Reference Data Program, The NIST Mass Spectrometry Data Center, Gaithersburg, MD 20899, January 1995.
- A. H. Grange, G. W. Sovocool, J. R. Donnelly, F. A. Genicola and D. F. Gurka, Rapid Commun. Mass Spectrom. 12, 1161 (1998).
- A. H. Grange and W. C. Brumley, LC-GC, 14, 478 (1996).
- Q. Wu, Anal. Chem. 70, 865 (1998).
- A. H. Grange and W. C. Brumley, J. Am. Soc. Mass Spectrom., 8, 170 (1997).
- R. B. Cole (Ed.), Electrospray Ionization Mass Spectrometry, Vol. 1, Wiley, New York, pp. 42-50 (1997).
- W. H. Schaefer and F. Dixon, Jr., J. Am. Soc. Mass Spectrom. 7, 1057 (1996).
- R. Willoughby, R. Voyksner and P. Price, Proc. 46th ASMS Conf. Mass Spectrom. Allied Topics, Orlando, FL, p. 65 (1998).
- A. H. Grange and W. C. Brumley, Rapid Commun. Mass Spectrom. 6, 68 (1992).
- P. Price, J. Am. Soc. Mass Spectrom. 2, 336 (1991).
- A. H. Grange, J. R. Donnelly, G. W. Sovocool and W. C. Brumley, Anal. Chem. 68, 553 (1996).
- D. R. Lide (Ed.), Handbook of Chemistry & Physics, 74th ed., CRC Press, Boca Raton, FL, pp. 1-110 - 1-111 (1994).
- K. Biemann, Mass Spectrometry: Organic Chemical Applications, McGraw-Hill, New York, p. 62 (1962).
- A. H. Grange and G. W. Sovocool, Proc. 46th ASMS Conf. Mass Spectrom. Allied Topics, Orlando, FL, p. 349 (1998).
- J. H. Beynon, Mass Spectrometry and Its Applications to Organic Chemistry, Elsevier, New York (1960).
- E. A. White and G. M. Wood, Mass Spectrometry: Applications in Science and Engineering, Wiley, New York, pp. 163-164 (1986).
- F. W. McLafferty, Interpretation of Mass Spectra, 3rd ed., Univ. Science Books, Mill Valley, CA, p. 23 (1980).
Acknowledgements: The authors thank Dr. Helmut Muenster from Finnigan MAT, Bremen, Germany, for reviewing this article for technical accuracy and for providing Figure 1. The authors also thank Dr. Christian Daughton, Dr. William Brumley and Dr. Alvin Marcus of the Environmental Chemistry Branch for reviewing the article.
The U.S. Environmental Protection Agency (EPA), through its Office of Research and Development (ORD), funded and performed the research described. This manuscript has been subjected to the EPA's peer and administrative review and has been approved for publication.