2656
Anal. Chem. 1985, 57,2656-2663
Thermally Assisted Fast Atom Bombardment: A New Approach toward Optimization of Analyses by Fast Atom Bombardment Mass Spectrometry B. L. Ackermann, J. T. Watson, and J. F. Holland*
Departments of Biochemistry and Chemistry, Michigan State University, East Lansing, Michigan 48824
The use of saturated soiutlons In the preparation of matrices for thermally asslsted fast atom bombardment (TA-FAB) mass spectrometry has been Investigated. Certain saturated matrices, when heated, simulate the desired properties of giycerol without creatlng as much unwanted background. I n addition, differences In the desorption proflles for analyte and matrix Ions are often observed as a result of controlled heatlog of the probe tlp during FAB. I n favorable cases, suitable data can be collected for valid background subtraction, whlch under conventional FAB condltlons are not avallable. Inltial results wlth TA-FAB using saccharide matrices, such as glucose, compare favorably wlth conventlonal FAB analyses using glycerol. Results are presented for the TA-FAB analysis of several nonvolatile blomoiecules whlch Indicate that detection limits In the submicrogram range can be obtained by this method, particularly when a matrix is found whlch Is physlochemically well-suited to the analyte being studied.
A group of mass spectrometric techniques known collectively as desorption-ionization (DI) methods have significantly broadened the scope of mass spectrometric analysis, particularly in the capacity to analyze molecules of either high molecular weight or low volatility. DI methods typically involve the measurement of ions originating from a surface mounted within the ion source chamber. DI methods differ from conventional ionization modes, such as electron impact (EI) and chemical ionization (CI), in that introduction of the sample into the gas phase is not a prerequisite for ionization. While the approach to ionization differs among the various DI methodologies, each process begins with the analyte in the condensed state prior to desorption/ionization. Following the introduction of field desorption (FD) by Beckey in 1969 ( I ) , many other DI techniques have been developed. Examples of these include: plasma desorption (PD) (2), secondary ion mass spectrometry (SIMS) (3),and laser desorption (LD) (4). More recently, a technique known as fast atom bombardment (FAB) was introduced by Barber and co-workers in 1981 (5). This method utilizes an energetic beam of atoms (6-10 keV) to effect desorption/ionization of nonvolatile analytes dissolved in a viscous, low vapor pressure matrix. While many matrices have been used with varying success, the most frequently chosen matrix is glycerol. The widespread use of FAB today is largely due to its low cost and instrumental simplicity, therefore leading to greater commercial availability than other DI methods. The importance of glycerol to the FAB process should be underscored as many workers have used energetic incident (kiloelectronvolt) beams of either atoms or ions to desorb/ ionize molecules with relatively limited success (6, 7). However, glycerol provides a unique physiochemical environment which significantly enhances the desorption/ionization process. Not only is glycerol a good solvent for many classes of polar
molecules such as peptides, OIigosaccharides,and nucleotides, it allows for diffusion of analyte molecules to the solution surface to replenish those “sputtered away during the process. This accounts for the longevity observed for samples run by FAB-MS (8). Finally, glycerol somehow mediates the vast amounts of incident energy into modes suitable for desorption/ionization (as opposed to pyrolysis). Regrettably, FAB mass spectra are frequently overwhelmed by the background generated from the matrix. In the case of glycerol it is not uncommon to see some contribution from the background at almost every mass. Unfortunately, the presence of unwanted background complicates the interpretation of data obtained by FAB and reduces the certainty of proposed structural and molecular assignments. This is particularly true when a predominant background ion is isobaric with the candidate peak of interest, a situation which, unfortunately, is often encountered. What then can be,done to contend with the problem of unwanted background in FAB spectra? In a previous communication (9) we proposed an objective set of guidelines for estimating signal to background (SIB) values for candidate peaks in FAB spectra using relative intensity values of a predetermined set of background peaks in the vicinity of the analyte peak. While this alone does not solve the problem, it serves as an index by which to judge the quality or merit of data obtained by FAB-MS. In essence, it provides a means of distinguishing a candidate peak from the background, thereby providing an objective, empirical index for comparison of FAB mass spectra. Another approach is to perform a computer-aided background subtraction. While this practice is routinely done by many investigators, it is far from a quantitative or reliable process (9). The fundamental problem is that both background and analyte have similar temporal characteristics. In other words, there is no one scan in a FAB analysis which contains a representative view of background only, on which to base a subtraction. Therefore, to create the backgroundonly situation, one usually removes the probe to apply the appropriate blank solution. The major problem here is that background behaves differently alone than in the presence of the analyte. For instance, since analyte desorption/ionization tends to suppress the contribution from glycerol, the use of a glycerol blank frequently leads to oversubtraction. Another problem is realized upon reinsertion where it is difficult to achieve identical focus conditions, particularly for magnetic instruments. Although the problems arising from probe removal may be circumvented by using the rotary, dual probe tip design introduced by Occolowitz (IO), there is still the problem of creating an acceptable blank since many other components may be present along with the analyte in the sample matrix in unknown quantities. I t is important to acknowledge that routine background subtraction may provide insight in many nonquantitative cases and, therefore, may be a useful tool. However, caution should be exercised, especially when viewing a weak analyte signal amidst formidable background.
0003-2700/85/0357-2656$01.50/00 1985 American Chemical Society
A,NALYTICAL CHEMISTRY, VOL. 57, NO. 13, NOVEMBER 1985
An increasingly popular solution to the problem of background in FAB mass spectra is to remove the interference by an instrumental means. One such approach is to use one of a series of techniques known collectively as mass spectrometry/mass spectrometry (MS/MS) (11). MS/MS utilizes two or more mass selection devices in series so that a parent ion may be selected by the first mass analyzer and induced to fragment after which the fragments are analyzed by a second mass analyzer. Another option is to record the spectra using a high-resolution mass spectrometer to differentiate the analyte from background. However, this second approach creates a compromise in sensitivity. Unfortunately, both of these techniques require an expensive array of instrumentation making their use prohibitive to many FAB users. In this paper we report a new approach to the problem of background in FAB mass spectra using a technique we call thermally assisted FAB (TA-FAB). TA-FAB refers to the process of heating the FAB probe surface, containing both the analyte and a saturated solution of a solid substrate, while performing FAB. Under these conditions it is possible to optimize for the desorption of the analyte as a function of temperature. The substrates which have yielded success to date include glucose, fructose, sucrose, thioglucose, and tartaric acid. All are hydrogen bonding solids which, until heated, do not promote optimal desorption of analyte ions under FAB conditions. The intent of this communication is 3-fold: first, to show that saturated matrices, when heated, often yield less background than the viscous liquids used in conventional FAB; second, to demonstrate that the desorption profiles of substrate and analyte under the controlled heating of TA-FAB are, in certain cases, distinguishable, thereby creating a viable situation for background subtraction; third, to show how the TA-PAI3 process may be optimized by selecting a matrix which is physiochemically suited to the particular analyte under study.
EXPERIMENTAL SECTION Materials. All reagents used either as substrates or as analytes in these studies were obtained through one of several commercial sources. Each was of reagent purity requiring no purification prior to use. The analyte solutions were prepared with either distilled water or an aqueous methanol solution depending upon the solubility of the particular solute. The saturated matrix solutions were prepared with 50% methanol in distilled water, primarily to facilitate rapid evaporation of excess solvent (HzOand MeOH) from the probe. All methanol used was of HPLC quality (MCB, Inc.). The glycerol used in the conventional FAB experiments (99.8% Mallinckrodt, Inc.) was distilled under vacuum (1torr) at 120 OC prior to use (12). For comparisonsto TA-FAB,0.5 pL of glycerol was applied to the same probe tip used for TA-FAB experiments from a 5-pL graduated glass micropipet (Fisher Scientific Products). Mass Spectrometry. A Varian MAT CH5 double focusing mass spectrometer with a reverse Nier-Johnson geometry and a 3-keV accelerating potential was used for all data collection. The magnet was scanned at a rate of 25 s/decade, while an electron multiplier set at minus 2.25 kV detected the positive ions transmitted through the sectors. Fast xenon atoms (6.8 keV) were generated by a Model BllNF fast atom gun equipped with a B50 current regulated power supply (Ion Tech, Ltd., Teddington, U.K.). The operating pressure inside the ion source housing during FAB was 1 x 10” torr. Samples were introduced into the mass spectrometer through a vacuum lock using a probe collinear to the optical axis which also accommodates field desorption samples. Once introduced, the probe containing the sample was inserted until it intercepted the atom beam at 90’. At this point, the probe position, lens potentials, and acceleration potential were adjusted to achieve a set of optimum focus conditions. Mass calibration was routinely obtained by performing FAB on saturated CsI/ glycerol on the TA-FAB probe tip without heating. All data were collected, stored, and processed with a dual PDP-8/e data system
W band (ION Source)
2657
1 1 ECP output
7 -
Figure 1. Circuit designed to increase the current supply of the emitter current programmer.
(9). The dual arrangement provided real-time display of the data during the analysis and was capable of performing background subtraction and other processing routines. In addition, the host PDP-8/e minicomputer communicated with a larger PDP-11/44 computer allowing additional storage and software flexibility. TA-FABInstrumentation. Under the conditions of TA-FAB, an analyte/matrix mixture applied to a conductingemitter surface is resistively heated in a precise and controlled manner while undergoing fast atom bombardment. The emitter was made by spot welding a piece of 97% W, 3% Re (0.235 in. X 0.018 in. X 0.001 in.) acroea the stainlesssteel posts of an emitter base identical with those used to prepare activated emitters for FD. The ends of the posts were filed in such a manner to approximate the preferred 30’ angle between the spot welded band and the incident atom beam (e.g., angle of incidence 60’) (12). Upon insertion into the ion source, a small stainless tab on the probe made contact with a metal plate which supplied the current. This enabled current to flow through the emitter to the reference return which in this case was the potential of the ion source (+3 kV). The same arrangement has been used previously for FD experiments. Current was supplied to the tungsten band probe tip by use of an emitter current programming unit (ECP) designed earlier in this laboratory to provide precise regulation of current through FD emitters (19). Because of the lower resistance of the tungsten band relative to an FD emitter, the current capacity of the ECP had to be fortified to achieve similar temperatures in the tungsten band. Hence, the ECP was modified to introduce the feedback circuit of Figure 1. For convenience of prototyping, this circuit incorporated a 12-V lead storage battery as the power source and was capable of providing currents in excess of 3 A through the tungsten emitter. The total current gain resulting from modification was approximately 30-fold. In addition, the battery was able to supply operational currents for periods greater than 2 weeks between recharging. Because the battery floats at source potential, it was necessary to isolate it from ground. A polyethylene battery box (Rubber Queen Marine Products, Jackson, OH) was used for this purpose. Sample Preparation and Analysis by TA-FAB. Samples were applied to the tungsten band in the following manner. First, a 0.5-pL droplet of the saturated matrix solution was applied from a 10-pL Hamilton syringe (Reno, NV) and the solvent removed with a heat gun. For the fructose solution, 1 pL was found to contain 60 f 5 fig of fructose. When the saccharide matrices were used, the desolvation process appeared to be incomplete,leaving a syruplike coating on the band rather than a well-defined solid. One microliter of analyte solution was then administered onto the matrix after which the solvent was gently removed with the heat gun. The probe was then introduced into the mass spectrometer and inserted until contact was made with the ECP. The FAB gun was turned on and the instrument focused on a background peak while applying little or no heat to the band so that early desorption of the analyte could be avoided or minimized. During the initial three scans in all the TA-FAB runs, data were collected similar to that by conventional FAB (no heat). During the fourth scan, the current was manually increased (slowly) to a value of 0.90 A. Beginning with the fifth scan, a heating rate of approximately 100 mA/min was initiated using the ECP. The current was allowed to increase beyond the point of maximum analyte desorption before returning to 0.00 A.
ANALYTICAL CHEMISTRY, VOL. 57, NO. 13, NOVEMBER 1985
2658
2
n/z
-"
8 -
i
1
123
(3G+H)+ 277
357 (4G'H)' 369
i
325
350
400
(M+G)+
I 450 n/z
500
550
600
Figure 2. Averaged mass spectrum produced from 0.5 pL of a saturated fructose matrix SOlUtlOn under the conditions of TA-FAB.
RESULTS AND DISCUSSION One reason saturated solutions were chosen for use as substrates in these investigations was the possibility that these solutions, when heated, might produce less background than that commonly observed when using glycerol under conventional FAB. At the same time, satufated solutions, when heated, might simulate the physiochemical properties of glycerol which enhance desorption/ionization by fast atom bombardment. In order to maximize the chemical similarity with glycerol, several hydrogen-bonding solids have been considered. Previous work using solids as matrices for other desorption/ionization methodologies had shown a positive effect on the data produced. Friedman and Beuhler reported increased yields for the desorption of small peptides while performing rapid heating in the presence of either urea or oxalic acid (14). More recently, Cooks and co-workers found that the desorption efficiency of molecular ions was increased for the SIMS spectra of organic salts when ammonium chloride was used as a matrix (15). They also observed a decrease in analyte fragmentation and less spectral background by this method. Analogous results were reported by Vestal et al. who used inositol and tartaric acid to enhance the respective desorption of erythromycin and histidine from a moving belt LCMS interface using laser desorption (16). Initial success in our laboratory was achieved with glucose when it was noticed that controlled heating had a pronounced effect on the desorption profile produced by FAB. Glucose was selected as a candidate because of its obvious structural similarity to glycerol. However, glucose holds at least one important advantage over glycerol as a FAB matrix: glucose does not tend to form ions by clustering. In contrast, glycerol forms many hydrogen-bonded clusters which are the basis for many background ions observed at higher mass. Figure 2 shows the spectrum obtained when 0.5 KL of fructose was subject to the conditions of TA-FAB. The result is very similar to that obtained for glucose under similar conditions. The absence of peaks in the high mass region of the mass spectrum is clearly apparent. The magnification by 100 past m/z 400 indicates that any contribution past this point is due to noise on the electron multiplier. A small degree of dimerization is witnessed by the peaks at m/z 365 and 325 which correspond to (2 fructose Na - H,O)+ and (2 fructose H - 2H20)+,respectively. However, the highest mass peak observed for glucose or fructose in many analyses is commonly the natriated parent (M + Na)+ which occurs at m / z 203. No (M H)+ ions are observed for glucose, as protonation at a hydroxyl group results in the facile loss of water to give a peak at mlz 163. Further fragmentation gives rise to the con-
+
+
+
Ix3 Mt
1 0
50
IO0
150
200
8 4 ,
250
/'""""I
300
""""'I
350
400
m/z
Figure 4. TA-FAB spectrum of 3 pg of thiamine hydrochloridein 0.5 ELLof glucose averaged over the region of analyte desorption.
sistently observed fragment ions series at m / z 145, 127, 115, 97,85,73,57, and 43. Even though these peaks are sometimes large in magnitude, they fortunately appear at low mass, a region of little interest during most analyses by FAB-MS. Comparison of FAB and TA-FAB for Thiamine Hydrochloride. Figure 3 is a conventional (averaged) FAB spectrum of 3 pg of thiamine hydrochloride (vitamin B,) dispersed in 0.5 pL of glycerol and bombarded with 6.8-keV xenon atoms. The dominance by glycerol in this spectrum is apparent at m/z 93 and 185, peaks which correspond to the protonated glycerol monomer and dimer, respectively. Because of the contribution by glycerol in this spectrum, only three peaks may be recognized as originating from the analyte. First, the ion of mass 265 is the intact cation of the thiamine hydrochloride salt. Secondly, the ion of mass 357 corresponds to the addition of glycerol to the intact thiamine cation. Finally, the peak at m/z 123 results from cleavage between the two heterocyclic rings of thiamine. The origin of this ion was revealed by Cooks et al. who used SIMS in connection with MS/MS to study the fragmentation of vitamin B, in glycerol (17). Their work demonstrated that the peak at m/z 123 results from collisionally activated dissociation of m/z 357 (e.g., (M + G)') and is not a daughter of M+. Although Cooks also showed that two resonance-stabilized daughters (mlz 122, 144) result from fragmentation between the two rings when the thiamine cation (M+)is the parent, neither is of sufficient intensity to make a reliable assignment in the FAB spectrum of Figure 3. As a comparison to conventional FAB, the experiment was repeated by TA-FAB using 0.5 p L of saturated glucose instead of glycerol. The two runs were performed consecutively, under identical experimentalconditions. Between runs, the tungsten band was thoroughly cleaned with acetone, dilute HC1, and
ANALYTICAL CHEMISTRY, VOL.
2659 I
I
I
0
57, NO. 13, NOVEMBER 1985
20
10
Yo loo%=1440
(C1H602 It
SCAN NUMBER
0
0
Figure 5. Reconstructed mass chromatograms for analyte (mlz 265) and background ( m / z 73) ions for the TA-FAB analysis of 3 pg of thiamine hydrochloride In 0.5 p L of glucose.
distilled water and then checked for the presence of memory effects by TA-FAB. Figure 4 displays the mass spectrum obtained by averaging the scans during which desorption of thiamine occurred. In this spectrum, clear evidence is given for the molecular ion peak at m / z 265 and also for the fragment ion peaks occurring at m / z 122 and 144, which were observed by Cooks under collisionally activated dissociation (17).The peak at m / z 52 is Crt which originates from the stainless steel present in the probe tip. Even without subtraction, the selection against background is superior for TA-FAB as compared to the conventional FAB spectrum in Figure 3 where glycerol contributes the most intense peaks in the mass spectrum. Not only is the influence of glycerol (conventional FAB) more intense than that of glucose in TA-FAB, but the multiplicity of its appearance is greater due to clustering. For example, in Figure 3 clustering accounts for the peak 12 mass units above the thiamine molecular ion (e.g., (3G H)+, m / z 277). In contrast, in Figure 4, the lack of clustering by glucose confines most background peaks in the TA-FAB spectrum to the region below mass 200; none of the series of glucose ions (previously listed) may be readily discerned in this case. It should be noted that selection against glucose is not always as complete prior to subtraction as the data in Figure 4 indicate. Selection is dependent upon many factors which include analyte concentration, choice of substrate, and experimental conditions. However, while matrix peaks are commonly observed in TA-FAB spectra, their overall contribution is generally less imposing than the background generated by glycerol. The chronology of the TA-FAB experiment for thiamine may best be visualized by following the course of the two reconstructed mass chromatograms appearing in Figure 5. For the duration of the heating process, these are essentially ion desorption profiles. The upper mass chromatogram corresponds to the molecular ion of thiamine (m/z 265), while the lower trace is the mass chromatogram of the predominant glucose ion a t m / z 73 (C3H502+).The plots in-Figure 5 indicate the temporal nature of the desorption profiles attained by this method. The first four scans represent data collected by FAB only (no heat). These scans contained mostly glucose peaks. However, since the same glucose peaks were observed before and during heating, this region is indicative of the background which codesorbs later with the analyte. Next, the current was increased manually to 0.90 A at scan 4, after which a controlled ramp of the current was initiated a t scan 5. The current increased linearly past the point of optimum analyte desorption, indicated by the point of maximum intensity in the mass chromatogram of m / z 265 (Figure 5). A current maximum of 1.54 A was reached at scan 18 after which the ECP was turned to 0.00 A. Examination of the mass chromatogram at m / z 265 in Figure 5 shows that the temperature providing maximum
+
50
100
150
200
250
I""""' 300
350
400
m/z
Flgure 6. TA-FAB mass spectrum of 3 pg of thiamine hydrochloride in 0.5 p L of glucose following subtraction of an averaged background
spectrum. desorption occurred at scan 13. We choose to refer to this point as the "best matrix temperature" or BMT. This name was given by analogy to field desorption where the term "best anode temperature" or BAT refers to the optimum desorption temperature for a particular FD analysis. The value for the BMT observed for thiamine in Figure 5 corresponds to a current of approximately 1.29 f 0.05 A. Differences in the observed BMT currents for repeated analyses can be attributed to chemical and instrumental variations experienced with sequential determinations. It is believed that the temperature at the BMT actually varies much less than this; studies are currently being conducted to confirm this assumption. Potential for Background Subtraction Utilizing Differential Desorption Profiles. Controlled heating of the analyt,e/matrix mixture while undergoing TA-FAB allows for a feature not available in conventional FAB, Le., the possibility of differential desorption between analyte and background ions. Since m/z 73 is a reliable indicator of matrix desorption for glucose, it is clear from the data in Figure 5 that differential desorption between analyte ( m / z 265) and background ( m / z 73) was achieved during the TA-FAB analysis of thiamine hydrochloride in glucose. For this example, background subtraction was acceptable since the instrumental conditions and sample composition were not disturbed by removal of the probe to create a background-only situation by means of a second analysis. To make background subtraction a reliable and quantitative procedure requires that the scan chosen for subtraction be a true representation of the background present during the region of analyte desorption. The best way to approach this ideal is to track several major background ions and find a scan (or an average scan) containing only background that most accurately represents the intensity of these ions during the region of optimum analyte desorption. This situation closely resembles the approach taken to background subtraction in GC/MS. Using the data in Figure 5, and assuming (for simplicity) that mlz 73 was representative of all background present, scan 2 would be a viable candidate as the spectrum for the background because the intensity for m / z 73 in this scan is approximately equal to the intensity found for this ion during the region of analyte desorption. However, the actual scan used was an averaged background spectrum formulated by averaging several spectra on both sides of the analyte desorption profile ( m / z 265) to provide a more accurate representation of all background ions present. When this averaged background spectrum was subtracted from the thiaminecontaining averaged spectrum of Figure 4,the result is the subtracted spectrum shown in Figure 6. The net result is a spectrum (Figure 6) virtually free from interfering background. Structural Information for Ala-Leu-Gly Obtained by TA-FAB. The conventional (averaged) FAB spectrum of 1
2660
ANALYTICAL CHEMISTRY, VOL. 57, NO. 13, NOVEMBER 1985
- 100%= 12760 -
'
"'I'
!
~
260
- 100% 7264
- 100%= 7512
1
0
100
50
I50
200
Y. I
250
I
300
SCAN NUMBER
Flgure 7. Conventional averaged FAB mass spectrum of 1 pg of alanyl-ieucyl-glycine in 0.5 pL of glycerol.
I
I
144
1
(M+H)*
0
50
100
260
I50
250
I
360
m/z
(M*H)'
86
260
0
50
100
150
200
250
300
m /z Flgure 8. (a) TA-FAB mass spectrum of 1 pg of alanyl-leucyl-glycine in 0.5 pL of fructose averaged over the BMT. (b) TA-FAB mass spectrum of 1 pg of alanyl-leucylglycine in 0.5 pL of fructose following subtraction of an averaged background spectrum.
of the tripeptide alanyl-leucyl-glycinedissolved in 0.5 p L of glycerol is shown in Figure 7 . Molecular weight confirmation is possible from the intense peak for the protonated molecule at mlz 260, while the peaks at mlz 185 and 189 can be assigned to fragments of the tripeptide (as indicated in Figure 7 ) . However, the undesired contribution from glycerol is apparent, especially at mlz 185 where the protonated glycerol dimer is isobaric with the proposed fragment (M + H - Gly)+. Hence, the sequence of the tripeptide may not be obtained from this mass spectrum. Barber et al. used linked-scanning to show that the (M + H)+ ion of mass 260 does, in fact, give rise to a daughter of mass 185 for this molecule (18). However, an investigator without access to MS/MS or high-resolution MS would be forced to seek an alternate FAB matrix to alleviate the intense interference at m l i 185. While this may be a viable approach to the problem, it is a cosmetic solution which merely substitutes one form of background for another. Figure 8a displays a spectrum, averaged over the region of the BMT, for 1 pg of alanyl-leucyl-glycineusing fructose as the TA-FAB matrix. The instrumental conditions were the pug
10
0
m/z
Flgure 9. Reconstructed mass chromatograms for analyte ( m l z 260, 189, 185) and background ( m l z 85) for the TA-FAB analysis of 1 pg of alanyl-leucyl-glycine In 0.5 pL of fructose.
same as those used to obtain the conventionalFAB spectrum of Figure 7. Complete sequence information is available from the assigned fragments, labeled on the mass spectrum, as very little competition from background is observed. Five major peaks in the mass spectrum (mlz 189,185,157,86,44) arise from fragmentation of the tripeptide. These peaks are consistent with the complete metastable analysis reported by Barber et al. for this compound with the observation that the peak at mlz 44 arises as a daughter of the fragment ion at mlz 157 (18). Figure 8b represents the result obtained following the subtraction of an averaged background scan from the averaged spectrum of Figure 8a. Again, the result yields a clear picture showing the major fragments of the tripeptide as nearly all of the competition from background has been removed. Furthermore, not only does this result verify the loss of glycine at mlz 185, it also suggests that the competing loss of alanine occurs with about equal probability. This ratio seems reasonable since there is no compelling structural reason to favor one fragment over the other. However, since large deviations were sometimes apparent in the ratios of fragment ion intensities, it should not be concluded that the 1851189 ratio shown here is correct. Reexamination of the spectrum obtained using glycerol (Figure 7) shows a 1851189 ratio of about 5:l which obviously does not reflect the fragmentation pattern of the analyte. Desorption Profiles Indicate Origin of Ion Species. As previously demonstrated, differences in the desorption profiles of matrix and analyte ions made possible by TA-FAB can lead to a favorable situation for background subtraction. Another consequence of this behavior is that it is often possible to distinguish analyte peaks from background simply from their respective mass chromatograms. This is because analyte fragment ions have desorption profiles very similar to those of the molecular ion and different from those of the background ions. This effect is illustrated in Figure 9 where the reconstructed mass chromatograms of the fragment ions at mlz 185 and 189 closely resemble that of the molecular ion at mlz 260. Each analyte peak gave maximum intensity during scan 6 indicating the BMT occurred at a current of about 1.0 A. Also present is the profile of the fructose fragment C4H502+ ( m / z 85) whose desorption is clearly different from those of the analyte peaks of the tripeptide. Unfortunately, the behavior of background ions in the region of the BMT is not always predictable, which can hinder the search for a background spectrum representative of the background at the BMT. Hence it is possible to have good selection against background but not have the proper conditions for a quantitative subtraction whenever anomalous behavior is observed for the background ions in the region of the BMT. The cause of this behavior is currently being investigated.
57, NO. 13, NOVEMBER 1985
ANALYTICAL CHEMISTRY, VOL.
Optimization in TA-FAB by Proper Selection of Analyte/Matrix Combinations. In addition to creating a favorable situation for background subtraction, the use of heated saturated solutions as FAB matrices offers new alternatives for the optimization of analysis by FAB. It is now possible to seek matrices with chemical and physical properties which are the best suited to the particular samples being analyzed by TA-FAB. Experimental confirmation of analyte/matrix interaction may be evidenced by two distinct types of observations. The first is the enhancement of anal@ peaks over background. This effect may be empirically evaluated by SIB ratios. The second type of observation indicative of analyte/substrate interaction would manifest itself by changes in the analyte fragmentation pattern from one substrate to another. Arginine was chosen to illustrate how important matrix selection is to the results obtained by TA-FAB. One microgram of arginine was analyzed by TA-FAB alternatively in tartaric acid and glucose keeping the conditions of analysis constant. One-half microliter of a saturated solution of each respective matrix was applied to the emitter in each case. The BMT observed for arginine in glucose was 1.02 A, whereas the BMT in tartaric acid was somewhat higher (1.22 A). Unlike most amino acids which form zwitterions, arginine contains an additional site of positive charge at the guanidinium group. Hence, it exists as an ion of charge +1 between pH 2.17 and pH 9.04. This ion has an m / z value of 175 and appears in the TA-FAB mass spectra of arginine using either glucose or tartaric acid. A comparison of the data is presented in Figure 10. Figure 10a is the mass spectrum obtained at the BMT for 1 pg of Arginine in glucose, while Figure 10b is the corresponding spectrum using tartaric acid. Arginine may be distinguished from background in each case; however, the absolute intensity of the protonated molecule at m/z 175 was 3.86 times greater when analyzed in tartaric acid than when analyzed in glucose. The difference in sensitivity reflects the greater ability of tartaric acid to act as a proton donor creating a larger concentration of (M + H)+ ions for arginine. A more cogent and realistic indication of the presence of an analyte peak in FAB mass spectra is the degree to which the peak extends above the neighboring background peaks. Unfortunately, the comparison above (using absolute intensities) gives no indication of the relationship between the peak at m/z 175 and competing background for the two substrates used in this study. To assist in such determinations we have developed a set of rules (9) for estimating signal to background (SIB)values for analyte peaks in FAB mass spectra. In this procedure the relative intensity of the candidate peak is compared to a select set of background ion intensities used to formulate an average background value. Application to the present example showed a relative enhancement in S I B of 3.0 for the TA-FAB analysis of 1pg of arginine in tartaric acid relative to that in glucose. Figure 1Oc and Figure 10d display the results achieved after the background subtraction of glucose and tartaric, respectively, from the mass spectra in Figure 10a and Figure lob; averaged background spectra were used in each case. Both subtractions yield more favorable data as the (M + H)+ion at mlz 175 becomes the base peak in each of the mass spectra shown. Yet, the presence of several background peaks of lesser intensity indicates that background subtraction was not 100% effective in either case. A visual inspection of Figure lOc,d reveals that less overall competition from background occurred for the tartaric acid example. Again, this conclusion was verified from SIB calculations which show that there was more than a %fold advantage in the observed SIB ratio when using tartaric acid instead of glucose.
2661
H2N-C-NH-CH2-CH2-CH2-CH-COO-
0
50
100
260
150
250
m/2
b
W E )= 27 M*H)+ 175
I
I
I
0
50
100
I
200
I50
250
m/z
1
!
(S/B)=22 43
I05
0
50
100
150
200
250
m/z
.M+H)' I75
d
W E ) = 72
0
50
IO0
150
200
2:
m/z
Figure 10. (a)TA-FAB mass spectrum recorded at the BMT for 1 pg of arginlne in 0.5 p L of glucose. (b) TA-FAB mass spectrum recorded at the BMT for 1 pg of arginine in 0.5 pL of tartaric acid. (c)TA-FAB mass spectrum of 1 pg of arginine in 0.5 p L of glucose following subtraction of an averaged background spectrum. (d) TA-FAB mass spectrum of 1 pg of arginine in 0.5 pL of tartaric acid following subtraction of an averaged background spectrum.
A dependency by fragment ion relative intensities or fragment ion type on the matrix employed would also indicate the presence of specific analytelmatrix interactions. Even though subtle differences are apparent in the arginine spectra presented, there is presently not enough evidence to suggest that analyte/matrix interaction is responsible. To make such a determination, one would need to know the origin of each
2682
ANALYTICAL CHEMISTRY, VOL. 57, NO. 13, NOVEMBER 1985 (Fructose+NaY I203
50
0
50
100
I50
200
25C
300
350
100
150
200
250
300
400
rn/z
Figure 11. TA-FA9 mass spectrum (averaged) of 1 fig of ampicillin in 0.5 pL of fructose following subtraction of an averaged background spectrum.
major fragment as well as the inter-run variance in fragment ion intensities; work in this area is continuing. At present, controlled experiments have been performed to assess the inter-run variations associated with the analysis of 1pg of alanyl-leucyl-glycine by TA-FAB using two matrices: fructose and tartaric acid. For each analysis, the intensities integrated over the BMT for the six most intense analyte ions were summed and ratioed to the total ion current integrated over the same interval. The calculated ratio appeared to be independent of the matrix used (e.g., fructose (22.0%),tartaric acid (21.2%)). In addition, the level of imprecision found in each case was encouraging considering the number of possible sources of error: fructose (average relative deviation lo%, n = 5) and tartaric acid (average relative deviation 8%, n = 7). It should be mentioned that it was necessary to consider the contribution from several major analyte ions in order to achieve the precision indicated above. Also, it was imperative that replicate analyses be performed consecutively and under identical experimental conditions. Additional Applications of TA-FAB to the Analysis of Nonvolatile Biomolecules. To date, TA-FAB has been successfully applied to the analysis of a variety of nonvolatile molecules of biological interest. The list presently includes ampicillin, acetaminophen sulfate (K+ salt), tryptophan, glycyl-prolyl-alanine, taurodeoxycholic acid (Na+ salt), raffinose, and riboflavin. In each case, 1-5 pg of analyte was analyzed by both TA-FAB and conventional FAB (using glycerol). The comparisons to conventional FAB were generally quite favorable, even without the aid of background subtraction. Furthermore, the large SIB values observed for many of the analytes tested suggest that detection in the submicrogram range should be possible for several compounds using TA-FAT3 (especially if matrix optimization is employed). One microgram of the free acid form of the antibiotic ampicillin was analyzed in frutose by TA-FAB. A background subtracted spectrum appears in Figure 11. In addition to the protonated molecule at mlz 350 five distinct fragment ions are present; each exhibited a desorption profile similar to that at m f z 350. The origin of the fragments at m f z 106,160, and 192 is shown in Figure 11and may be readily assigned to stable even-electronspecies. On the other hand, several possibilities could be proposed to account for the peaks at mlz 118 and 174. The latter might be the loss of water from the ion of mass 192. As a final example, Figure 12 displays an unsubtracted averaged scan obtained over the region of the BMT, for the TA-FAB analysis of 5 pg of taurodeoxycholic acid in fructose. The base peak at mlz 544 corresponds to the addition of sodium to the intact sodium salt of this molecule. The most intense contribution of fructose is the (M + Na)' ion at mlz 203. The large degree of fragmentation seen below mass 200 originates not only from fructose but from successive decom-
rn/z
Flgure 12. Unsubtracted TA-FAB mass spectrum of 5 pg of taurodeoxycholic acid (Na' salt) in 0.5 pL of fructose averaged over the region of analyte desorption.
position of the steroid ring structure. The peaks at mlz 224 and 252 can be explained by fragmentation between the C and D rings of taurodeoxycholic acid.
CONCLUSION The use of heated saturated solutions as matrices for FAB-MS has been shown to result in several important consequences for the analysis of nonvolatile molecules. Among these are (1)the observation of less background in FAB mass spectra, (2) the possibility for valid background subtraction, and (3) extension of viable matrix candidates to materials other than viscous liquids. In addition, the sensitivity available by TA-FAB for a variety of nonvolatile biomolecules compared favorably to that of conventional FAB spectra run under identical instrumental conditions. It appears from bur investigations that analyte desorption is facilitated by the increased "fluidity" given to the matrix by a nonionizing form of energy (e.g., resistive heating of the emitter surface). This is to be contrasted from "melting", since melting is a defined macro phase change and does not appear to be appropriate under these experimental conditions as the matrices appeared to exist as viscous syruplike liquids in the initial stages. Regardless of the initial physical state of the matrix, intense analyte desorption relative to background, does not occur until the matrix is heated. As the sample is heated in a programmed fashion, it passes through an optimum or "best matrix temperature" where maximum analyte desorption occurs. Differences in the desorption profiles for analyte and matrix ions thus reflect differences in the optimum thermal conditions for each. Further, it was noticed that glucose had a less specific desorption profile than the analytes tested. Smooth desorption profiles occur for analytes by TA-FAB because the increased diffusion created by heating enables analyte molecules to move toward the surface to replace molecules sputtered away by the atom beam. This effect is also responsible for sample longevity. Optimizing the chemical relationship between analyte and matrix is certainly not unique to TA-FAB. Proper matrix selection has long played a major role in analyses by conventional FAB, where several alternatives to glycerol have been reported in particular applications (8, 19,ZO). In addition, many in situ derivatizations are routinely performed to enhance the quality of the data obtained by FAB (12,21,22). The most common example is the addition of a proton donor to the FAB matrix to promote formation of (M + H)' ions of the analyte, thereby enhancing sensitivity. We believe saturated matrices could offer even more possibilities for optimization by TA-FAB due to the greater number of candidate solids which exist (as compared to viscous liquids). The range of applications of TA-FAB will be dependent upon success in the search for new matrices. Acceptable
Anal. Chem. 1985, 57,2663-2668
2663
candidates are those which become fluid (rather than decomposing) under the conditions of TA-FAB. Several solids me currently being evaluated for use: citric acid, inositol, urea, oxalic acid, and ammonium chloride. Polymers shall also be investigated. Work in progress will delineate several indices of analytical merit such as sensitivity, precision, and dynamic range. Finally, a continuing objective will be to more clearly understand the phenomena responsible for the observed effects associated with this technique.
(8) Barber, M.; Bordoli, R. S.; Sedgwick, R. D.; Tyler, A. N.; Elliott, 0. J. Anal. Chem. 1982, 5 4 , 645A. (9) Ackermann, B. L.; Watson, J. T.; Newton, J. F., Jr.; Hook, J. B.; Braselton, W. E., Jr. 6lomed. Mass Spectrom. 1984, 1 1 , 502. (IO) Gilliam, J. M.; Landis, P. W.; Occolowitz, J. L. Anal. Chem. 1983, 55, 1531. (11) Yost, R. A,; Fetterolf, D. D. Mass Spectrom. Rev. 1983, 2, 1. (12) Martin, S. A.; Costello, C. E.; Biemann, K. Anal. Chem. 1982. 54, 2362. (13) Maine, J. W.; Soltmann, B.; Holland, J. F.; Young, N. D.; Gerber, J. N.; Sweeiey, C. C. Anal. Chem. 1976, 4 8 , 427. (14) Friedman, L.; Beuhler, R. J. 26th Annual Conference on Mass Spectrometry and Allied Topics, St. Louis, 1978; Workshop on Field De-
ACKNOWLEDGMENT The authors thank M. R. Davenport for his continued technical assistance in support of this research. Registry No. Fructose, 57-48-7; glucose, 50-99-7;tartaric acid, 87-69-4; thiamin hydrochloride, 67-03-8; alanyl-leucyl-glycine, 60030-20-8;L-arginine, 74-79-3;ampicillin, 69-53-4; sodium taurodeoxycholate, 1180-95-6.
sorption. (15) Busch, K. L.; Hsu, B. H.; Xle, Y.; Cooks, R. G. Anal. Chem. 1983, 55,
LITERATURE CITED (1) Beckey, H. K. Int. J. Mass Spectrom. Ion Phys. 1969, 2, 500. (2) Torgerson, D. F.; Skowronski, R. P.; Macfarlane, R. D. Blophys. Res. Common. 1974, 60, 616. (3) Grade, H.; Wlnograd, N.; Cooks, R. G. J. Am. Chem. SOC. 1977, 99, 7725. (4) Posthumus, M. A.; Kistemaker, P. A,; Meuzelaar, H. Anal. Chem. 1978, 50, 985. (5) Barber, M.; Bordoli, R. S.; Sedgwick, R. D.; Tyler, A. N. J. Chem. Soc.. Chem. Commun. 1981, 325. (6) Devienne, F. M.; Roustan, J. C.; C . R . Hebd. Seances Aced. Sci., Ser. 6 1976, 283, 397. (7) Benninghoven, A,; Jaspers, D.; Sichtermann, W. Appl. Phys. 1978, 11,35.
1157. (16) Hardin, E. D ' Fan, T. P.; Vestal, M. L. Proceedings of the 32nd Annual (17)
(18) (19) (20) (21) (22)
Conferencejbn Mass Spectrometry and Aiiied Topics, San Antonio, TX, 1984, 246. Glish, G. L.; Todd, P. J.; Busch, K. L.; Cooks, R. G. Int. J. Mass Spectrom. Ion Proc. 1984, 56, 177-192. Barber, M.; Bordoli, R. S . ; Sedgwlck, R. D.; Tetler, L. W. Org. Mass Spectrom. 1981, 16, 258. Germain, P.; Prome, J. C. Org. Mass Spectrom. 1984, 19, 448. Meili, J.; Seibi, J. Org. Mass Spectrom. 1984, 19, 561. Busch, K. L.; Unger, S. E.; Vincze, A.; Cooks,%. G.; Keough, T. J. Am. Chem. SOC. 1982, 104, 1507. Morris, H. R.; Dell, A.; Etienne, A. T.; Judkins, M.; McDonell, R. A.; Panico, M.;Taylor, 0. W. Pure Appl. Chem. 1982, 5 4 , 267.
RECEIVED for review March 14, 1985. Accepted July 5,1985. This work was supported by the Biotechnology Research Program under the Division of Research Resources of the National Institutes of Health (NIH Grant No. RR00480-16). In addition, B. L. Ackermann received support from the Dow Chemical Co. in the form of a research fellowship.
On-Line Ion Implantation for Quantification in Secondary Ion Mass Spectrometry: Determination of Trace Carbon in Thin Layers of Silicon Howard E. Smith and George H. Morrison*
Department of Chemistry, Cornel1 University, Ithaca, New York 14853
The duopiasmatron ion source of the Cameca IMS 3f secondary Ion mass spectrometer has been used to generate a mass-filtered ion beam for the purpose of Implanting an internal standard Into a semiconductor matrlx. Depth profile analyses of C+ Implants superimposed upon residual C concentrations In a 5800-A film of polycrystalllne Si have shown that unlform, accurate doses of primary Ions can be implanted. Residual C concentrations, determined from nine such on-line Ion Implant analyses, gave values of 3.4 (f0.4) X 10'' atoms/cm3. This value agreed accurately wlth an off-line Ion implant concentratlon determination, wlth a relatlve dlfference of -8 %. This technlque extends the quantltatlve capablllties of the Ion mlcroanalyzer.
Secondary ion mass spectrometry (SIMS) offers high sensitivity for the detection of most elements,high dynamic range, and excellent depth resolution (50-100 A). Hence, it is an excellent technique for quantitative elemental depth profiling analysis. In empirical quantification, the signal level from a known concentration of an external standard is used to obtain the analytical curve or sensitivity factor (1). The variability in sputtering yields and practical ion yields observed in sam0003-2700/85/0357-2663$0 1.50/0
ples of different matrices places several restrictions on the analysis. The matrix of the standard must be identical or nearly identical to that of the analyte of interest. The standard and the analyte samples must be analyzed under the same instrumental and experimental conditions. Any variation in the experimental conditions that takes place during the analyses can be corrected for by referencing the analyte signal level to that of an internal reference element, usually a matrix signal. This method is known as the relative sensitivity factor (RSF) method. The method assumes that the analyte and reference signals will be affected in the same way by changes in the experimental conditions. The method of solid-state internal standard addition elegantly circumvents the matrix-effect problem. Ion implantation is used to superimpose a quantification standard upon a homogeneously distributed analyte: typically, a mass-filtered ion beam is accelerated to 50-300 keV and rastered uniformly over the surface of the sample. The ions are stopped in the sample as atoms with a distribution that approximates that of a Gaussian (2). Subsequent depth profiling of this additive distribution, and the ratio of the integrated signals from each, gives a measurement of the volume concentration of the residual analyte, if the depth of analysis and the implanted fluence are known. The methodology and requirements for 0 1985 American Chemical Society
