Chemical ionization mass spectrometry of thermally labile compounds

May 14, 1979 - in part provided by the Computer Science Center of the. University of ... Department of Chemistry, University of Delaware, Newark, Dela...
0 downloads 0 Views 492KB Size
Anal. Chem. 1980, 52, 245-248 (17) Aras, N. K.; Zoller, W. H.; Gordon, G.E.; Lutz, G. J. Anal. Chem. 1973, 45, 1481-1490. (18) Olmez, I.; Aras, N. K.; Gordon, G. E.; Zdler, W . H . Anal. Chem. 1974, 46, 935-937.

245

Department of Energy under Contract No. Ey-76-S-05-5173 and the Other by the NSF/ASRA under Grant No. ENV75-02667 and the NSF/Office of Polar Programs under Grant No. DPP-7623423. Computer time was in part provided by the Computer Science Center of the University of Maryland.

RFZEIVEJI for review May 14,1979. Accepted October 30,1979. T h e PGAA experiments were supported in part by the U S .

Chemical Ionization Mass Spectrometry of Thermally Labile Compounds Gordon Hansen' and Burnaby Munson" Department of Chemistry, University of Delaware, Newark, Delaware

19711

EXPERIMENTAL The experiments were performed with a DuPont 21-llOB mass spectrometer modified for high pressure operation, which has been described previously (8, IO). The instrument was operated in the configuration that the electron entrance aperture and the ion exit

slit are on the same axis. The distance between the sample introduction axis and the ion exit is 3 mm. The probe (Figure 1)was designed to fit the direct introduction system of the DuPont 21-llOB mass spectrometer and for the tip to be at the previously determined optimum distance (2.5 nim) from the electron entrance/ion exit axis (8). A 2.5-cm section of 4-mm 0.d. Pyrex tubing, sealed on one end, was sealed to a carefully selected (straight and uniform diameter) piece (35 cm) of 7-mm 0.d. Pyrex tubing. The thermocouple wires were separated in the probe tip by passing them through a two-hole alumina insulator, the upper part of which was wrapped with Nichrome heating wire. The thermocouple/heater assembly was fixed in the probe tip by Cermacast with the thermocouple touching the sealed tip of the glass tubing. The tip of the probe was lightly sanded to reduce the thickness of the glass. The probe temperature is controlled by a power proportioning temperature programmer which enables the probe tip to be heated either linearly at rates from 2.0 to 30 "C/min or heated using a ballistics curve a t an average heating rate as high as 60 "C/s. For most of the experiments, a thin layer of Teflon tape was placed over the tip of the probe to reduce the possibility of surface catalyzed decomposition. Cross contamination was eliminated by use of a new Teflon layer for each sample. No significant ionic contribution to the background was observed from the Teflon. This probe is similar to one previously reported ( 3 ) . We can measure directly the temperature of the probe tip and the thermocouple responds rapidly to the gas temperature. However, we cannot report accurately the sample temperatures during rapid heating, since the pulse of heat must be transmitted through the glass to the sample and the gas temperature is significantly lower than the probe temperature. Samples were either deposited on the tip of the probe as a light dusting of the solid or from evaporation of 1ILLof dilute solution. In both cases, the samples were present on the surface as small crystals. The solvent was evaporated from the probe tip by heating to 100 "C in air until the sample appears as a dry powder on the tip of the probe. For routine analysis of unknown compounds, the probe temperature was normally increased at a linear rate of 30 "C/min to a final temperature of 400 "C. Spectra were recorded continuously throughout the heating cycle: 4.5-s scan from m / z = 28 to 600, every 6.5 s. The source and gas temperatures were normally kept at a low value (7C-100 "C) to minimize discharges through the probe as well as to reduce the possibility of thermal decomposition of the sample. Subsequent experiments have shown that the spectra are dependent on the gas temperature as well. Isopentane (99+%) was used as the reagent gas a t a pressure of approximately 0.5 Torr. The compounds studied were obtained commercially from different sources and used without further purification.

'Present address: Department of Pharmacology and Experimental Therapeutics, School of Medicine,Johns Hopkins University, Baltimore, Md. 21205.

RESULTS AND DISCUSSION One set of experiments conducted with the in-beam probe used ballistics heating to heat the sample rapidly t o a pre-

The use of in-beam sample introduction for obtaining analytically useful mass spectra of thermally labile compounds is facilitated by an Independently heatable probe. The probe is capable of heating rates from 2 'Clmin to 60 ' C I S . The performance of the probe Is compared to performance of Teflon caplllarles for sample insertion. The use of a Teflon surface on the probe has been found to increase both the absolute and relative yield of the (M H)' ions from thermally labile compounds like creatine and arginine.

+

T h e apparently simple sample introduction technique, direct source insertion, reported by Baldwin and McLafferty ( I ) , has been the subject of intensive recent work (2-8) since interest in this technique was rekindled by the work of Hunt and co-workers (9). In somewhat different modifications, the technique has also been called field desorption/chemical ionization, in-beam surface chemical ionization, direct exposure and desorption chemical ionization mass spectrometry. There are as many names as there are workers in the field, t h e results are not highly reproducible, and the desorption/ ionization process is not well understood. In our previous experiments on this technique, the samples were introduced on the tips of Teflon tubing and were heated by the hot reagent gas. The spectra obtained with this procedure were analytically interesting, but were strongly dependent on gas temperature and also exhibited a strong time dependence, perhaps resulting from competing processes of vaporization and decomposition. Several experiments were necessary for each compound t o obtain near optimum conditions for observing ions characteristic of sample molecular weight and the spectra were not highly reproducible. A more rapid and efficient method for sample handling and heating was required in order for this technique to be routinely useful. We wish to report a n independently heatable sample probe which provides rapid and reproducible control of sample temperature and which enables the use of in-beam sample introduction on a routine basis.

0003-2700/80/0352-0245$01 .OO/O

C

1980 American Chemical Society

246

ANALYTICAL CHEMISTRY, VOL. 52, NO. 2, FEBRUARY 1980

f

CU~CONSTANTAN

NICHROME ALUMINA

I

THERMOCOUPLE

I

I

15

23

/%

H E A T E R COIL

1

INSULATOR

,

10 T!T.

P

Figure 3. Time dependence of ion currents from arginine with in-beam probe. Isopentane reagent gas. Rapid ballistics heating to 410 OC at average rate of 47 "C/s

L

I

I

1

j

Figure 1. Schematic diagram of separately heatable in-beam probe I

0.02

I

I

I

I

I

FR ?E

1

E"

QATLQ.

"r

Figure 4. Ion currents from creatine vs. final probe temperature. In-beam probe. Isopentane reagent gas Linear heating rate, 30 OC/min

,

!00

I

I

320 F lML :R@M TEMF.?A7LPE

I

I

36"

~Z

+

Figure 2. Ratio of integrated ion currents of (M H)' and (M - OH)' vs. final probe temperature. Ballistics heating. In-beam probe. Isopentane reagent gas determined temperature to simulate the method described previously for Teflon capillary insertion (8). For these experiments, 1.9 pg of creatine was deposited from solution on the Teflon-coated probe tip. The samples were introduced into the source (gas temperature 100 OC) and heated to a final probe temperature at average rates of 10-15 "C/s. Scanning of spectra was initiated simultaneously with activation of the temperature programmer. The results of these experiments (Figure 2) are similar to the results obtained previously using the capillary insertion technique: an increase in sample temperature (final probe temperature) results in an increase in the absolute and relative abundances of (M + H)+ ions for a thermally labile compound such as creatine. The spectra obtained with the independently heatable probe are more reproducible than those obtained previously from a capillary heated by the gas. The spectra are time-dependent in both types of experiments, and the temperature/time profiles can be reproduced with this probe. Quantitation is likely from this probe since spectra may be obtained over the entire time of sample evolution. For the three experiments in Figure 2, for example, the integrated sample ion currents were the same, 76 f 1 X lo3. In addition, the optimization experiments can be done rapidly since the probe can be heated and cooled much more rapidly than the source. Heating rates as large as 40-60 OC/s may be used with the in-beam probe. However, the cycle time of our present equipment is too long to allow full mass range spectra to be obtained repetitively during the 5-10 s heating period.

Consequently, two masses were monitored by rapidly switching the accelerating voltage to scan over the ions. Figure 3 shows typical results of such an experiment. The (M + H)' and (M - OH)' ions ( m / z 175, 157) were monitored as 1 kg of arginine was rapidly heated (about 8 s) to a final probe temperature of 410 OC, at an average heating rate of 47 OC/s. During the rapid heating regime of the experiment, the (M + H ) + ion is the most abundant sample ion, five times as abundant as the (M - OH)' ion. These experiments accurately demonstrate the strong time dependence of the direct exposure spectra. Since our previous introduction technique did not allow accurate timing of the beginning of the rapid heating cycle, we can now see why the spectra were not highly reproducible. A moderate linear heating ramp is used for obtaining preliminary in-beam CI spectra of unknown samples, since their optimum temperature characteristics are not known. The heating rate normally chosen for these analyses is 30 OC/min. The spectra obtained using this slower rate of heating are characterized by less abundant (M + H)' ions (Figure 4) than are obtained with rapid heating. However, they are useful for locating an optimum temperature for rapid heating. A comparison was made of spectra obtained using the inbeam probe with and without the Teflon covered tip. For each of these experiments, a sample of 1.9 pg of creatine was deposited from solution on the probe. The probe temperature was increased linearly at a rate of 30 OC/min from 100 to 410 "C, at a constant source (gas) temperature of 100 "C. The results of those experiments are shown in Figure 5. These results indicate that there is less decomposition from a Teflon surface than from a glass surface. Similar surface effects have been reported previously, although the explanation is not obvious (11). In addition to a greater abundance of the (M H)+ ions, there also appears to be an increase (perhaps a factor of two) in the overall sensitivity with the Teflon surface

+

r-ANALYTICAL CHEMISTRY, VOL. 52, NO. 2, FEBRUARY 1980 7

I

247

1

C R E A T I N E / I soPENTANE

+

n

,,L

0,051

0

,

i

7

i

281 PF331

Figure 6. Matrix effects on isopentane CI spectra of creatine. In-beam

Flgure 5. Comparison of glass vs. Teflon probe surfaces. In-beam probe. Isopentane reagent gas. Linear heating rate, 30 "C/min

Table I. Total Yields (M t H)+, (M - OH)' for Glass and Teflon Probe Surfacesa Z(M -

probe surface

z ( M t H)+ m / z 132

Teflon Teflon

19.3 x 103 14.2 x 103

2 4 2 263 x io7 x io4 x

5.2 x 103 4.9 x 103

glass glass

m / z 114 ~103 103 103 103

1.9 pg creatine, heating rate: 30 "CI min, reagent gas: 0.50 Torr C.H.,. source temr,: 100 "C.

as indicated by the greater ion currents of both (M + H)' and (M - OH)+ (Table I). Perhaps it is merely the result of the greater total ion current from the samples deposited on the Teflon-coated surface, but useful spectra are obtained over a longer time period and over a wider temperature range for the samples deposited on the Teflon surfaces than for those deposited on the glass surfaces. T h e larger ratio of (M H)'/(M - OH)' with the Teflon surface perhaps results from the inability of Teflon to catalyze dehydration, whereas glass surfaces can. The overall increase in total ion current from the Teflon surface is not so readily explained, but may result from lower activation energies of desorption for samples from Teflon than from glass ( I l b ) . Microscopic examination of the samples after evaporation of the solvent, however, shows small crystals of samples on both Teflon and glass. However, at the temperatures of analysis, the sample is present as a liquid on the probe tip. If creatine is placed on the tip of the in-beam probe in an acidic matrix, like oxalic acid, essentially no (M + H ) + ions are observed in the isopentane CI spectra under conditions where abundant (M H)+ ions are observed from creatine. The samples are evolved from an acidic liquid at these temperatures. However, if the matrix is inert, like urea or ammonium chloride, no significant matrix effects are observed (Figure 6). The contrast in spectra obtained with a separately heatable conventional probe in which the sample is placed in a capillary in a heating well and in spectra obtained with the in-beam probe is shown for citric acid as the analyte in Table 11. Conventional spectra were obtained by heating the probe at a linear rate of 20 "C/min, from 100 to 400 "C. Acquisition of data was initiated at the start of the heating cycle and continued throughout the course of the experiment. In-beam spectra were obtained by depositing the sample as a light dusting on the Teflon covered probe tip and heating the probe a t a linear rate of 30 "Clmin from 150 to 250 "C. The conventional spectra of citric acid obtained from 190-300 "C are dominated by the decomposition ions corresponding to the losses of 1-2 molecules of water and CO,

+

Table 11. Citric Acid Conventional Probe Spectrum vs. In-Beam Probe Spectrum'

OH)'

' Conditions:

+

probe. Isopentane reagent gas. Linear heating rate, 30 "C/min

385 261 193 175 147

131 113 103

m lz ( 2 M + H)' ( M - H,O - CO,),H+ (M t H)' (M-OH)'

Base Peak, % conventional in-beam probe probe ( 2 1 3 "C), (200 "C), 20 "Clmin 30 "Clmin

( M H - CO, - H,O)+ (MH - CO, - 2H,O)+

' 0.45 Torr i-C5Hl,,source temp =

+

___ 14.5 -

37.4 --_ 100.0

17.3 69.1

10.5 6.7

__

100.0

_--

35.8 19.8 100 "C.

(Table 11). The (M H)+ ions is not observed in the spectra obtained with a conventional probe. The presence of an ion at m / z = 261 together with the high intensity of m / z = 131 suggests that the loss of water and C 0 2 is at least partially a thermal decomposition of the neutral species, forming a neutral species of mass 130 which is subsequently protonated to give the abundant m at mlz = 131 which reacts with the neutral compound of mass 130 to form m / z = 261. The in-beam spectra of citric acid (Table 11) contain the protonated parent, m / z 193, as the base peak throughout the temperature range studied (2OC-250 "C). The presence of the protonated dimer (2M H)' at m / z 385 is indicative of the high vapor phase concentration of the intact molecular species. The in-beam spectra contain small amounts of the characteristic fragment ions whose relative abundances increase slightly with increasing temperature. The results of the present experiments are very similar to those reported recently for "direct exposure" CI mass spectra obtained from samples deposited on a Vespel probe tip (3, 4 ) . Quantitative comparisons cannot yet be made of the spectra obtained on the three surfaces. Double maxima of the total sample ion currents were reported during the heating of a thermally labile sample, p-nitrophenyl-@-1)-glucuronide with significantly different spectra at different temperatures (3). The maxima in ion currents for (M + H ) + and the decomposition ions occurred at different temperatures ( 3 ) . Similar observations are shown in Figure 3 for arginine: a small maximum during the period of rapidly rising temperature with abundant (M + H)' ions and a subsequent, larger maximum at longer times, but at constant temperature, with spectra that show extensive decomposition. The data of Figure 3 and the observations of Cotter and Fenselau (3) indicate that the (M + H)' ions in the spectra #ofthese thermally labile compounds are produced by reactions of the reactant ion with

+

240

Anal. Chem. 1980. 52.248-252

sample molecules and not by reactions of sample fragment ions (weaker acids than the ions from the reagent gas). If the (M + H)+ ions were produced by reactions of sample fragment ions with neutral sample molecules, one would expect to observe a decrease in the ratio of (M + H ) + / ( M - OH)+ with the addition of an inert diluent to the sample. Figure 6 shows the results of experiments in which a large amount of an inert material was added to the analyte. Mixtures of creatine/urea and creatine/NH&l in the ratio of -1:lO were heated with the in-beam probe and their ion profiles compared with pure creatine. The results in Figure 6 show no evidence of a diluent effect as the curves for the mixtures are essentially the same as that obtained for the pure creatine. One can also obtain good CI spectra of thermally stable compounds with this technique. For nonpolar compounds the major difficulty arises because the samples vaporize so rapidly a t the minimum temperature at which the instrument can be operated, -50 “C. A separately heatable in-beam or direct exposure probe appears to be more useful for obtaining mass spectra of thermally labile compounds than Teflon, Vespel, or glass rods which are heated only by the reagent gas or mass spectrometer source. This separately heatable probe allows better operator control over the narrow window of time and temperature over which (M + H)+ ions can be detected and improves the re-

producibility with which one can obtain spectra of these compounds.

LITERATURE CITED (1) M. A. Baldwin and F. W. McLafferty, Org. Mass Spectrom., 7, 1353 (1973). (2) W. A. Woistenholme, U. Rapp, and H. Kaufman, Paper #083, 1979 Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Cleveland, Ohio, March 1979. (3) R. J. Cotter and C. C. Fenselau, Biomed. Mass Spectrom., 6, 287 (1979). (4) R. J. Cotter, Anal. Chem., 51, 317 (1979). (5) M. Ohashi and N. Nakayama, Org. Mass Spectrom., 13,642 (1978). (6) M. Ohashi, S. Yamada. H. Kito. and N. Nakavama. Biomed. Mass Spectrom., 5 , 579 (1978). (7) W. R. Anderson, W. Frick, and G. D. Daves, J . Am. Chem. Soc., 100,

1974 (1978) -,. (8) G:Hansen and B. Munson, Anal. Chem., 50, 1130 (1978). (9) D. F. Hunt, J. Shabanowitz, F. K. Botz, and D. Brent, Anal. Chem., 48, 1160 (1977). (IO) F. Hatch and 8. Munson, Anal. Chem.. 49, 189 (1977). (11) (a) R. J. Buehler, E. Flanigan, L. J. Greene, and L. Friedman, Biochem~stry,13,5060 (1974); (b) R. J. Buehler, E. Flanigan, L. J. Greene, and L. Friedman, Biochem. Biophys. Res. Commun., 46, 1082 (1972). \

RECEIVED for review May 29, 1979. Accepted November 5 , 1979. Taken in part from the M.S. thesis of Gordon Hansen and presented at the 26th Annual Conference on Mass Spectrometry and Allied Topics, St. Louis, Mo., June 1978. Supported in part by a grant from the National Institutes of Health, GM 25269-02.

Accuracy of Isotopic Label Calculations for Spectra with a (Molecular Ion - Hydrogen) Peak Wolfgang Benz Chemical Research Department, Hoffmann-La Roche Inc., Nutley, New Jersey 07 110

Afler a review of assumptions underlying the label calculation by mass spectrometry, theoretical spectra with superimposed random error are used to identify factors influencing the accuracy. The observable fit (calculated-observed spectrum) allows only a very crude estimate of accuracy. In deuterated compounds the presence of an M - H peak requires an additional parameter and leads to a set of nonlinear equations. A program was written that solves this problem by approximation, subject to certain ilmitations. In some cases two solutions are given. Additional criteria are discussed that will select the correct solution.

Mass Spectrometry is the most commonly used method for the determination of stable isotopes, and methods for the calculation of isotopic labels have been published repeatedly (1-3). Curiously, in most cases nothing is said about the achievable accuracy and how accuracy can be estimated. We became aware of this neglect when we extended the method to a special case which had been considered unsolvable and wanted to assess the reliability of the new procedure. Deuterated compounds with an M - H peak constitute this special case for which the standard method fails because the “M H” will split into M - H and M - D with an unknown split ratio. A program was written that finds a solution by approximation, subject to certain limitations. The first part of this paper discusses the accuracy of isotopic label calculations in general, and the latter part is concerned with deuterium 0003-2700/80/0352-0248$0 1.OO/O

determinations in the presence of an M - H peak. It turns out that some difficulties still remain. Therefore, additional criteria are discussed that help in assessing the accuracy. It seems worthwhile to briefly outline the basis of our procedure because most of the published methods are much more complicated. We treat label determinations as a special case of mixture analysis. The mass spectrum of a mixture kill allow the quantitative determination of the components, if the spectra of the pure components and their “response factors” are known. The response or sensitivity factor expresses the fact that different compounds have different ionization efficiencies, Le., an equimolar mixture will, in general, not give equal peak heights. Mathematically, this can be expressed as a set of linear equations. Let Ak be the intensity at mass k , x i the concentration of component i, Si the response factor of component i, and Fik the normalized intensity of the pure component i a t mass k . Then we can write: n

Ah = z F i k S , ~ i i=l

(1)

for a mixture of n components. Since a mass spectrum consists of many measurable peaks, say m, there are m such linear equations for n unknowns x i . As long as m Z n, this system can be solved by standard methods. Usually, m > n, Le., it is an overdetermined set. Therefore, full use of all the information is made by using a least-squares approach. This, of course, is equivalent to a consideration of experimental error and is well documented in the literature (4, 5). ‘C 1980 American Chemical Society