A Quadrupole Ion Trap with Cylindrical Geometry Operated in the

A cylindrical geometry ion trap is used to record mass spectra in the mass-selective instability mode. The geometry of the cylindrical ion trap has be...
10 downloads 0 Views 218KB Size
Anal. Chem. 1998, 70, 438-444

A Quadrupole Ion Trap with Cylindrical Geometry Operated in the Mass-Selective Instability Mode J. Mitchell Wells, Ethan R. Badman, and R. Graham Cooks*

Department of Chemistry, Purdue University, West Lafayette, Indiana 47907

A cylindrical geometry ion trap is used to record mass spectra in the mass-selective instability mode. The geometry of the cylindrical ion trap has been optimized to maximize the quadrupole field component relative to the higher-order field content through field calculations using the Poisson/Superfish code and through experimental variation of the electrode structure. The results correspond well with predictions of the calculations. The trap has been used to record mass spectra with better than unit mass resolution, high sensitivity, and a mass/charge range of ∼600 Th. Multistage (MS3) experiments have been performed, and the Mathieu stability region has been experimentally mapped. The performance of this device compares satisfactorily with that of the hyperbolic ion trap.

is also currently exploring this topic. There have been no reports on the use of cylindrical traps in the standard mass-selective instability mode of operation. The CIT geometry was patented by Langmuir et al. in 1962.16 Our interest in cylindrical traps is derived from the ease with which these devices may be machined compared to hyperbolic traps. This feature makes cylindrical traps attractive for miniaturization, provided adequate performance is available. Fieldportable17-19 and even smaller scale micromachined mass spectrometers20 are receiving increasing attention as a means for providing rapid, sensitive screening for environmental contaminants, to name just one example. The goal of our research is to optimize the geometry of the CIT and to characterize the best available performance of the device as a preliminary step to the investigation of its usefulness as a small-scale instrument.

Quadrupole ion traps were introduced by Paul in 19531 and have since developed through several stages to their current state of relatively high performance and increasing popularity.2-4 Alternative geometries for Paul ion traps have been suggested,5 and in the past, cylindrical ion traps (CIT) received the attention of a number of research groups. Most of this interest in cylindrical traps has been for use as a simple ion storage device.6-10 Exceptions are studies by Bonner et al.11 and by Mather et al.12 in 1977 and 1980, respectively, and some earlier work by Benilan and Audoin.13 In addition to our efforts,14 at least one other group15

EXPERIMENTAL SECTION Figure 1 shows the cylindrical ion trap in one of the versions implemented in this study. The toroidal ring electrode of the hyperbolic trap is approximated by a simple cylinder, while the two hyperbolic end cap electrodes are replaced with planar electrodes. Each end cap has one center hole, which is 2 mm in diameter. The trap electrodes and mounting flanges were machined in-house from 304 stainless steel and are supported on alumina rods. Electrode spacing and electrical isolation is achieved with machinable glass ceramic (Macor, Corning Inc., Corning, NY) spacers. The stainless steel support flange for the electrode assembly also houses an electron ionization filament block as used in the Finnigan ITMS (Finnigan Corp., San Jose, CA) and other Finnigan ion trap instruments. The filament end cap was machined to allow the ITMS electron gate lens to be mounted and used to control the ionizing electron beam. The cylindrical traps examined here were mounted in a prototype Finnigan ITMS instrument which has been described elsewhere.21 This instrument is mounted on rails in a large vacuum chamber and uses an rf trapping voltage with a frequency of 1.1 MHz and

(1) Paul, W.; Steinwedel, H. Z. Naturforsch. 1953, 8A, 448. (2) Practical Aspects of Ion Trap Mass Spectrometry; March, R. E., Todd, J. F. J., Eds.; CRC Press: Boca Raton, FL, 1995; Vols. 1-3. (3) March, R. J. Mass Spectrom. 1997, 32, 263-276. (4) Jonscher, K. R.; Yates, J. R. Anal. Biochem. 1997, 244, 1-15. (5) Beaty, E. C. J. Appl. Phys. 1987, 61, 2118-2122. (6) Mikami, N.; Miyata, Y.; Sata, S.; Toshiki, S. Chem. Phys. Lett. 1990, 166, 470-474. (7) Mikami, N.; Sato, S.; Ishigaki, M. Chem. Phys. Lett. 1991, 180, 431-435. (8) Mikami, N.; Sato, S.; Ishigaki, M. Chem. Phys. Lett. 1993, 202, 431-436. (9) Grebner, T. L.; Neusser, H. J. Int. J. Mass Spectrom. Ion Processes 1994, 137, L1-L6. (10) Ji, Q.; Davenport, M. R.; Enke, C. G.; Holland, J. F. J. Am. Soc. Mass Spectrom. 1996, 7, 1009-1017. (11) Bonner, R. F.; Fulford, J. E.; March, R. E.; Hamilton, G. F. Int. J. Mass Spectrom. Ion Phys. 1977, 24, 255-269. (12) Mather, R. E.; Waldren, R. M.; Todd, J. F. J.; March, R. E. Int. J. Mass Spectrom. Ion Phys. 1980, 33, 201-230. (13) Benilan, M. N.; Audoin, C. Int. J. Mass Spectrom. Ion Phys. 1973, 11, 421432. (14) Gill, L. A.; Wells, J. M.; Badman, E.; Cooks, R. G. Proceedings of the 45th ASMS Conference on Mass Spectrometry and Allied Topics, Palm Springs, CA, June 1-5, 1997; p 127. (15) Arkin, R. C.; Laude, D. A. Proceedings of the 45th ASMS Conference on Mass Spectrometry and Allied Topics, Palm Springs, CA, June 1-5, 1997; p 123.

(16) Langmuir, D. B.; Langmuir, R. V.; Shelton, H.; Wuerker, R. F. U.S. Patent No. 3,065,640, 1962. (17) Hemberger, P. H.; Alarid, J. E.; Cameron, D.; Leibman, C. P.; Cannon, T. M.; Wolf, M. A.; Kaiser, R. E. Int. J. Mass Spectrom. Ion Processes 1991, 106, 299-313. (18) Hemond, H. F. Rev. Sci. Instrum. 1991, 62, 1420-1425. (19) Wise, M. B.; Thompson, C. V.; Buchanan, M. V.; Merriweather, R.; Guerin, M. R. Spectroscopy 1993, 8, 14-22. (20) Orient, O. J.; Chutjian, A.; Garkanian, V. Rev. Sci. Instrum. 1997, 68, 13931397. (21) Louris, J. N.; Cooks, R. G.; Syka, J. E. P.; Kelley, P. E.; Stafford, G. C.; Todd, J. F. J. Anal. Chem. 1987, 59, 1677-1685.

438 Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

S0003-2700(97)01198-0 CCC: $15.00

© 1998 American Chemical Society Published on Web 01/01/1998

Figure 1. (a) Concept and (b) experimental implementation of a cylindrical ion trap as an approximation to the hyperbolic Paul ion trap.

a maximum peak amplitude of 7500 V. Mass-selective instability scans22 are effected by ramping the amplitude of the rf at a nominal rate of 5555 Da/s so that ions of increasing mass are successively made unstable, ejected from the ion trap through an aperture in the end cap electrode, and detected by an external electron multiplier. A supplementary ac signal source is also available on this instrument to allow axial modulation during mass analysis and resonant excitation to induce collisional activation during tandem mass spectrometry experiments. The geometry of the CIT was varied with respect to the z0 dimension, the closest distance between the end cap and the center of the trap, and with respect to the ring/end cap spacing, the closest distance between the ring and end cap electrodes (Figure 1b), to minimize the higher-order field contributions to the trapping field. These factors were investigated by changing the size of the Macor spacers and by using ring electrodes of various sizes. However, the internal radius, r0, of the ring electrode was held constant at 1.0 cm. Calculations of the electric field in the CIT were used to guide the selection of dimensions used in the construction of the CITs studied experimentally. The fields were calculated as the dimensions were varied using the Poisson/Superfish code23 maintained at Los Alamos National Laboratory. The program runs on a PC platform and allows harmonic analysis of the field created in a system of specified geometry through a Fourier series approximation (see Results and Discussion section for more detail on the calculational procedure). This analysis gives the relative contributions made to the field in the device by the quadrupole term and higher-order field terms (octapole, dodecapole, etc.). Krypton and the headspace vapors of organic compounds were leaked into the vacuum manifold via Granville-Phillips (Granville Phillips Co., Boulder, CO) leak valves. Organic compounds were obtained from Aldrich (Aldrich, Milwaukee, WI) and were degassed through three freeze-pump-thaw sequences before each

use. Sample pressures were typically 1 × 10-6 Torr (uncorrected, 1 Torr ) 133 Pa) as read from a Bayard-Alpert-type ionization gauge. Helium buffer gas was admitted to the manifold through a Granville-Phillips leak valve to nominal pressures of between 2 and 8 × 10-4 Torr.

(22) Stafford, G. C.; Kelley, P. E.; Syka, J. E. P.; Reynolds, W. E.; Todd, J. F. J. Int. J. Mass Spectrom. Ion Processes 1984, 60, 85-98. (23) Billen, J. H.; Young, L. M. Proceedings of the 1993 Particle Accelerator Conference Vol. 2, 1993; pp 790-792.

(24) Syka, J. E. P. In Practical Aspects of Ion Trap Mass Spectrometry; March, R. E., Todd, J. F. J., Eds.; CRC Press: Boca Raton, FL, 1995; Chapter 4. (25) Wang, Y.; Franzen, J. Int. J. Mass Spectrom. Ion Processes 1994, 132, 155172.

RESULTS AND DISCUSSION Optimization of the Geometry. It has been shown that truncation of the electrodes, machining imperfections, and apertures in the electrodes cause the trapping field in the hyperbolic Paul trap to deviate from a pure quadrupolar field.24,25 A general expression for the potential in a real device can be derived from a solution to the Laplace equation and is given in eq 1 in spherical ∞

Φ(F,θ,φ) )

∑A F P (cos θ) 0 n n n

(1)

n)0

polar coordinates (F, θ, φ). In this expression, the A0n are weighting factors, and Pn(cos θ) are Legendre polynomials. When the summation in eq 1 is written out with FnPn(cos θ) expressed in cylindrical coordinates, the more familiar expression for the potential shown in eq 2 results. The A0n terms for n ) 0-4

1 3 Φ(r,z,φ) ) A00 + A01z + A02 r2 - z2 + A03 r2z - z3 + A04 2 2 34 r - 3r2z2 + z4 + ... (2) 8

(

)

(

(

) )

correspond to the monopole, dipole, quadrupole, hexapole, and octapole components, respectively. Only when the coefficients for n ) 0 and n ) 2 are nonzero is the field purely quadrupolar. Truncation of the electrodes and other imperfections cause the higher-order coefficients to have nonzero values, resulting in a

Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

439

Figure 2. (a) Contribution of the octapole and dodecapole terms relative to the quadrupole term as a function of z0 with r0 ) 1.0 cm and ring/end cap gap of 1.0 mm calculated using the Poisson/Superfish code; (b) calculated field lines for a CIT with z0 ) 0.897 cm.

field that changes nonlinearly from the center; hence real traps are necessarily nonlinear devices. Nonlinearities in the field result in resonances that can cause undesirable ion losses26-30 although they can also be exploited to improve ion trap performance.31,32 Benilan and Audoin13 and Bonner et al.11 have both shown that the solution for the Laplace equation for the field in a cylindrical ion trap is more appropriately described by Bessel functions than Legendre polynomials. The latter authors developed an expression for the potential in the cylindrical trap for the case where the end caps are grounded and the trapping rf voltage is applied to the ring electrode. Their expression is given in eq 3, where pn ∞

Φ(r,z) )

∑B I (p r) cos(p z) n 0

n

n

(3)

n)0

is given in eq 4, Bn are weighting factors and I0 is a zero-order

pn ) (2n + 1)π/2z0

(4)

modified Bessel function of the first kind. Despite the differences in the form of the potential in the hyperbolic and cylindrical traps, the experimental observation is that cylindrical traps are capable of trapping ions and of producing mass spectra in the mass-selective stability mode.11,12 This is explained by numerical calculations by Bonner et al. of the fields in the cylindrical trap, and by Beaty5 of the field in a related trap, (26) Guidulgi, F.; Traldi, P. Rapid Commun. Mass Spectrom. 1991, 5, 343348. (27) Morand, K. L.; Lammert, S. A.; Cooks, R. G. Rapid Commun. Mass Spectrom. 1991, 5, 491. (28) Guidugli, F.; Traldi, P.; Franklin, A. M.; Langford, M. L.; Murrell, J.; Todd, J. F. J. Rapid Commun. Mass Spectrom. 1992, 6, 229-231. (29) Eades, D. M.; Yost, R. A. Rapid Commun. Mass Spectrom. 1992, 6, 573578. (30) Alheit, R.; Kleineidam, S.; Vedel, F.; Vedel, M.; Werth, G. Int. J. Mass Spectrom. Ion Processes 1996, 154, 155-169. (31) Franzen, J. Int. J. Mass Spectrom. Ion Processes 1993, 125, 165-170. (32) Franzen, J. Int. J. Mass Spectrom. Ion Processes 1994, 130, 15-40.

440

Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

which show that the shapes of the field in both traps are similar to that in the hyperbolic trap. We have calculated the field in a number of geometries of the cylindrical device to determine its harmonic content and to minimize the contributions of higherorder field components, i.e., to minimize Bn in eq 3 for values of n other than 0 and 2. The Poisson program23 used for this purpose accepts an input file that defines a set of electrodes via lists of r, z coordinates and also specifies the voltage applied to each electrode. The space between and around the electrodes is covered with a triangular mesh, and the potential at each point on the mesh is calculated from the applied voltages using Ampere’s law. The negative derivative of a least-squares fit to the potential at each point is used to give the electric field. An iterative process is used to provide an optimum coverage of the specified volume. The multipole content of the field may be calculated by representing the field as a Fourier series and solving for the coefficients by interpolating the potential over an arc through the specified volume. As is evident from the results of this study, this procedure seems to be an excellent guide in field optimization even though the accuracy of the calculations is not known. Figure 2 shows the contribution of octapole and dodecapole potential components to the trapping potential relative to the quadrupole component calculated by Poisson as the z0 dimension is varied, while r0 is held constant at 1.0 cm and the ring/end cap spacing is held constant at 1.0 mm. These data indicate that the octapole component of the field is minimized at a z0 value of ∼0.9 cm, while the dodecapole component is still relatively large. Calculations at larger values of z0 than those given in Figure 2 show that the dodecapole component does not approach zero until the value of z0 is nearly 2 cm, by which time the octapole potential component has again become quite large in the negative direction. Poisson calculations were also conducted at a fixed z0 of 0.897 cm and r0 of 1.0 cm while the ring/end cap spacing was varied. These calculations (data not shown) indicate that the smaller this gap, the lower the higher-order content of the field. One millimeter was deemed the lower practical limit for this gap in

Figure 3. Experimental optimization of the CIT geometry. For all spectra, Pbromobenzene ) 1 × 10-6 Torr, PHe ) 8 × 10-4 Torr, and ionization time, 500 µs. For (a) and (b), z0 was held constant at 0.783 cm, while the ring/end cap spacing was varied from 3.3 to 1.0 mm. No axial modulation was used. For (c) and (d), the ring/end cap gap was held constant at 1.0 mm. (c) z0 ) 0.783 cm and axial modulation at 470 kHz, 2 V. (d) z0 ) 0.897 cm and axial modulation at 450 kHz, 2.5 V.

the experimental work: beyond this point the danger of arcing of the trapping voltage to the end caps was too great. Additional Poisson calculations in which the hole in each end cap was considered showed a minimum higher-order field content at a z0 value of 0.909 cm. It was found that this additional stretch had little effect on overall performance with respect to resolution; however, it did affect mass assignment as discussed later. Figure 3 shows mass spectra of the molecular ion region of bromobenzene and demonstrates the improvement in resolution obtained by varying the geometrical parameters. This was done by using ring electrodes of various sizes and by changing the size of the Macor spacers. The abscissa on this and subsequent mass spectra are in terms of nominal mass charge units as calibration was not performed and m/z values were simply obtained from the data system display. It should further be noted that both the axial modulation frequency and amplitude play a significant role in the nominal m/z ratio at which peaks are displayed by the data system. Minimizing the ring/end cap gap at a fixed z0 value of 0.783 cm increases the separation of the two peaks (Figure 3a and b), while “stretching” the trap to a z0 of 0.897 cm and holding the gap constant at 1.0 mm provides further improvement to unit mass resolution (Figure 3c and d). It is important to note that the spectra in Figure 3a and b were taken without axial modulation, while Figure 3c and d utilize axial modulation to further improve resolution. Comparison of panel b to c, which were taken using the same CIT without and with axial modulation, respectively, indicate that axial modulation does little to improve the resolution of the CIT when the geometry is not optimized.

Figure 4. Mass spectrum of n-butylbenzene recorded with a cylindrical ion trap with z0 ) 0.897 cm, r0 ) 1.0 cm, Pn-bb ) 1 × 10-6 Torr, PHe ) 8 × 10-4 Torr, ionization time, 100 µs, and axial modulation at 425 kHz, 3 V.

The optimum performance of the CIT with respect to resolution is illustrated in Figure 4, which shows the mass spectrum of n-butylbenzene recorded with a CIT with z0 ) 0.897 cm and a ring/end cap gap of 1.0 mm. The peaks at 91 and 92 Th are fully resolved, and the full width at half-maximum for both peaks is ∼0.3 Th. Experiments with perfluorotributylamine (PFTBA) show that the mass scale is linear with rf voltage, and so calibration is readily achieved. Figure 5 shows the mass spectrum of PFTBA using the cylindrical trap with z0 ) 0.897 cm and a ring/end cap Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

441

Table 1. Mass Shift (Experimental - Theoretical, (0.05 Da) for Nitrobenzene and Acetophenonea theoretical mass (Da) nitrobenzene acetophenone a

Figure 5. Mass spectrum of PFTBA with a cylindrical trap with z0 ) 0.897 cm, r0 ) 1.0 cm, PPFTBA ) 1 × 10-6 Torr, PHe ) 2 × 10-4 Torr, ionization time, 1 ms, and axial modulation at 465 kHz, 3 V. Inset: calibration plot of ejection time vs true mass, R 2 ) 0.999 998.

gap of 1.0 mm. This demonstrates both that a mass range of at least 600 Th is achievable with the CIT and that mass calibration is straightforward. The inset shows the mass calibration plot of ejection time vs true mass for PFTBA, with an R2 of 0.999998. Further experimental characterization of the higher-order field content of the trapping field was conducted by making careful measurements of the masses of the nitrobenzene and acetophenone molecular ions. These ions have been observed to have a mass shift of ∼-0.5 Th and ∼-0.7 Th, respectively, in an ideal (r0 ) 1.00 cm, z0 ) 0.707 cm) hyperbolic ion trap on a relative mass scale determined under particular operating conditions.24,33 These shifts were attributed, inter alia, to the presence of higherorder field components caused by the truncation of the electrodes and were alleviated by stretching the z0 of the device to 0.783 cm, which introduced a positive octapolar field component.24 Cleven et al. showed that when the structure of the electrodes was disrupted further, by machining slots in the electrodes to allow entrance of a laser beam for photodissociation studies, the mass shifts increased for the z0 ) 0.783 cm trap, and a further stretch to z0 ) 0.870 cm was then necessary to compensate for the additional higher-order field effects.33 To characterize mass shifts in the CIT, PFTBA ions of mass 69, 100, and 131 were each isolated in rf/dc fashion, the ionization time was adjusted to give a constant peak area (constant number of ions trapped), and the ejection time was measured from the amplified electron multiplier signal which was observed on a TDS 540 digital oscilloscope (Tektronix Inc., Beaverton, OR) triggered at the start of the rf mass analysis ramp. The exact masses of these ions and their ejection times were used to construct a calibration line. The ejection times of the nitrobenzene molecular ion (m/z 123) and the acetophenone molecular ion (m/z 120) were measured in the same way as the calibrant ions, and their experimental masses were then calculated from the calibration line. Although these experiments did not match the numbers of ions in the trap or the He pressures used in the earlier hyperbolic trap experiments,24,33 mass shifts of similar magnitude were observed for nitrobenzene (33) Cleven, C. D.; Cooks, R. G.; Garrett, A. W.; Nogar, N. S.; Hemberger, P. H. J. Phys. Chem. 1996, 100, 40-46.

442 Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

123.1123 120.1519

mass shift z0 ) 0.897 cm z0 ) 0.909 cm -0.38 -0.52

-0.32 -0.23

All ions have unit charge.

and acetophenone. The results for the CIT with z0 ) 0.897 cm and a ring/end cap gap of 1.0 mm, and for a further stretch of z0 to 0.909 cm with a ring/end cap gap of 1.0 mm, are shown in Table 1. These results indicate that the stretch to 0.909 cm has reduced the mass shift for these two ions and so has clearly affected the field, probably by removing some of the negative octapole component. These results also indicate that, with respect to mass accuracy, further optimization of the CIT geometry is required. This may be achieved via further stretches of z0, or by “shimming” of the field with small auxiliary electrodes, as has been demonstrated for linear quadrupoles with circular rods and for cylindrical Penning traps.34-36 For the data shown in Figures 3 and 4, unit mass resolution was achieved by increasing the He pressure to 8 × 10-4 Torr, which is too high for long-term operation of an electron multiplier. An alternative structure to that shown in Figure 1 was explored therefore to alleviate the need for very high buffer gas pressures. The end cap electrodes were modified to take the form of end caps, with sidewalls that fit over the ring electrode, and Macor rings are placed between the ring and end cap electrodes (Figure 6). This structure effectively seals the volume of the ion trap and requires that gas be leaked directly into the trap volume through a Teflon line butted up against a small hole drilled through the sidewall of the end cap closest to the ion source. The high pressures inside the trap required for maximum resolution can be obtained without having to operate with the manifold at high pressures, and hence, the electron multiplier and turbomolecular pump can be operated under standard, safe conditions. Poisson calculations showed that the effect of the modified electrodes on the higher-order field components is minimal in comparison to the unmodified electrodes; therefore, a direct comparison of performance with and without the modified end caps is possible. Figure 6 shows a diagram of the modified CIT and compares the m/z 91 and 92 region for n-butylbenzene for (Figure 6b) the conventional CIT at the standard 2 × 10-4 Torr He pressure in the manifold and (Figure 6c) the “gastight” CIT at 8 × 10-5 Torr, both experiments being performed with axial modulation. The gastight CIT shows a significant improvement in resolution and it is noteworthy that it can be operated at a manifold pressure less than that of the conventional CIT. Because gas is introduced directly into the trap, an estimate of sample and He pressure is not available. Stability Region. To better characterize the cylindrical ion traps used in this study, and to facilitate tandem mass spectro(34) Gabrielse, G.; Mackintosh, F. C. Int. J. Mass Spectrom. Ion Processes 1984, 57, 1-17. (35) Gabrielse, G.; Haarsma, L.; Rolston, S. L. Int. J. Mass Spectrom. Ion Processes 1989, 88, 319-332. (36) Tan, J.; Gabrielse, G. Appl. Phys. Lett. 1989, 55, 2144-2146.

Figure 6. (a) Modification to the CIT end caps to limit gas conductance out of the cell. For (b) Pn-bb ) 1 × 10-6 Torr, PHe ) 2 × 10-4 Torr, ionization time, 200 µs, and axial modulation at 450 kHz, 2 V; (c) Pmanifold ) 8 × 10-5 Torr, ionization time, 200 ms, and axial modulation at 400 kHz, 4.5 V. (b) Conventional CIT; (c) gastight CIT.

Figure 7. Stability diagram determined for a CIT with z0 ) 0.897 cm and r0 ) 1.0 cm by monitoring the abundance of m/z 84 from Kr as a function the amplitude of the rf and dc potentials applied to the ring electrode. Axial modulation at 460 kHz, 4 V.

where r0 is the internal radius of the hyperbolic ring and r1 is the internal radius of the cylindrical ring. In addition, Mather et al.12 have experimentally mapped the stability diagrams for several geometries of the CIT and found them to be similar to that of the hyperbolic ion trap. The stability diagram obtained for a CIT with z0 ) 0.897 cm and a ring/end cap gap of 1 mm is shown in Figure 7. The m/z 84 isotope of krypton was isolated in rf/dc fashion and allowed to cool for 5 ms using the sample krypton, pressure 1 × 10-6 Torr, for this purpose. No He buffer gas was used for this experiment. Axial modulation, 460 kHz, 4 V, was used in order to sharpen the stability boundaries. The rf and dc voltages applied to the ring electrode were then manipulated through the ITMS software and the point at which the m/z 84 peak was just distinguishable from the noise (S/N ) 2) was recorded as the stability boundary. The agreement between the CIT stability region and that of the hyperbolic trap is reasonably good. Johnson et al.37 reported that when mapping the stability diagram for a hyperbolic trap, the voltage actually delivered by the supply was lower than the voltage requested by the software, thus causing a shift in their experimental diagram which is similar to that seen in Figure 7. No attempt to calibrate the dc supply was made in this work, but it is expected that such a calibration will improve the correspondence of the theoretical and measured stability diagrams. In addition, the rate at which the dc and rf voltages

change can affect the shape of the experimental diagram, and in particular this accounts for the deviation of the experimental βz ) 1 boundary, as also noted for hyperbolic traps by Johnson et al. Tandem Mass Spectrometry. The CIT has been used to record tandem mass spectra for a number of model compounds including PFTBA and acetophenone. MS3 data for acetophenone taken with a cylindrical trap with z0 ) 0.897 cm, are shown in Figure 8. The parent ion, m/z 120, was isolated using the rf/dc isolation method. By ramping the rf voltage and applying an offset dc voltage to the ring electrode at the same time, the ion of interest is placed at the apex of the stability diagram. Under these conditions, all other ions but those of the single mass-to-charge ratio of interest are destabilized and ejected from the trap. The resonance excitation frequencies used were found by calculating the q value the parent ion would have in a hyperbolic trap of the same internal radius, calculating the associated frequency, and then empirically optimizing the frequency and amplitude to give the best fragmentation. The last step is exactly the same procedure used with the hyperbolic trap, and the adjustments needed were of similar magnitude in the two devices. The isolated parent ion (m/z 120) is shown in Figure 8a (note the lowabundance ion of m/z 105 formed during or after isolation). After the parent ion had been isolated and allowed to cool for 5 ms, collision-induced dissociation was effected via resonant excitation with an ac potential of 1 V at 118 kHz, applied between the end cap electrodes for 20 ms. The product ions were allowed to cool for 5 ms, and the resulting mass spectrum is shown in Figure 8b. The highly abundant m/z 105 fragment ion was then isolated (Figure 8c), brought to a qz value of 0.3, and fragmented in the same way with an ac potential of 120 kHz, 1.8 V, to give the sequential product spectrum38 shown in Figure 8d. Note that these data were taken at a helium pressure of only 2 × 10-4 Torr and not at the higher pressures that give the best resolution as previously mentioned, which accounts for the broadness and slight tailing to higher mass of the peaks. In summary, a cylindrical ion trap used in the mass-selective instability mode with optimized geometry gives sufficiently good

(37) Johnson, J. V.; Pedder, R. E.; Yost, R. A. Rapid Commun. Mass Spectrom. 1992, 6, 760-764.

(38) Schwartz, J. C.; Wade, A. P.; Enke, C. G.; Cooks, R. G. Anal. Chem. 1990, 62, 1809-1818.

metric experiments using the CITs, the stability region of the optimized geometry trap with z0 ) 0.897 cm and r0 ) 1.0 cm was experimentally mapped. Benilan and Audoin13 have shown that the Mathieu stability parameters a and q can be approximated by two parameters R and χ using the relationships given in eq 5,

Ru ) au(r0/r1)2, χu ) qu(r0/r1)2

(5)

Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

443

Figure 8. MS3 data for acetophenone using a CIT with z0 ) 0.897 cm and axial modulation at 450 kHz, 4 V. (a) Isolated molecular ion, (b) product ion spectrum resulting from resonant excitation of m/z 120 using 118 kHz, 1 V, (c) isolated MS/MS product ion at m/z 105, and (d) sequential product scan due to CID of m/z 105 using resonant excitation with 120 kHz, 1.5 V. Pressure of acetophenone, 1 × 10-6 Torr; helium pressure, 2 × 10-4 Torr; ionization time, 4 ms.

performance to warrant further consideration as a mass analyzer. Using the Poisson/Superfish code for PCs, higher-order fields were minimized through calculations upon which the experiments were based. The calculations provide information that served to guide the optimization procedure. Mass shifts of nitrobenzene and acetophenone that result from nonlinear higher-order fields were measured, and it was found possible to reduce their magnitudes by adjusting the geometry of the ion trap. In terms of resolution, sensitivity, and mass range, the CIT compares well with the performance attainable with a typical hyperbolic ion trap, but the simplicity and relative ease of design relative to a hyperbolic ion trap make it an ideal candidate for miniaturization, (39) Soni, M.; Frankevich, V.; Nappi, M.; Santini, R. E.; Amy, J. W.; Cooks, R. G. Anal. Chem. 1996, 68, 3314-3320.

444 Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

particularly in conjunction with alternative modes of operation, such as nondestructive image current detection.39 Current efforts involve optimization of the rf trapping frequency and improvement in the mass range and resolution using established methods. Future work will include nondestructive detection and construction and characterization of smaller scale devices. ACKNOWLEDGMENT This work was supported by the Office of Naval Research Grant N00014-94-K-2002. Received for review December 5, 1997. AC971198H

October

28,

1997.

Accepted