Introduction

Mass spectrometry has become one of the most widely used analytical methods because of its its high sensitivity, selectivity, and fast analysis. Mass spectrometers (MS) have historically been large, laboratory-confined instruments. Though the motivation for developing smaller MS was first driven by space exploration [1, 2], MS that are small enough to be portable now find their way into such diverse fields as environmental monitoring [3,4,5], threat detection [6,7,8], oceanography [9, 10], and clinical diagnosis [11,12,13]. Although portability typically requires a reduction of the physical size, weight, and power consumption of the whole instrument, miniaturizing the mass analyzer often leads to reduced weight, size, and power of other system components. Previous efforts have miniaturized a variety of conventional mass analyzers, including time-of-flight (TOF) [14,15,16], magnetic sector [17, 18], and quadrupole mass analyzers [19,20,21]. Radio frequency (rf) ion traps [22,23,24,25] are of particular promise because of several key characteristics: their simple structures allow less stringent electrode alignment; their tolerances to high pressure that help reduce the vacuum requirements; and their compact size and low weight. Although the resolution in ion traps is typically lower than that in other mass analyzers, they have nonetheless seen a fast pace of development for portable MS [12, 26,27,28,29].

Several issues frequently arise in development of miniaturized mass analyzers. For instance, electric breakdown is more likely as electrode surfaces are closer together and pressures are higher [30]. Electronic stability (rf and amplitude) must meet more stringent absolute requirements when using lower voltages and smaller physical dimensions [31]. The absolute mechanical tolerances must be tighter on the smaller dimensional scale of smaller analyzers [32], often pushing the limits of conventional machining techniques. Space-charge effects are a greater concern in analyzers with smaller dimensions [31, 33], often resulting in decreased sensitivity. With ion traps, this reduced sensitivity has been recovered in two ways: either by using parallel arrays of traps or by using traps with inherently larger storage volumes due to an extended trapping dimension (e.g., linear, rectilinear, or toroidal traps).

Compared with quadrupole ion traps and cylindrical ion traps, toroidal ion traps have a compact structure and a larger trapping capacity for the same characteristic trapping dimension [25], making them appealing for developing a portable MS [26, 34], though mechanical fabrication and assembly may limit miniaturization [35]. The 2D linear ion trap (LIT) also effectively expands the trapping capacity [36], although performance may be more sensitive to geometric deviations, including misalignment of electrodes [23]. The rectilinear ion trap (RIT) substitutes flat surfaces for the hyperbolic electrodes of the LIT, simplifying fabrication for a given level of machining precision [24, 37]. Ion trap arrays based on cylindrical [38,39,40], linear/rectilinear [41, 42], and either in 2-D [42,43,44,45] or 3-D quadrupoles [46, 47] have been extensively explored and developed. A common feature of trap arrays is the involvement of microfabrication techniques to produce very small traps [31].

Microfabrication, including photolithography and micropatterning, provides high precision and accuracy in two dimensions, but limited capabilities in the third (out-of-plane) dimension. To leverage the high precision of in-plane features and reduce or eliminate dependence on out-of-plane fabrication, we previously demonstrated an approach to making ion traps using two plates, with patterning on the facing surfaces of each plate to produce the needed trapping potential [48,49,50]. Several trapping geometries were demonstrated, including a linear ion trap [51]. This two-plate linear ion trap included sets of electrodes lithographically patterned on the facing surfaces of two ceramic (Al2O3) plates. Ions were ejected through laser-drilled slits in the plates. Capacitive voltage dividers created different rf amplitudes for each electrode. The plates were positioned 4–6 mm apart in initial experiments [52], and 1.9 mm apart in later experiments [52], and near-unit mass resolution was demonstrated in both cases. However, subsequent efforts to further miniaturize this two-plate LIT ran into several issues. Foremost among these were difficulties in ion ejection through an increasingly narrow ejection slit.

Although the dimensional scale of lithographic patterning can easily be reduced to make smaller features and smaller ion traps in the two-plate approach, the plate thickness cannot be reduced without making the resulting device more fragile. However, using a plate that is thick compared with the trapping dimensions can lead to problems with ion ejection. In previous efforts (unpublished), we developed and tested a two-plate LIT with plate spacing of 724 μm (y0 = 362 μm). The ejection slit was fabricated to be 100 μm wide and 20 mm long. The plate thickness was still 0.5 mm as with the larger devices, resulting in an ejection slit that was many times deeper than it was wide. Unfortunately, ions were not successfully ejected through this narrow, deep slit. Two factors likely prevented ion ejection: ions traveling at small angles during ejection would not make it through the slit but would hit the walls, and build-up of charge on the inside walls of the slit repelled subsequent ions. In a LIT, ions can pick up energy in the non-ejection direction, which has a secular frequency close to the ejection frequency, resulting in ions being ejected with some angular dispersion. In addition, depending on operating frequency and voltage, ion trajectories in miniaturized ion traps can be larger, relative to the available space between electrodes, than in full-size traps [30], exacerbating the issues with the narrow ejection slit.

One possible solution to the above problem is to eject ions parallel to the plates rather than through a slit. However, simulations show that it is more difficult to obtain desirable combinations of higher-order terms in the trapping field when ions are ejected between and parallel to the plates. Early results from the halo ion trap [49, 53] showed the same difficulties. A slit allows ions to be ejected with a potential in the ejection direction that is carefully controlled all the way to the point of ejection from the trap.

We have opted for another solution, which is to taper the ejection slit profile. The slit walls open out, allowing ions to pass through even with a large angular dispersion. Charge build-up on the slit walls is thereby prevented, and ion ejection can proceed. The tapered slits profile also allows the slit walls to be coated with a conducting or semiconducting material, which is impractical in straight-walled slits. Tapered or open-out slits are used on other ion traps made using metal electrodes, including the original Finnigan quadrupole ion trap, the linear ion trap [23], and the rectilinear ion trap [24]. In these cases, the slit is not to reduce charge build-up but simply to improve the efficiency of ejection for ions emerging with some angular dispersion.

The previous plate design also had several issues in making electrical connections between the front and back side of the plates, and in making electrical connection with the printed circuit board behind the ceramic plates. These issues were solvable at the dimension scale of the full-size device, but were difficult to miniaturize further. Both of these issues are solved in the present design by patterning the connections on the trapping side of the plates and wire-bonding the connections from ceramic plate to printed circuit board.

The above issues presented significant challenges in ion trap miniaturization using the two-plate approach. This article demonstrates that a tapered ejection slit combined with single side patterning and connection solves these issues, enabling further miniaturization of this mass analyzer. In addition, performance is significantly improved even for the full-scale device, in terms of mass resolution and also ruggedness and ability to operate for long periods of time without signal loss due to charge build-up. Further, in characterizing this new design, one of the two plates was mounted on a motorized translation stage so that plate spacing could be optimized in vacuum during operation.

Experimental

Design and Fabrication

Whereas our prior efforts all used ceramic substrates, in the present study we used borosilicate glass. Glass has similar electrical and structural properties as alumina ceramic, but is more amenable to the tapering cuts used to make the ejection slit. The fabrication process is shown in Figure 1. Plates were milled and diced to 57 × 37 mm, with a thickness of 0.50 mm. A 2.5-mm long, 0.50 mm-wide ejection slit was machined in the glass with a back side taper angle of 45°. Holes were milled into the plates for plate alignment during assembly. Aluminum electrodes and traces were patterned onto the plates using photoresist and photomasks. After patterning, a 100 nm Ge layer was evaporatively deposited on both sides of the plates, covering both the trapping side and the back side of each plate.

Figure 1
figure 1

(a) The work flow of the fabrication process; (b) simulation model in SIMION 8.1; (c) cross-section view of a complete plate assembly with trapped ions

Full size image

The electrode pattern is shown in Figure 1b. Five pairs of rf electrodes are used on each plate, with the positions and dimensions identical to those of the previous ceramic plate design, with a few exceptions. This was done to allow comparison between the designs to be as close as possible. Axial confinement of ions was accomplished using patterned end bars on each plate. The PCB and glass plate were assembled as shown in Figure 2a, b.

Figure 2
figure 2

(a) Photo of a complete glass plate with PCB; (b) the assembly of glass plate and PCB with plate holder; (c) ion trap assembly, there are several pogo pins on the thin PCB so that conduction is made by pressing it on the back side of the thick PCB

Full size image

The specifications and the optimized rf amplitudes for the five pairs of rf electrodes are shown in Table 1. Electrodes #0 surrounded the ejection slits, and were grounded on one plate but floated (nominally at ground) with an applied AC resonance waveform on the other, electrodes #1 were always grounded, #4 and #5 were always applied with the full rf driving voltage, and the voltages applied on #2 and #3 electrodes were scaled as 0.14 times of the full driving amplitude to adjust the shape of the trapping field [54].

Table 1 The Geometric Specification and the Optimized Voltage Distribution on Each Electrode
Full size table

After patterning, each glass plate was attached to a printed circuit board (PCB) using vacuum-grade epoxy (Torr-Seal; Agilent Technologies, Lexington, MA, USA) as seen in Figure 2a, b. The connection pads of the plate were then wire-bonded to corresponding pads on the PCB. In addition, a second PCB with capacitors was attached to the back side of the mounting PCB to establish rf amplitudes and conduct the applied waveforms to the trapping electrodes. Although fairly fragile by themselves, glass plates bonded to the PCBs and mounted in an ion trap assembly can withstand routine handling in the laboratory with no problems.

Mounting and Assembly

Previous designs relied on mounting screws and sapphire balls (for plate separation and alignment), resulting in a rigid assembly. To provide flexibility with the present design, one of the two plates was mounted onto a translation stage and the other plate was fixed. Plate alignment was adjusted prior to putting the instrument in vacuum. Plate spacing was controlled using a motorized stage and was adjustable in vacuum. The assembly is shown in Figure 2c. To mount the moveable plate on the motorized stage (MAX 343, Thorlab Inc., Newton, NJ, USA), a plate holder was designed that can be attached to the glass plate and its support PCB. The stage was mounted on a base, and the top plate was fixed onto a ceiling piece. The bottom plate was immobilized on the stage together with and right above an electron multiplier (Detector Technology, Inc., Palmer, MA, USA). An electron gun (Torion Inc., American Fork, UT, USA) was mounted on the side of the trap with the electron beam directed along the axis of the trap. The two plates were pre-aligned using four sapphire balls (Swiss Jewel Company) before the bottom plate was completely immobilized on the stage. The bottom plate was first loosely placed above the detector, and the two plates were separated apart so that the four balls could be placed in the alignment holes on the bottom plate. The bottom plate was then moved closer to the top plate by the actuator to the position where the four balls were seated in the alignment holes on the top plate. This position was stored in the motorized stage controller. Next, the bottom plate was immobilized and the plates were separated apart again to remove the alignment balls. Finally, the bottom plate was moved back to the stored position. The repeatability of the piezo positioners is 50 nm in each direction. However, larger variances in position are expected when plates are replaced due to imprecision of mounting the plates in the positioning system.

Experiments

The electronic timing control of resonance ejecting ions was similar to previous experiment [51] but with a shorter period (60 ms for the entire experiment). The ionization time was reduced to 0.5 ms; ions were trapped with the rf at 200 V0-p and cooled for 1.825 ms. Ions were then ramped out by sweeping the rf voltage from 200 to 560 V0-p. An AC resonance waveform was applied during the rf ramp with an optimized frequency of 620 kHz and amplitude of 1 V0-p. A +5 V DC voltage applied to the two end bars contained the ions axially inside in the trap.

The sample used in this experiment was a mixture of toluene and deuterated toluene (D8-toluene) at a volume ratio of 1:1, and a mixture of xylenes (o-, m-, p-). Sample pressure for toluene and D8-toluene was 1–2 × 10−6 Torr, whereas the sample pressure for xylene mixtures was 4–5 × 10−6 Torr, nearly three times higher than that of toluene mixtures, so that the fragment ions m/z 91 as well as their isotopic peak m/z 105 and m/z 107 could be seen within a wider range of plate spacing distances. The electron gun (e-gun) acceleration voltage was −62 V for toluenes and was decreased to −52 V for xylenes to reduce the fragmentation. A −100 V/+23 V pair of voltages were used on the gate to block or focus the electron beam. The electron multiplier was used with −1600 V applied. As plates and coaxial cables added about 100 pF to the circuit load, the drive frequency was reduced to 1.35 MHz instead of the rf power supply’s built-in frequency of 1.6 MHz (Ardara Technologies, L.P., Lewiston, NY, USA). The background pressure was in the 10−7 Torr level. Sample was preloaded in a glass tube that was connected to a leak valve (203 variable leak; Granville Phillips, Boulder, CO, USA). Sample vapor was leaked into the vacuum. Helium at 3–4 × 10−3 Torr was introduced for collisional cooling of the trapped ions. Variations of plate spacing and scan rate were explored to observe the effect on signal intensity and mass resolution. Computer simulations utilizing SIMION 8.1 were used to plan experiments and anticipate optimal conditions (voltages, plate spacing, etc.).

Results and Discussion

Simulations showed an optimized plate spacing of 4.4 mm with driving frequency at 1.6 MHz, while best resolution was achieved in practice when the spacing was 5.00 mm with the same applied rf. Although end cap stretching can improve performance in many ion traps because of improvement in higher-order terms of the trapping field [55], these higher-order terms were taken into account in the simulations. It may be that the fields were not exactly as intended, or that some other factor is present that was not adequately accounted for in the SIMION simulations. Such factors may include the effect of the deposited germanium layer, or fabrication tolerances of the slit. We observed small plate-to-plate differences, including a different optimal AC frequency when different plates were used, which may be the result of fabrication variance or inconsistent alignment.

The spectra for the toluene mixture and xylenes are shown in Figure 3. The scan rate was 2300 Th/s. In Figure 3a the toluene molecular ion peak at m/z 92 and the H-loss peak at m/z 91 were completely resolved, as were the corresponding peaks at m/z 98 and 100 from D8-toluene. Peaks at m/z 105, 106, 107, 112, 113, and 114 are expected products of ion-neutral reactions (methyl abstraction) within the trap, including the various combinations of deuterated and non-deuterated species [48]. In Figure 3b, the molecular ion peak at m/z 106 from xylene is well resolved from the H-loss peak at m/z 105 and the 13C peak at m/z 107. The S/N ratio in the xylene spectrum was higher than that in toluene mixture spectrum as the xylene pressure was much higher [56].

Figure 3
figure 3

Representative spectra of toluene and D8-toluene (a) and xylenes (b). Peak widths were calculated using a Gaussian curve fitted to the peaks in Matlab

Full size image

Spectra at different plate spacings for the toluene mixture and xylenes are shown in Figures 4 and 5, respectively. The AC frequency, AC amplitude, end-bar voltage, rf trapping voltage, and rf were same as above and constant for all these experiments. Plate spacing in this experiment can be changed without having to break vacuum, so a relatively constant pressure was maintained, greatly reducing changes in signal intensity attributable to sample or helium pressure. Thus, the variation of signal intensity should be due primarily to differences in plate spacing. In the process of collecting toluene mixture spectra, the smallest spacing was 4.8 mm as the peak intensity was too weak to observe if the plate spacing was closer than 4.8 mm. When the spacing was increased from 4.8 mm to 5.0 mm, a clear increase of signal intensity was observed, similar to the xylene mixture. On the other hand, the resolution in toluene mixture spectra seemed to vary little at small spacing from 4.8 mm to 5.0 mm, compared with the resolution variation in xylene mixture spectra when the spacing was increased from 4.2 mm to 5.0 mm. When the plate spacing was further increased from 5.0 mm to 5.5–5.6 mm, peak intensity from toluene (m/z 91 and 92) started to decrease more obviously than ion peaks from D8-toluene. A clear decrease in resolution was also seen in spectra of both samples, indicating there was a maximum point for both toluene mixture and xylene mixture within the variation range. That optimal point occurs at 5.0 mm under the current conditions, at which the minimum peak width (full width of a peak at its half maximum height, FWHM) was 0.35 Th in toluene spectra (m/z 98) and 0.32 Th in xylene spectra (m/z 106), corresponding to a resolution of 119 and 140, respectively. The variation in resolution with plate spacing was probably due to the variation in higher order components of the trapping field at different plate spacings. The variation in peak intensity might be because the e-gun was fixed relative to the stationary plate, and was not always directed precisely along the trapping center. However, a more likely explanation is that the ion ejection efficiency changed with the magnitude of the higher-order field components. With a greater contribution of octopole and dodecapole, ions more quickly fall out of resonance during AC excitation, reducing the fraction that gain sufficient energy to be ejected at that point in the scan. Another phenomenon observed is that when the plate spacing is increased, all the peaks gradually shift to higher ejection rf voltages (as seen in Figures 4 and 5), consistent with the increased effective trap size (y0).

Figure 4
figure 4

Spectra of the toluene mixture with plate spacing ranging from 4.8 mm to 5.5 mm

Full size image
Figure 5
figure 5

Spectra of the xylene mixture with plate spacing ranging from 4.2 mm to 5.6 mm

Full size image

Figure 6 shows the fraction of octopole and dodecapole as a function of plate spacing, calculated using SIMION and using a previously published procedure for calculating the potential on the germanium surface [52]. This procedure cannot be used for ion trajectory calculations and differs from the procedure used above to predict performance. Interestingly, the optimal experimental performance occurs when the octopole is still negative relative to the quadrupole, and even the sum of octopole and dodecapole is slightly negative.

Figure 6
figure 6

Higher order components (octapole and dodecapole) as a fraction of quadrupole field at different plate spacings

Full size image

It is well known that resolution in ion traps is dependent on the scan rate [57].

Figure 7 shows the spectra for toluene and D8-toluene at different scan rates ranging from 1700 to 12,000 Th/s, with each spectrum averaged 50 times. Plate spacing was always at the optimal value of 5.0 mm. It can be seen that the resolution (FWHM) was better with slower scan rate. At 3700 Th/s, the m/z 91 and 92 peaks started to combine. Peaks at m/z 98 and 100 remain resolved even with reduced resolution.

Figure 7
figure 7

Toluene and D8-toluene spectra at different scan rates and spacing of 5.0 mm

Full size image

An issue with previous LITs made using patterned plates is that charge build-up in the slits prevented long-term operation. Signal intensities would decrease over a timescale of minutes, and would return to normal only after the system was turned off for several tens of minutes. Although the cause of this behavior was previously unclear, the absence of the problem with the present design points to charge build-up as the culprit. With the present design we saw no reduction in signal even after many hours of continuous operation. This likely is the result of the tapered slit because other aspects of the design have not significantly changed (e-gun operation, trap design). From this we also conclude that the signal reduction shown in Figures 4 and 5 is likely the result of differences in ejection efficiency due to different electric fields with each value of plate spacing, and not due to charge build-up.

The above results demonstrate that the tapered slit design is effective at solving one of the main limitations of miniaturizing the two-plate LIT. Specifically, the taper allows ion ejection to be unimpeded by either collision with the slit wall or by build-up of charge on the slit wall, without compromising plate ruggedness by requiring that the plate thickness scale with the trapping dimensions. In addition, the improved electrical connections with the new design proved to be more reliable and easier to miniaturize. The tapered slit can be machined much narrower than what was used in the present study; smaller versions have already been fabricated, patterned, and mounted, and are being prepared for testing.

Conclusions

Key improvements for a two-plate planar LIT were demonstrated to work and to provide solutions encountered in efforts to miniaturize the previous design. Better than unit resolution was achieved by optimizing the plate spacing under a given voltage distribution at a scan rate of 2300 Th/s. This new design consistently showed stable spectra with good resolution. The fabrication process for this one-side ion trap is easier and faster, and the connections are more durable than those of the old one, again an advantage for further miniaturization efforts.