Microsecond-pulsed Glow Discharge Time-of-flight Mass Spectrometry: Analytical Advantages WEI HANG, CYNTHIA BAKER, B. W. SMITH, J. D. WINEFORDNER and W. W. HARRISON* Department of Chemistry, University of Florida, Gainesville, FL 32611, USA A microsecond-pulsed glow discharge time-of-flight mass mass analyser for a pulsed ion source. It has arguably the highest ion transmission among all mass spectrometers. In spectrometer was constructed and evaluated for elemental analysis. Mass spectra from the instrument show significant addition, because the TOF mass spectrometer is able to operate at a high repetition rate, a large number of spectra can be advantages, including higher signal-to-noise ratios than those of a dc glow discharge source.Important temporal advantages acquired and integrated (or averaged) in a short period of time, resulting in a significant signal-to-noise enhancement. result from the pulsed discharge and pulsed mass analyser. Mass discrimination among different elements is very small. For the observation of pulsed events, TOFMS offers the distinct advantage of multiplex detection of a high percentage The instrument currently has a resolving power of 360 in linear mode and 1600 in reflectron mode (full width at half of ions of all masses during each pulsed event.Even with a continuous source, Mahoney et al.8 have shown some promis- maximum). Present detection limits are at the low ppm level, limited primarily by the detection and data acquisition system.ing results from an orthogonal ICP-TOFMS instrument, which shows results competitive with those given by commercial Because the detector is easily saturated, the present data acquisition system has limited dynamic range and sensitivity. instruments. An earlier paper has described our ms-pulsed GD-TOFMS Possibilities exist to overcome this constraint. system design considerations and preliminary results in the Keywords: Glow discharge; pulsed glow discharge; time-of- linear mode.9 In this paper, further developments are given, flight mass spectrometry ; atomic mass spectrometry including ion optics, detection system, and data from the reflectron mode.Characteristics of this system, such as mass discrimination, temporal stability, and S/N have been meas- The application of mass spectrometry to the direct trace ured. Some standard samples have been tested for the determielemental analysis of solid samples was fairly limited until nation of the spectral resolution and detection limits, which about 10 years ago.Since then, it has become a standard are currently limited to the ppm level because of detector method in this field, basically due to the introduction of constraints. commercial glow discharge mass spectrometry (GDMS) instruments. INSTRUMENTATION The GD source has existed as an active analytical and diagnostic tool for more than 70 years,1 and it has been A diagrammatic representation of the ms-pulsed GD-TOFMS assembly is shown in Fig. 1. The whole instrument has been extensively used for the elemental analysis of solids for the past 20 years.2 A continuous GD source, either in dc mode or described in detail elsewhere.9 In the existing stage, several modifications have been made to improve the performance of in rf mode, produces a steady continuous beam of sample ions, with a small energy distribution (#10 eV). These advantages the system. One important modification was the optimization of the ion permit the manufacture of a relatively simple, and therefore more cost-effective, GDMS instrument, in sharp contrast to optics before the orthogonal extraction. High negative potential is applied to both the skimmer and slit for a higher other direct solid sampling mass spectrometers, such as laser ablation ICP-MS, spark source MS or secondary ion MS transmission, and all the other lenses are also adjusted accordingly.The sampler plate (source chamber) is fixed at ground instruments.Only recently, however, have analysts begun to extract the full potential of this low power, yet effective, source. potential. The final optimized conditions are listed in Table 1. To cope with the increased transmission, two 25 mm diam- A dc GD source is routinely operated at a power level of 1–4 W, and an rf GD source usually works at 20–50W, which eter extended dynamic range microchannel plates (MCP, Galileo Electro-Optics, Sturbridge, MA, USA), mounted in gives the same order of magnitude of signal intensity as a dc source at the lower operating power.By contrast, the instan- Chevron mode, are used in the system to replace the standard MCP assembly (Galileo) as the linear TOF detector. The taneous power of the microsecond (ms)-pulsed GD can reach several hundred watts, and the sputtering rate during the standard MCP assembly was used instead as the reflectron TOF detector. pulse-on region is about two orders of magnitude higher than that of the dc mode.3 High power also results in high excitation The increased transmission greatly increased the peak intensity of the ion signals.With the operation of the normal and ionization efficiency. In an argon gas discharge with a copper cathode, such a high instantaneous power turns the potential on the detector (1.6–2 kV), the signal intensity of the matrix can reach several volts with a 50 V impedance. The fast normally observed blue glow into a distinctive green, due to the stimulation of higher lying atomic states.A 1–4 orders of preamplifier (SR445, Stanford Research System, Sunnyvale, CA, USA), which was formerly used, is not able to handle magnitude greater signal intensity, relative to the dc source, has been observed in emission, fluorescence and mass such an intense peak. Thus, the signal from the MCP detector is connected directly to the digital oscilloscope (TDS 520A, spectrometry.4–7 With the development of ms-pulsed GD studies, it became Tektronix, Beaverton, OR, USA) with a 50 V input impedance.Ultrahigh-purity grade argon gas (99.999%, Liquid Air, San clear that the combination of a pulsed GD with a time-of- flight (TOF) mass spectrometer could be an attractive combi- Francisco, CA, USA), was used throughout. The sample used in each experiment will be described later. No cryogenic nation. In many ways, the TOFMS instrument is an ideal Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 (143–149) 143Fig. 1 Schematic diagram of ms-pulsed GD-TOFMS system. Table 1 Typical pulsed GD-TOFMS operating conditions Pulsed GD: Pulse magnitude: 1–3 kV Source pressure (argon): 0.8–1.5 Torr Pulse width: 5–20 ms Cathode–orifice distance: 4–5 mm Pulse frequency: 100 Hz Pulse current: 4–200 mA T OFMS: Orifice diameter=skimmer diameter=1 mm Orifice–skimmer distance: 4 mm Potential of L1: -800 V Flight tube: -2000 V L2: -150 V Skimmer: -400 V L3: -200 V Slit: -800 V Repelling plate bias voltage: 0 to ca.-4V Repelling pulse magnitude: 84 V (linear mode), 150 V (reflection mode) Width: 1–5 ms Microchannel-plate detector: -1600 to 1900 V Second stage pressure: 1×10-4 Torr Third stage pressure: 1.5×10-5 Torr cooling was used; hence, spectral purity may still be subject to orthogonal TOFMS instrument.9,11 Thus, the sampling efficiency of a TOF mass spectrometer is up to two orders of improvement.magnitude higher than that of a quadrupole instrument. In addition to the high sampling efficiency of the ms-pulsed RESULTS AND DISCUSSION GD-TOF mass spectrometer, this system has the opportunity for temporal separation of discharge gas species from sputtered Sampling Strategy sample components. It is believed that mass spectrometry In a sequential scanning mass analyser, such as a quadrupole samples ions primarily near the vicinity of the sampler orifice.12 instrument, after the ions pass through the interface (the Gas species (argon, nitrogen, oxygen, water, etc.), owing to sampler and skimmer), only part of the specific m/z ions will their high ionization potentials, are ionized principally by the be transmitted to the detector at any given moment.Other energetic GD electrons. These species show little ionization ions simply hit the walls and are lost. Thus, the transmission after the termination of the discharge because of the fast of a scanning instrument is only 0.1–0.01%.In a TOFMS scattering loss of electrons. On the other hand, sputtered instrument with orthogonal extraction, the sampling strategy particles diffuse across the cathode–sampler orifice distance, consists of two processes. The first is the interface sampling, are ionized by long-lived metastable argon atoms through which is similar to that used in a quadrupole mass spec- Penning collisions, and then successfully sampled by the mass trometer.The second is the ion extraction in the repelling spectrometer. Thus, most of the sputtered sample ions reach region, which is the key part of an orthogonal structure the repelling region several hundred microseconds later than instrument. The transmission of this type of instrument is the gas species. By setting the delay time of the repelling pulse determined by the speed of the ions passing through the in favour of the sample ions, interferences from gas species can repelling region and by the opening of the extraction grid (grid be greatly reduced.This process is illustrated in Fig. 2. During the pulse-on region, cathode atoms are sputtered out, but they G1 in Fig. 1).10 A 20% transmission has been observed in the 144 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12without loss of analytical signal. Therefore, this advantage offers a new approach to the analysis of samples that contain small amounts of ‘troublesome’ elements, such as C, Al, Si, S, P, K, Ca and Fe.Fig. 4 demonstrates the separation of gaseous carbon (mainly from hydrocarbons in the source) and cathodecontaining carbon by varying the delay time of the repelling pulse. Mass Discrimination In ICP-MS, because of the high pressure and frequent collisions, all ions pass through the interface at virtually the same speed. There is no concern about the transport time difference for different m/z ions from the interface to the repelling region in an orthogonal TOFMS instrument.However, GD sources operate at alow pressure (#1 Torr), and the collision frequency in the interface is also much smaller than that of an ICP. Thus, apotential problem must be considered in pulsed GD-TOFMS: smaller m/z ions have a higher velocity than larger m/z ions. Thus, different m/z ions will arrive at the repelling region at different times, which will result in some degree of mass discrimination. It was necessary to determine the extent of this problem with our system.Another potential problem, which Fig. 2 Diagram of the diffusion and transport processes in ms-pulsed has been observed in an ICP-TOFMS system,13 is the response GD-TOFMS: (a) during the pulse-on region; (b) after the termination of different m/z ions to the deflector potential. Extensive of the pulse; (c) several hundred microseconds after the pulse. experiments have been conducted to identify whether these two potential problems are significant in the present system.A mixed metal powder sample was used in this experiment are very close to the cathode surface [Fig. 2(a)]. After the (mass proportions: Al5Ti5Cu5Ag5W=151515151). The most termination of the GD, gas species ions cannot be formed, due intense Ar peak is usually observed when the repelling pulse to the rapid disappearance of electrons; sputtered atoms are delay time is#20 ms, which is close to the result from SIMION diffusing to the sampler orifice, which requires several hundred simulation (if the initial kinetic energy is set at 7 eV).If we microseconds under normal GD conditions [Fig. 2(b)]. Thus, assume that Ar ions need 20 ms to travel from the sampler to sputtered particles and gas species are physically separated the repelling region [shaded region in Fig. 2(c)], then Al ions when they arrive at the repelling region. By applying a narrow need 16.5 ms, and W ions need 50 ms. This 33.5 ms difference pulse to the repelling plate, ions in front of the extraction grid would then introduce mass discrimination into the system.[shaded region in Fig. 2(c)] will be accelerated into the flight Fortunately, the packet of mass ions from the sample has a tube, and begin their TOF mass separation. large spatial distribution. That is, the repelling pulse delay Fig. 3 is a TOF mass spectrum acquired when the temporal from the time when the mass spectrum peaks of sample ions repelling window is set to be favourable for copper ions.The initially appear to the time that those peaks have completely peak intensities of Ar and ArH are at the millivolt level, which passed is about 500 ms (when the source pressure is 1 Torr and is not evident in Fig. 3. Although, in this approach, the the cathode–sampler orifice distance is 5 mm). This large interference from gas species cannot be reduced to absolute spatial distribution is due to the diffusion time of the sputtered zero (owing to their spatial and energy distribution), it is able particles from the cathode surface to the sampling orifice and to reduce all the peak intensities of the gas species by up to 3 the initial kinetic energy distribution when the atoms are orders of magnitude (depending on the source pressure, cathsputtered from the cathode surface.Thus, compared with the ode–sampler orifice distance, and the repelling pulse delay) large spatial distribution of the ion package, the flight time differences among different elements from the interface to the Fig. 3 Demonstration of the temporal resolution of ions from sputt- Fig. 4 Carbon ion signal intensity versus repelling pulse delay time, ered atoms in a ms-pulsed GD-TOF mass spectrum of a copper disc taken at a delay time of 450 ms. Source pressure: 1 Torr argon; cathode– showing the signal from background hydrocarbons at 25 ms delay and from the cathodic carbon at 250 ms. Sample: Fe5C=9555 mass ratio; sampler orifice distance: 5 mm; pulse frequency: 100 Hz; pulse width: 15 ms; pulse magnitude: 3 kV; repelling pulse delay: 450 ms.source: 0.9 Torr, 5 mm; pulse: 100 Hz, 15 ms, 1.8 kV. Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 145Fig. 6 Mass spectra of (a) dc and (b) ms-pulsed GD. Source pressure: 1 Torr; conditions for dc: 800 V, 3 mA; pulse: 2 kV, 15 ms. Optimized cathode–sampler orifice distance for maximum Cu intensity: 7 mm (dc), 4 mm (pulse).(2.5×10-6 V)15]. Under such conditions, the signals from subppm analytes are too small to be digitized by the oscilloscope, limiting the detection of small analytical signals. An improved Fig. 5 Signal intensities of Al–Ti–Cu–Ag–W mixed disc sample versus repelling pulse delay time, demonstrating the relatively small separation data readout system is now planned for acquisition. in the repeller region. Source pressure: 1 Torr; pulse: 100 Hz, 15 ms, 2 kV; cathode–sampler orifice distance: (a) 3, (b) 5, (c) 7 mm.Stability of Signal A NIST unalloyed Cu 495 disc sample was used in this repelling region become insignificant. Experimental data show experiment, where the concentrations of Cu and Fe are 99.94% that the discrimination is small (Fig. 5). Different elements are and 96 ppm, respectively. After the sample had been pre- evenly mixed when they arrive at the repelling region. Under sputtered for 10 min (under the same pulse conditions), the normal operating conditions [Fig. 5(b)], the intensities of most measurement was begun.For a period of 1 h, the temporal elements can be optimized at the same delay time. stabilities of both the matrix and the trace elements, primarily Experiments relating to the effect of deflecting potential have long-term drift, are within 10% of the starting values (Fig. 7), been carried out at different source pressures, cathode–sampler which is close to the stability of other types of GD sources.16 orifice distances, and repelling pulse delay times.Although different operating conditions need different deflecting potentials to optimize the system, all five elements studied are optimized at the same deflecting potential. This would imply that ions of different mass have approximately the same kinetic energy after being sampled, in contrast to ICP-MS, where kinetic energy is mass-dependent. The same result was also observed in another TOFMS system where an rf-GD source was used.14 S/N The ms-pulsedGD Cu signal intensity is one order of magnitude higher than that of the dc GD, as shown in Fig. 6. Increasing the pulse power will further enhance the signal intensity in the ms-pulsed mode, while further increasing the voltage/power in the dc mode will overheat the sample. The limitation at present in reaching lower detection limits with the ms-pulsed source is the large dark current observed for the TOF system. With the Fig. 7 Temporal stability of ms-pulsed GD-TOFMS system.Sample: GD off, approximately 350 mV is measured, arising primarily NIST SRM 495 Cu; source: 1 Torr, 5 mm; pulse: 100 Hz, 15 ms, 2 kV. from the input noise associated with the digital oscilloscope Measurement started 10 min after pre-sputtering under the same conditions. [the dark current of the MCP detector is only 5×10-8 A 146 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Note that the MCP detector was not working in its linear range owing to the high intensity of the Cu signal.Resolving Power A ms-pulsed GD source permits higher resolving power and sensitivity compared with a continuous source. In the repelling region, the repelling pulse only extracts a certain volume of ions, those present in front of the extraction grid [the shaded region in Fig. 2(c)], into the flight tube. These ions have been sorted to some extent according to their speed when they move from the source to the repelling region.The ion optics (including the skimmer, three cylindrical lenses, steering plates, and slit) can shape the ion beam for less spatial distribution in the repelling region. The TOF mass spectrometer actually extracts an ion packet with limited spatial and energy distribution, which then enables the deflector plates to guide the ion package to the detector. Sputtered sample ions are sampled several hundred microseconds after the termination of the plasma. Their speed passing the repelling region is probably determined by diffusion energetics in the source rather than by the plasma potential.Because ions are formed several hundred microseconds after the termination of the discharge, their slower speed of transit across the repelling region further improves the transmission in the orthogonal extraction. On the other hand, for a continuous GD source, the ion optics after the sampler cannot efficiently focus an ion beam with a larger energy distribution. In that case, the TOF mass spectrometer extracts a volume of ions with a larger spatial and energy distribution compared with a ms-pulsed GD source, which makes it more difficult for the deflector plates to make a smooth transition of this ion packet to the detector.Some Fig. 8 Mass spectra of lead disc: (a) reflectron mode, (b) linear mode. of the ions may also strike the flight tube wall, decreasing both Source: 1 Torr, 5 mm; pulse: 100 Hz, 15 ms, 2.8 kV.the resolving power and the signal intensity. The resolving power achieved to date with our TOF instruto us. This interference was also observed with other samples ment is 1600 in the reflectron mode and 360 in the linear mode when an expanded oscilloscope scale and a high MCP voltage [Fig. 8(a) and (b)]. For a continuous source, the resolving were used. The 63Cu2563Cu65Cu dimer peak ratio is also not power is about 300. It may be noted that a small peak appears matched, because the peak height of these species just exceeds in front of every major peak in the reflectron mode [Fig. 8(a)]. the limit of the oscilloscope sensitivity. There should be a peak Several factors could contribute to the production of such a for 65Cu2 at mass 130, but the signal is too small to be sampled peak. There are two grids that act as the extraction system; by the oscilloscope. Intense 63Cu and 65Cu peaks cover a larger hence, some ions may be trapped inside the grids at the mass range than just masses 63 and 65, because of the energy repelling pulse-off time.Also, a grid is installed in front of the and spatial distribution. High detector voltage (1.9 kV) is detector, where some ions may hit it and produce electrons, probably another reason for the broad peaks, if we make a which will arrive at the detector ahead of the ions due to the comparison with the Cu peaks in Fig. 6, where the detector electric field between the grid and the MCP. Experiments are voltage is 1.6 kV.currently underway to clarify the origin of the small peak. The main limitation of this instrument at present is the detection and data acquisition system. The high sampling Mass Spectra of Standard Samples efficiency of the system results in the non-linear response of the MCP detector, which reduces the ability for semi- Fig. 9(a) is a linear mode spectrum of NIST SRM 1264a Low quantitative analysis without a calibration graph. The ana- Alloy Steel taken for an average of 1000 spectra.The concenlogue- to-digital converter has a minimum sensitivity of 40 mV trations of major and trace compositions are listed in the when the y-axis scale is set at 1 mV per division (largest upper right-hand corner of Fig. 9(a). All the constituents can expanded scale). Thus, all signals below 40 mV are simply not be clearly identified. Fig. 9(b) is a part of the spectrum of the sampled. A spectrum of 1000 averaged spectra has a small same sample with the TOF mass spectrometer operating in background which cannot be observed in the oscilloscope; the reflectron mode, where the resolving power is seen to be hence, its noise level cannot be readily measured.Based on five times better than for the linear spectrum. the high signal intensities, we anticipate a power of detection NIST SRM 495 Unalloyed Copper was also tested with this significantly below the ppm level on modification of our detec- system. The reflectron mode spectrum of 1000 averaged spectra tion system.is shown in Fig. 10. At 100 Hz, 10 s are required to acquire 1000 spectra. Signals from trace elements at the ppm level are clearly apparent. CONCLUSIONS The aim of this work was to develop an instrument capable Detection Limits of reaching low detection limits (lower than those of quadrupole- based instruments). The energetic ms-pulsed discharge Referring to Fig. 10, some problems in the spectrum are readily apparent.The 107Ag5109Ag isotope ratio is not matched well, source and high transmissive TOFMS should make the goal possible. With the use of a TOF mass spectrometer, a short perhaps due to some interference at mass 109 which is unknown Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 147Fig. 9 Mass spectra of NIST SRM 1264a Low Alloy Steel in (a) linear mode and (b) reflectron mode. Source: 1 Torr, 4 mm; pulse: 100 Hz, 15 ms, 2 kV. analysis time is expected, although the maximum frequency is last stage (presently 1.5×10-5 Torr).A matrix deflector and an energy discriminator are under construction, which will be limited to 100 Hz at present, because of the maximum averaging speed of the oscilloscope. mounted at the end of the reflectron tube, thus mitigating the MCPsaturation problem, and background noise will be further In order to determine the detection limits, a high performance preamplifier will be used in the system to increase the reduced.Even with these improvements, we still require a wider dynamic range detector which can have a linear response signal. A boxcar integrator with a minimum gate width of 2 ns (Model 4402, EG&G) will also be used for determining the for matrix and trace element signals. A wider dynamic range, more sensitive, multichannel data acquisition system is also detection limits. A 330 l s-1 turbo pump will be used to replace the 80 l s-1 turbo pump to improve the poor vacuum in the needed for multi-element analysis. 148 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Fig. 10 Mass spectra of NIST SRM 495 Cu in reflectron mode. Source: 1 Torr, 4 mm; pulse: 100 Hz, 15 ms, 1.5 kV. 6 Hang, W., Walden, W. O., and Harrison, W. W., Anal. Chem., Even with the limitations at this stage, this approach shows 1996, 68, 1148. potential as a powerful analytical technique. The temporal 7 Hang, W., Yang, P. Y., Wang, C. L., Su, Y. X., and Huang, B. L., advantage inherent to the pulsed source and analyser permits Rapid Commun. Mass Spectrom., 1994, 8, 590.strong discrimination against background discharge gases. 8 Mahoney, P. P., Ray, S. J., Li, G., and Hieftje, G. M., paper Mass discrimination of this system is also very small. Further presented at the 1996 Winter Conference on Plasma studies will include instrumentation improvement, GD param- Spectrochemistry, Fort Lauderdale, FL, ThP41. 9 Harrison, W. W., and Hang, W., J. Anal. At. Spectrom., 1996, eter optimization, determination of relative sensitivity factors, 11, 835. and isotope ratio measurements. 10 Myers, D. P., and Hieftje, G. M., Microchem. J., 1993, 48, 259. 11 Myers, D. P., Li, G., Mahoney, P. P., and Hieftje, G. M., J. Am. This research is supported by US Department of Energy, Soc. Mass Spectrom., 1995, 6, 400. Division of Chemical Sciences, University of Florida Division 12 Hess, R. K., and Harrison, W. W., Anal. Chem., 1986, 58, 1696. of Sponsored Research, and Hewlett-Packard Laboratories. 13 Myers, D. P., Li, G., Yang P., and Hieftje, G. M., J. Am. Soc. Mass Spectrom., 1994, 5, 1008. 14 Heintz, M. J., Myers, D. P., Mahoney, P. P., Li, G., and Hieftje, REFERENCES G. M., Appl. Spectrosc., 1995, 49, 945. 1 Aston, F. W., Isotopes, Longmans, New York, 1924. 15 Data sheet, Galileo Electro-Optics, Sturbridge, MA, 1996. 2 Harrison, W. W., J. Anal. At. Spectrom., 1992, 7, 75. 16 Shick C. R., Jr., and Marcus, R. K., Appl. Spectrosc., 1996, 50, 454. 3 Pollmann, D., Ingeneri, K., and Harrison, W. W., J. Anal. At. Spectrom., 1996, 11, 849. Paper 6/04454I 4 Farnsworth, P. B., and Walters, J. P., Anal. Chem., 1982, 54, 885. Received June 26, 1996 5 Huang, B., Yang, P., Lin, Y., Wang, X., and Yuan, D., Fenxi Huaxue, 1991, 19, 259. Accepted October 17, 1996 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 149