Anal. Chem. 2000, 72, 5787-5791
Space Charge Effects on Resolution in a Miniature Ion Mobility Spectrometer Jun Xu, William B. Whitten,* and J. Michael Ramsey
Oak Ridge National Laboratory, P.O. Box 2008, MS 6142, Oak Ridge, Tennessee 37831
Miniaturization of ion mobility spectrometry (IMS) is expected to have many advantages, as well as difficulties, in the separation of chemical species at atmospheric pressure. We report the results of studies of a miniature ion mobility spectrometer that has a drift channel 1.7 mm in diameter, the smallest cross section reported to date. The miniature cell contains a homogeneous drift field and is operated at atmospheric pressure. The miniature IMS has been characterized by measuring both negative and positive ion spectra using a frequency-quadrupled Nd: YAG laser on samples of NO, O2, and methyl iodide; a useful resolution (>10) was achieved with an operating voltage of 500 V. Peak broadening due to Coulomb repulsion was determined to have a major effect on the resolution of the miniature device. The purpose of this paper is to show that Coulomb repulsion effects can be an important contribution to pulse broadening in miniature ion mobility spectrometers. Ion mobility spectrometry (IMS) is capable of separating ionic species at atmospheric pressure and has demonstrated usefulness in detecting chemical warfare agents, explosives, and drugs.1-4 There is a consensus among researchers that miniaturization of such instruments is desirable for the next generation of IMS.5,6 While several commercial hand-held IMS devices have been demonstrated, fundamental questions associated with miniaturization still remain, such as the minimum cross section of the drift channel, limitations on resolution, and other technical barriers. This work describes experiments with a miniature IMS that has a drift channel of 1.7 mm in diameter, the smallest diameter reported to our knowledge. We have shown that the resolution of this device is moderate, but sufficient for separating certain molecules. The resolution was found to be dependent on ion concentration, decreasing with (1) Eiceman, G. A.; Karpas, Z. Ion Mobility Spectrometry; CRC Press: Boca Raton, FL, 1994. (2) Wu, C.; Siems; W. F.; Asbury, G. R.; Hill, H. H., Jr. Anal. Chem. 1998, 70, 4929-4938. (3) Taylor, S. J.; Piper, L. J.; Connor, J. A.; FitzGerald, J.; Adams, J. H.; Harden, C. S.; Shoff, D. B. The Proceedings Supplement of the Sixth Int. Symp. On Protection against Chemical & Biological Warfare Agents, Stockholm, Sweden, May 10-15 1998; pp 81-89. (4) Baumbach, J. I.; Eiceman, G. A. Appl. Spectrosc. 1999, 53, 338A-355A and references therein. (5) Baumbach, J. I.; Berger, D.; Leonhardt, J. W.; Klockow, D. Int. J. Environ. Anal. Chem. 1993, 52, 189-193. (6) Miller, R. A.; Eiceman, G. A.; Nazarov, E. G.; King, T. A. Technical Digest, Solid-State Sensor and Actuator Workshop, Hilton Head Island, SC, June 4-8, 2000; p 120. 10.1021/ac0005464 CCC: $19.00 Published on Web 10/21/2000
© 2000 American Chemical Society
increasing initial ion density. We attribute this effect to Coulomb repulsion of the ions as they transit down the drift tube to the detector. Just as for condensed-phase ion mobility measurements, i.e., electrophoresis, the resolving power of IMS under diffusion-limited conditions is independent of drift length. Resolution depends primarily on the electric potential drop experienced by the ions from the point of injection to detection. However, as the dimensions of the ion drift channel are reduced, it becomes increasingly difficult to achieve diffusion-limited conditions. This paper explores resolving power issues associated with small-diameter drift tube devices. Resolution (R) is generally defined by the ion drift time (td) divided by the full width at half-maximum (w) of the ion peak detected. Factors that determine resolution include the following: (1) initial ion pulse width, (2) broadening by Coulomb repulsion between ions in both the ionization and drift regions,7,8 (3) spatial broadening by diffusion of the ion packet, and (4) ionmolecule reactions in the drift region.9 If ion-molecule reactions are neglected, the square of the full width at half-maximum (w) is determined by
w2 ) (∆ti)2 + (∆tC)2 + (∆tD) 2
(1)
where ∆ti includes the initial pulse width, ∆tC is the temporal broadening induced by Coulomb repulsion in both the ionization region and drift region, and ∆tD is the diffusional broadening. Laser ionization was adopted for testing the miniature IMS system because of its simplicity.10 The laser radiation was focused into a small region with enough intensity for multiphoton ionization of the analytes. However, ions produced in this small region can generate a relatively large amount of Coulomb repulsion, especially in the initial stages of drift. If the effective volume of the laser ionization is assumed be to a cylinder with an initial radius of r0 (0.2 mm) and a length of h (3 mm), from Gauss’ law, the electric field, Er, induced by space charge at the edge of the ion cylinder is,
Er ) Nze/(2π0hr0)
(2)
where N is the total number of ions per laser pulse, e is the charge (7) Spangler, G. E.; Collins, C. I. Anal. Chem. 1975, 47, 393-402. (8) Spangler G. E. Anal. Chem. 1992, 64, 1312-1312. (9) Revercomb, H. E.; Mason, E. A. Anal. Chem. 1975, 47, 970-983. (10) Lubman D. M.; Kronick, M. N. Anal. Chem. 1982, 54, 1546-1551.
Analytical Chemistry, Vol. 72, No. 23, December 1, 2000 5787
Figure 1. Experimental diagram of the miniature ion mobility spectrometer.
of the electron,0 is the permittivity of free space, and z is the charge state of the ion. N is estimated to be ∼1 million ions per pulse for the 0.1-mJ laser power used; therefore, Er is ∼50 V/cm. This field is comparable to the drift field, which is ∼200 V/cm. The initial ion packet expands along the radius of the ionization cylinder due to the space charge field with a velocity, dr/dt ) KEr, while the packet drifts with velocity, Vd ) KEd, where K is the ion mobility. Solving this differential equation for ions drifting from the location of the laser beam to a distance L, we obtain the square of the radial expansion,
(∆r)2 ) r2 - r02 ) NzeL/(π0hEd)
(3)
where r is the radius of the expanded cylinder and Ed is the drift field. This expansion results in a temporal variance,
(∆tC)2 ) Nze/(π0h)(td2/V)
(4)
where td is the drift time and V is the drift bias. Assuming a Gaussian distribution for the ion diffusion in the drift channel, the root-mean-square displacement of the ions is given by (2Dtd)1/2. Replacing the diffusion constant, D, with the Einstein relation, KkT ) zeD, the broadening introduced by diffusion ∆tD is9
(∆tD)2 ) 16 ln2 kTtd2/zeV
(5)
where k is the Boltzmann constant, T is the drift temperature in kelvins, and V is the drift bias. For 200 °C and an 800 V drift bias, the predicted resolution would be 48 under diffusion-limited conditions. Moreover, under diffusive-limited spatial spread of the ion packet, the diameter of the drift channel could be made as small as 0.27 mm in a 35-mm-length IMS device without a considerable loss of ions. Typically, commercial atmospheric pressure IMS units have a resolution of 30 or less.11 Part of the reduction in resolution from the predicted value is believed due (11) Cohen, M. J.; Karasek, F. W. J. Chromatogr. Sci., 1970, 8, 330.
5788
Analytical Chemistry, Vol. 72, No. 23, December 1, 2000
to the Townsend energy factor, η,12 which represents an increase in the effective temperature due to the increased kinetic energy imposed by the drift field. Taking the Townsend factor into account and substituting eq 4, eq 1 can be rewritten as
w2 ) (∆ti)2 + [Nze/(π0h) + η16 ln2 kT/ez]td2/V (6)
It should be noted that the coefficient of td2/V in eq 6 includes both Coulomb repulsion and enhanced thermal diffusion terms. EXPERIMENTAL SECTION Figure 1 shows the experimental setup for the miniature IMS, which has a 1.7-mm-diameter drift channel, 35 mm in length. The channel is comprised of 25 stacked metal electrodes separated by insulator spacers. The electrodes are made of oxygen-free highdensity copper (OFHC) with a thickness of 0.33 mm, an inner diameter of 1.7 mm and an outer diameter of 15.7 mm. The insulator spacers are made of Macor with an inner diameter of 3 mm, outer diameter of 15.7 mm, and thickness of 1.1 mm. Note that the inner diameter of the metal electrodes is smaller than that of the insulator spacers, which is intended to minimize charging of the internal surface of the channel. The IMS sensor is combined with two 35-W cartridge heaters (Omega Engineering, Stanford, CT) and a thermocouple for temperature control in a range between 22 and 150 °C. Connected between the electrodes to form a voltage divider were 24 miniature resistors (DIGI-KEY, Thief River Falls, MN), which were chosen to have 1 MΩ resistance each with 1% uncertainty. A power supply (PS350, Stanford Research Systems, Sunnyvale, CA) provided the drift voltage to the end electrode, which was distributed to sequential electrodes through these resistors. A simulation calculation (SIMION13) showed that the potential as a function of the longitudinal distance at the center of the channel should decrease linearly with less than 0.1% deviation, much smaller than the variations induced by the 1% (12) McDaniel, E. W. Collision Phenomena in Ionized Gases; Wiley: New York, 1964; Chapter 11. (13) Dahl, D. A. SIMION 6.0; Idaho Natl. Eng. Lab., INEL-95/0403, 1995.
uncertainty in the values of the miniature resistors. The linearly decreasing potential indicated that a nearly homogeneous electrostatic field was produced in the drift region. The electrode next to the detector plate was ac-grounded through a capacitor so the electric current due to ion movement would not be coupled to the detector. The apparatus was baked at 120 °C overnight in a flowing Ar atmosphere to minimize detection of background ions. Sample gas was introduced through a mass flow controller (MKS Instruments, Andover, MA) with a flow rate of 7.5 standard cubic centimeters per minute (sccm). Sample gases used were 0.8 ppm NO in Ar and vapor from a CH3I liquid carried by N2. The N2 drift gas was purified by a filter (Lab Clear, Scientific Instrument Services, Ringoes, NJ) and regulated by a flowmeter (MKS-500 sccm). The drift gas passed through a specially designed spiral channel and then flowed in a direction counter to the ion flow. The spiral channel was made in such a way that it yielded a 900mm flow length inside the 35-mm-length IMS. This spiral channel allowed the drift gas to reach an equilibrium temperature with the spectrometer prior to entering the drift channel. Sample and drift gases exited from the ionization region through a valve attached to a pump which controlled the pressure in the IMS in the range between 380 and 1000 Torr. Ultraviolet laser radiation (266 nm) from a Nd:YAG laser (Quanta-Ray, model DCR2, Spectra-Physics, Mountain View, CA) or from an excimer laser (248 nm, PSX-100, MPB Technologies, Dorval, PQ, Canada) entered the ionization region through a quartz window, intercepted the sample gas, and exited through another quartz window. The laser pulse width from the Nd:YAG laser was 5 ns, and the repetition frequency was 10 Hz. Laser photoionization provides efficient ionization for large molecules with primarily molecular ions being formed..14,15 The laser was operated with an energy of less than 0.1 mJ/pulse and the beam was focused to a spot ∼0.4 mm in diameter in the ionization region for most of the measurements. To ionize NO molecules, a twophoton process is needed. The free electrons generated by photoionization were captured by NO, O2, and CH3I to form negative ions. The excimer laser, with a maximum energy of 1.6 mJ/pulse was used for a study of output pulse width as a function of laser pulse energy. Photoionization ions entered the drift region, were separated according to their mobilities, and reached a detector plate (OFHC) located at the end of the IMS channel. The ion current was sent to a homemade current amplifier coupled with a digital oscilloscope (TDS410A, Tektronix, Wilsonville, OR). The current amplifier was based on an Analog Devices preamplifier (model 549LH). The current amplifier had a gain of 0.08 V/nA with a 1-kHz bandwidth. The oscilloscope was triggered by the laser pulse, which served as a start signal for the mobility spectra. The oscilloscope digitized and averaged the mobility spectra. The spectra were subsequently stored by an Apple Macintosh computer running Labview software. RESULTS AND DISCUSSION Figure 2 shows ion mobility spectra of negative ions detected by the miniature IMS under laser radiation of NO, NO + impurity, (14) Dietz, T. G.; Duncan, M. G.; Liverman, M. A.; Smalley, R. E. Chem. Phys. Lett, 1980, 70, 246-250. (15) Lubman, D. M.; Naaman, R.; Zare, R. N. J. Chem. Phys., 1980, 72, 3034.
Figure 2. Ion mobility spectra of negative ions using laser ionization: (a) NO, (b) NO + impurity, and (c) NO + impurity + CH3I. The drift bias used for (a), (b), and (c) are -500, -500, and -800 V, respectively. The drift times in (a) and (b) to that of (c) are rescaled (see text).
and NO + impurity + CH3I mixtures. All these spectra were averaged over 20 scans. In Figure 2a, a single peak appeared when 0.8 ppm nitric oxide was introduced into the ionization region with Ar as the drift gas. The spectrum was obtained at room temperature. The gas pressure in the drift region was adjusted to ∼1 atm, and the drift potential was -500 V. NO molecules were ionized by a two-photon ionization process, while the background Ar was not ionized under the current laser conditions. The free electrons that were liberated by photoionization were captured by NO since the molecule has a positive electron affinity (0.025 eV). Positive ions were also observed when a positive bias was applied. However, the positive ion peak was relatively broad, most likely due to reaction ionization between NO+ and the residual moisture in the system. For the purpose of testing the separation power of the miniature IMS, we focused on the negative ions since the number of species of ions was limited in this case by electron affinity. IMS spectra were also measured at 64 °C, which showed a resolution slightly lower than that at room temperature. Since our primary interest is development of IMS operated at ambient temperature, the IMS spectra presented below were measured under room-temperature conditions. After switching from Ar to N2 drift gas, two ion peaks appeared in the IMS spectrum, as shown in Figure 2b. The peak at longer Analytical Chemistry, Vol. 72, No. 23, December 1, 2000
5789
Figure 3. Measured temporal peak width squared as a function of tD2/V.
drift time is attributed to NO- since its intensity varied accordingly to the NO flow rate. The earlier peak is believed to be due to residual O2 in the drift gas because (1) as the flow rate of the drift gas was increased, the earlier peak intensity increased accordingly, and (2) O2 has 0.45-eV positive electron affinity. The initial N2 gas was contaminated by 1.2 ppm oxygen and 1.1 ppm of moisture. After passing through the filter, the impurity concentration was ∼50 ppb. Figure 2c shows the ion mobility spectrum as CH3I gas was added to the ionization region. The CH3I molecules were generated from the evaporation of liquid CH3I from a vial that was attached in the sampling line. The concentration was estimated from the known vapor pressure to be a few tens of ppb. A third peak that appeared in Figure 2c is most likely I- since the electron affinity of iodide is larger than the bonding energy between CH3 and I. This is similar to Schumate and co-workers’ finding that all brominated alkanes yield electron-attached species (Br-).16 The drift bias was increased to -800 V for this spectrum to obtain a higher sensitivity. However, as discussed later, the resolution for this bias is worse than that at -500 V. Since the three spectra were obtained at different pressures and different biases, which affect the mobility of the samples, we rescaled the drift times in (a) and (b) to that of (c) for comparison of these spectra. In Figure 3 we have plotted the square of the full width at half-maximum (w2) as a function of td2/V for NO- ions produced with different potentials applied to the drift region. This plot is expected to be a linear relation as indicated by eq 6, similar to that observed in previous work.17,18 A linear fit to the data shows that the square of the initial time width is 0.58 ms2, corresponding to an initial time spread of 0.76 ms, much longer than would be predicted from the axial extent of the ionizing laser. This initial time width is believed to be dominated by the spatial extent of the electron swarm broadened by the drift field in the ionization region before the electrons were captured. The resolution was measured for the NO- peak as a function of drift bias, as shown in Figure 4. The resolution was not (16) Shumate, C.; St. Louis, R. H.; Hill, H. H., Jr. J. Chromatogr. 1986, 373, 141. (17) Rokushika, S.; Hatano, H.; Baim, M. A.; Hill, H. H., Jr. Anal. Chem. 1985, 57, 1902-1907. (18) St. Louis, R. H.; Hill, H. H., Jr. Anal. Chem. 1990, 21, 321-355.
5790
Analytical Chemistry, Vol. 72, No. 23, December 1, 2000
Figure 4. ontributions of initial pulse width (dot-dash), Coulomb repulsion (dot), and diffusion (dash) to resolution as a function of drift bias. Solid line is a combination of all three contributions. The filled circles are experimental data.
improved by increasing the potential applied to the drift region but, instead, reduced. This result is similar to previous observations17 for the case of a larger initial pulse width. To obtain an expression related to resolving power, we divide eq 6 by td2, as in ref 2:
R-2 ) Ri-2 + RC2- + RD-2
(7)
Ri ) L2/(∆tiKV)
(8)
RC ) (π0hV/Nze) 1/2
(9)
RD ) (zeV/16 ln2 KT)1/2
(10)
where
The estimated individual contributions to the resolution of initial pulse width, Coulomb repulsion, and diffusion, eqs 8-10, as well as the total, eq 7, are also plotted in Figure 4. For these estimates, we have assumed an initial pulse width of 0.76 ms, obtained from Figure 3. The linear fit of Figure 3 has a slope of 1.95 V, and according to eq 6, this slope is the sum of the Coulomb and diffusion contributions to the broadening. Since the Townsend energy factor is close to unity for atmospheric pressure,19 the diffusion contribution to the slope at room temperature is 0.28 V, so the Coulomb contribution is 6 times higher than the diffusion contribution, under these assumptions. The resolution at higher bias voltage approaches the ratio of drift time to the initial time spread and thus decreases inversely with voltage. For low-bias conditions, Coulomb repulsion and diffusion become the dominant contributions, with the estimated diffusion broadening being small compared to that generated by Coulomb repulsion under our (19) Siems, W. F.; Wu, C.; Tarver, E. E.; Hill, H. H.; Larsen, P. R.; McMinn, D. G. Anal. Chem. 1994, 66, 4195-4201.
Figure 5. Instrumental resolution as a function of laser pulse energy.
conditions. From eq 6 with h ) 3 mm, a Coulomb contribution to the slope in Figure 3 of 1.7 V requires an initial pulse of 9 × 105 ions. The number of ions in a pulse at the detector electrode can be obtained independently from the output current integrated over the duration of the pulse. A typical IMS pulse had peak amplitude of 75 pA and 1.43-ms width, giving a value of 0.67 × 106 ions in the output pulse. This value should be somewhat smaller than the input pulse because of the losses to the walls of the drift tube. The good agreement between the two values is confirmation that Coulomb repulsion is a major contribution to pulse broadening in the experiments. The Coulomb contribution to the resolution depends on the total number of ions initially generated and thus should vary with the energy of the ionization laser pulse at low fluences. We measured the resolution as a function of laser energy for different drift biases, as shown in Figure 5. For each bias, the resolution
was found to be lower at higher laser pulse energy. This effect was significantly larger at the lower bias voltage. As indicated in Figure 4, for a field higher than that generated by -600 V, the initial time spread dictated the resolution and the effect of Coulomb repulsion was less important. These results support the Coulomb repulsion hypothesis predicted in eq 9. In conclusion, we have presented the results of studies of a miniature ion mobility spectrometer with a drift channel 1.7 mm in diameter by 35 mm long. The miniature cell featured a homogeneous drift field and could be operated at atmospheric pressures and room temperature. The miniature IMS has been characterized by measuring both mobility spectra of negative and positive ions produced by a frequency-quadrupled Nd:YAG laser with samples of NO, O2, and methyl iodide. Resolution was modest but sufficient to resolve the ions studied. To make the device more compact, replacement of laser ionization with a corona discharge source is being explored. Future devices may be possible that exploit the manufacturing advantages of photolithography and micromachining. Microfabricated IMS devices could find application as process and field chemical sensors where conditions are well characterized. These miniature devices may also be relevant to more complex analysis scenarios when coupled to microfabricated gas- or liquid-phase chemical separation devices or even microscale mass spectrometry devices.20 ACKNOWLEDGMENT This research was sponsored by the US DOE, Office of Research and Development. Oak Ridge National Laboratory is managed by UT-Battelle, LLC for the U.S. Department of Energy under Contract DE-AC05-00OR22725. Received for review May 15, 2000. Accepted September 19, 2000. AC0005464 (20) Kornienko, O.; Reilly, P. T. A.; Whitten, W. B.; Ramsey, J. M. Anal. Chem. 2000, 72, 559-562.
Analytical Chemistry, Vol. 72, No. 23, December 1, 2000
5791
