Miniaturized Low-Cost Ion Mobility Spectrometer ... - ACS Publications

Ion mobility spectrometry (IMS) is a well-known method for detecting hazardous compounds in air. Typical applications are the detection of chemical wa...
6 downloads 0 Views 159KB Size
Download PDF
Anal. Chem. 2008, 80, 6671–6676

Miniaturized Low-Cost Ion Mobility Spectrometer for Fast Detection of Chemical Warfare Agents Stefan Zimmermann,*,† Sebastian Barth,† Wolfgang K. M. Baether,† and Joachim Ringer‡ Research Unit, Draegerwerk AG & Co. KGaA, D-23542 Luebeck, Germany, and Armed Forces Scientific Institute for Protection Technologies-NBC-Protection, D-29623 Munster, Germany Ion mobility spectrometry (IMS) is a well-known method for detecting hazardous compounds in air. Typical applications are the detection of chemical warfare agents, highly toxic industrial compounds, explosives, and drugs of abuse. Detection limits in the low part per billion range, fast response times, and simple instrumentation make this technique more and more popular. In particular, there is an increasing demand for miniaturized low-cost IMS for hand-held devices and air monitoring of public areas by sensor networks. In this paper, we present a miniaturized aspiration condenser type ion mobility spectrometer for fast detection of chemical warfare agents. The device is easy to manufacture and allows single substance identification down to low part per billion-level concentrations within seconds. The improved separation power results from ion focusing by means of geometric constraints and fluid dynamics. A simple pattern recognition algorithm is used for the identification of trained substances in air. The device was tested at the German Armed Forces Scientific Institute for Protection TechnologiesNBC-Protection. Different chemical warfare agents, such as sarin, tabun, soman, US-VX, sulfur mustard, nitrogen mustard, and lewisite were tested. The results are presented here. Ion mobility spectrometry (IMS) emerged as an analytical technique in the early 1970s. Detection limits in the part per billionrange and fast response times of just a few seconds make this technique more and more popular. Typical applications are the detection of chemical warfare agents, toxic industrial compounds, explosives, and drugs of abuse. IMS is well-established today for real-time monitoring of ambient air. It compares favorably to other analytical methods with respect to size, weight, handling, and instrumentation. There is an increasing demand for miniaturized, low-cost IMS for hand-held personal air monitoring devices and the detection of highly toxic compounds and explosives in public areas by sensor networks. In this paper, we present a miniaturized aspiration condenser type IMS that is easy to manufacture at low cost. Recent measurements at the Armed Forces Scientific Institute for Protection Technologies-NBC-Protection, Germany, show that our device has sufficient separation power for reliable * To whom correspondence should be addressed. Phone: +49-451-882-4546. Fax: +49-451-882-3921. E-mail: [email protected]. † Draegerwerk AG & Co. KGaA. ‡ Armed Forces Scientific Institute for Protection Technologies-NBC-Protection. 10.1021/ac800559h CCC: $40.75  2008 American Chemical Society Published on Web 07/30/2008

and reproducible identification of single substances in air down to low part per billion-level concentrations. It is important to note that the response time depends on the concentration. However, identification was possible within a few seconds even for the lowest concentrations tested. The device was challenged with sarin, tabun, soman, US-VX, sulfur mustard, nitrogen mustard, and lewisite. Ion Mobility Spectrometer Principles and Miniaturization Issues. Ion mobility spectrometry is based on the characteristic drift velocity of gas phase ion clusters in an electric field at atmospheric pressure. The ion formation process is described in detail by Eiceman et al.1 For low electric fields, the drift velocity vd is proportional to the electric field strength E as in eq 1. Low electric field conditions in analytical IMS hold as long as eq 2 is valid, where N is the density of the drift-gas molecules. νd ) KE

(1)

E e 10-17 V cm2 2N

(2)

The ion mobility K of a particular ion species depends on various parameters, such as pressure, temperature, ion mass, and ion collision cross section. A theoretical model for K is described in the literature.1,2 Under constant conditions, the ion mobility is a characteristic parameter for a particular ion species. In classical time-of-flight IMS, a shutter grid is recurrently opened for a short time allowing an ion swarm to enter the drift tube. A homogeneous electric drift field forces the ions toward a Faraday plate acting as the ion detector. The drift time td needed for a particular ion species to overcome the distance d between the ion shutter grid and the detector can be calculated as in eq 3.

td )

d d ) νd KE

(3)

Thus, different ion mobilities K lead to different drift times td. The initial ion swarm ideally separates in a number of individual ion swarms containing just one ion species. An ion mobility spectrum results by plotting the ion current over the drift time. A detailed description of time-of-flight IMS is given elsewhere.1 Most time-of-flight IMS have high resolution of R > 25. The definition (1) Eiceman, G. A.; Karpas, Z. Ion Mobility Spectrometry, 2nd ed.; CRC Press, Taylor & Francis Group: Boca Raton, FL, 2005. (2) Revercomb, H. E.; Mason, E. A. Anal. Chem. 1975, 47 (7), 970–983.

Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

6671

of the resolution R is given in eq 4, where td is the drift time and Wt,1/2 the temporal peak width measured at half of the peak height.3 In time-of-flight IMS the resolution R is independent of the peak position and drift time, respectively, since peak width proportionally increases with drift time.

R)

td Wt,1/2

(4)

The major drawback of time-of-flight IMS is the complex design compared to other IMS principles. Precise manufacturing is required, which makes miniaturization difficult and fabrication costs higher. Furthermore, peak broadening caused by inhomogeneous drift fields, space charge effects, and diffusion becomes more significant at small dimensions, which limits the resolution of miniaturized time-of-flight IMS.4–6 In particular, short drift lengths require very fast ion shutters and amplifiers for acceptable separation capabilities. The typical resolution of miniaturized timeof-flight IMS is R ≈ 10-15.1,6,7 A different miniaturization approach is based on field asymmetric ion mobility spectrometry (FAIMS), also referred to as differential mobility spectrometry (DMS), where in contrast to eq 2 a strong high-frequency electric field is used for ion separation. The separation principle is based on the effect of the electric field strength on the ion mobility. A detailed description of FAIMS is given elsewhere.8 The typical resolution of FAIMS is R ≈ 5-20.9 Miller et al. developed a planar FAIMS that can be easily manufactured by standard MEMS technologies.10 The device has good separation power but suffers from high power consumption due to the required strong high-frequency electric field. This makes miniaturized, lightweight, hand-held FAIMS difficult to develop. Another drawback is the reduced separation power for large molecules with high molecular weight, such as chemical warfare agents.11 Furthermore, traditional dopant gases, typically used in real environment applications to reduce the number of interferences, perform poorly in FAIMS.12 The objective of our research is the development of a miniaturized, low-cost, and easy to manufacture IMS with low power consumption and sufficient separation capabilities for single substance identification in air. Our concept is based on the aspiration condenser type IMS developed by Puumalainen and Paakkanen et al., where ions travel within a gas stream through (3) Siems, W. F.; Wu, C.; Tarver, E. E.; Hill, H., Jr. Anal. Chem. 1994, 66 (23), 4195–4201. (4) Spangler, G. E.; Vora, K. N.; Carrico, J. P. J. Phys. E: Sci. Instrum. 1986, 19 (3), 191–198. (5) Baumbach, J. I.; Berger, D.; Leonhardt, J. W.; Klockow, D. Int. J. Environ. Anal. Chem. 1993, 52 (1-4), 189–193. (6) Xu, J.; Whitten, W. B.; Ramsey, J. M Anal. Chem. 2000, 72 (23), 5787– 5791. (7) Teepe, M.; Baumbach, J. I.; Neyer, A.; Schmidt, H.; Pilzecker, P. Int. J. Ion Mobility Spectrom. 2001, 4 (1), 60–64. (8) Buryakov, I. A.; Krylov, E. V.; Nazarov, E. G.; Rasulev, U. K. Int. J. Mass Spectrom. Ion Processes 1993, 128 (3), 143–148. (9) Eiceman, G. A.; Nazarov, E. G.; Miller, R. A Int. J. Ion Mobility Spectrom. 2000, 3 (1), 15–27. (10) Miller, R. A.; Eiceman, G. A.; Nazarov, E. G.; King, A. T. Sens. Actuators, B: Chem. 2000, 67 (3), 300–306. (11) Miller, R. A.; Nazarov, E. G.; Krylov, E.; Eiceman, G. A. Method and Apparatus for Control of Mobility-Based Ion Species Identification. U.S. Patent 7,005,632. (12) Ross, S. K.; McDonald, G.; Marchant, S. Analyst 2008, 133, 602–607.

6672

Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

a transverse electric field, which forces the ions toward a number of detector electrodes.13,14 Different ion species with different ion mobilities reach different detector electrodes. Pattern recognition is used for ion identification. One major drawback of this approach is the low resolving power due to poor spatial ion separation caused by space charge effects and diffusion. Both effects can be reduced by increasing the flow rate so that ion concentration and drift time decrease. However, the flow rate is limited since laminar flow conditions are required. Furthermore, insufficient spatial focusing of ions before entering the separation region significantly reduces the resolving power of the aspiration condenser IMS developed by Puumalainen and Paakkanen et al. An improved aspiration condenser type IMS was developed by Sacristan et al.15,16 This swept-field aspiration condenser IMS uses a variable electric drift field to move all ion species across a single detector electrode. The individual detector electrodes are replaced by a single detector electrode in combination with a variable deflection voltage. The resulting I(V) curve is transformed into an ion mobility spectrum by applying the discrete inverse Tammet transform.17 However, reconstruction of the ion mobility spectrum is difficult in practice, especially for low signal-to-noise ratios. The required regularization of the discrete direct Tammet transform limits the resolution of the swept-field aspiration condenser IMS, and again, insufficient focusing of ions before entering the separation region further reduces resolving power. Our approach to gain resolving power is effective ion focusing before separation. As seen below, resolving power is significantly improved, while the system still meets all requirements for a lowcost, hand-held device as mentioned above. Miniaturized Ion-Focusing Aspiration Condenser IMS. The concept of our ion-focusing aspiration condenser IMS is shown in Figure 1. It is based on two different gas flows, the sample gas and the drift gas, that form parallel gas streams under laminar flow conditions. Because of the small channel dimensions, laminar flow conditions are given at the flow rates used. The ions travel within the sample gas stream, which is focused by means of geometric constraints before entering the separation region. In particular, a rectangular aperture of 100 µm × 5 mm defines the sample gas stream at the point of entry. Furthermore, a drift gas flow rate significantly higher than the sample gas flow rate prevents sample gas expansion in the separation region. Actually, the parallel drift gas stream fills most of the separation region, so that a thin layer of sample gas forms at the counter electrode. However, the mean sample gas velocity is about 2 times higher than the mean drift gas velocity since a certain sample gas flow rate is required in the ionization region. Further increasing the drift gas flow rate would lead to vortices. Still, geometric constraints in combination with fluid dynamics result in effective focusing of the sample gas stream and carried ions, respectively. (13) Puumalainen, P. Method for Detection of Foreign Matter Contents in Gases. U.S. Patent 5,047,723. (14) Paakkanen, H.; Ka¨rpa¨noja, E.; Ka¨tto ¨, T.; Karhapa¨a¨, T.; Oinonen, A.; Salmi, H. Method and Equipment for Definition of Foreign Matter Contents in Gases. Pat. Appl. WO 94/16320. (15) Sacristan, E. Ion Mobility Method and Device for Gas Analysis. U.S. Patent 5,455,417. (16) Sacristan, E.; Solis, A. A. IEEE Trans. Instrum. Meas. 1998, 47, 3, 769– 775. (17) Tammet, H. The Aspiration Method for the Determination of AtmosphericIon Spectra, Scientific Notes of Tartu State University No. 195, 1967; Israel Program for Scientific Translations: Jerusalem, Israel, 1970.

Figure 2. Dimensions of the planar separation region of the miniaturized ion-focusing aspiration condenser IMS. Figure 1. Concept of the miniaturized ion-focusing aspiration condenser IMS. The ions are carried in the sample gas stream, which is focused by means of geometric constraints and fluid dynamics before entering the separation region. A transverse electric field forces the ions out of the sample gas stream while traveling through the separation region, through the drift gas stream, and toward the bottom electrode setup. Ion species with different ion mobilities eventually separate into individual ion beams. The I(V) curve can be recorded by varying the drift voltage, so that all ion beams are successively directed onto the detector electrode. The ion mobility spectrum results from differentiating the recorded I(V) curve.

A transverse electric field forces the ions out of the sample gas stream while passing through the separation region, through the drift gas stream and toward the bottom electrode setup. Ion species with different ion mobilities ideally separate into individual ion beams. An ion mobility spectrum can be derived from varying the drift voltage, so that all ion beams are successively directed onto the detector electrode. The ion mobility spectrum results from differentiating the recorded I(V) curve with respect to V. Because of a high signal-to-noise ratio, differentiation is feasible. Separation power is improved compared to previously mentioned aspiration condenser type IMS since all ions travel through the entire drift field, which results in a more effective ion separation. Only ions close to the counter electrode are fully separated without focusing. Still, space charge effects and diffusion cause limited spatial ion separation. As mentioned above, both effects can be reduced by increasing the flow rates in order to minimize ion concentration and drift time. However, the flow rates are limited since laminar flow conditions are required. A more detailed description and a theoretical model of our ion-focusing aspiration condenser IMS are published elsewhere.18–20 EXPERIMENTAL SECTION A semi-integrated prototype of the miniaturized ion-focusing aspiration condenser IMS was used for all measurements. Pumps, filters, and most of the electronics, such as the pA-amplifier, control circuits, drivers, pressure, temperature, and mass flow sensors, were already integrated. A laptop was used for data acquisition. The total volume of the experimental setup was about 800 cm3 excluding the laptop and an external 15 V dc power supply. The dimensions of the planar separation region are given in Figure 2. The IMS device was made from PEEK using standard manufacturing techniques. The integrated counter, detector, and (18) Zimmermann, S.; Abel, N.; Baether, W.; Barth, S. Sens. Actuator, B: Chem. 2007, 125, 428–434. (19) Zimmermann, S.; Barth, S. Proc. Transducers 2007, 1501–1504. (20) Barth, S.; Baether, W.; Zimmermann, S. Proc. IEEE Sens. 2007, 79–82.

Figure 3. Pneumatic configuration of the miniaturized ion-focusing aspiration condenser IMS.

auxiliary gold electrodes were MEMS fabricated at IMTEK, Germany. Since the IMS component is not fully integrated into a hand-held, product-level device with power and size optimized electronics and gas handling, the mentioned volume just gives a first impression. A total volume of 600 cm3 including batteries and filters for 12 h of continuous operation seems to be realistic for the final product. Sarin, tabun, soman, US-VX, sulfur mustard, nitrogen mustard, and lewisite were measured. From earlier experiments, it is known that lower concentrations of sulfur mustard, nitrogen mustard, and lewisite are not detectable under humid conditions. Thus, purified dry air with an absolute humidity of 20 ppm was used for all experiments to investigate the separation power of the concept independent of any inlet system. β emission from a radioactive Ni-63 source with an activity of 100 MBq and an active area of 30 mm2 was used for ionization. All measurements were carried out at room temperature and with 300 mL/min of dry sample gas containing the analytes and 1000 mL/min of dry purified drift gas. The I(V) curves were recorded by varying the drift voltage from -24 to +24 V in 300 discrete steps. The ion current was measured for 100 ms per voltage step and averaged afterward to improve the signal-to-noise ratio. Thus, the total recording time was 30 s per I(V) curve. Both positive and negative ion mobility spectra were derived from differentiating averaged ¯I(V) curves with respect to V. Negative deflection voltages represent the negative ion mobility spectrum, and positive deflection voltages represent the positive ion mobility spectrum, respectively. Ions with high mobilities appear at low deflection voltages (absolute values). The pneumatic configuration of the miniaturized ion-focusing aspiration condenser IMS is shown in Figure 3. A molecular sieve of type 13X was used to remove contaminants and analytes from the drift gas. Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

6673

Figure 4. jI(V) curves of sarin at different concentrations. The saturation current increases with an increasing sarin concentration due to lower recombination rates of the sarin product ions compared to the reactant ions.

Figure 5. Differentiated jI(V) curves (ion mobility spectra) of sarin at different concentrations. At a sarin concentration of 120 µg/m3, just a single peak is visible at Ud ) 13.3 V.

RESULTS AND DISCUSSION As an example, Figure 4 and 5 show the recorded and differentiated ¯I(V) curves of sarin at different concentrations. As seen from Figure 4, the ion current saturates at higher deflection voltages since all ion species hit the detector electrode. Furthermore, the saturation current increases with an increasing sarin concentration. This can be explained by lower recombination rates of the sarin product ions compared to the reactant ions. The positive and negative reactant ion peaks (RIP) are located at Ud ) 8.3 V and Ud ) -8.5 V respectively, see Figure 5. The resolution R can be calculated as in eq 5 to R ) 2.4 for the positive reactant ion peak. R)

Ud WU,1/2

(5)

It is important to note that the resolution R of our ion-focusing aspiration condenser IMS is not independent of the peak position compared to time-of-flight IMS.18 Because of identical ion trajectories at their point of detection, time-of-flight is identical for all ion species. Since peak broadening mainly depends on diffusion and the diffusion coefficient is proportional to the ion mobility, spatial peak broadening is less for ion species with lower mobilities. This leads to an increased resolution at higher deflection voltages. 6674

Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

Figure 6. Differentiated jI(V) curves (ion mobility spectra) of nitrogen mustard at different concentrations. At higher nitrogen mustard concentrations a single positive product ion peak is visible at Ud ) 11.4 V.

As seen from Figure 5, only positive sarin product ions are formed. Because of the limited number of reactant ions in our system and the fact that reactant ions deplete when forming product ions, the positive reactant ion peak significantly decreases with an increasing sarin concentration. At higher sarin concentrations, all positive reactant ions turn into product ions. Furthermore, at a Sarin concentration of 120 µg/m3, just a single peak is visible at Ud ) 13.3 V. This peak is most likely the proton-bound dimer of sarin. However, further investigations by mass spectrometry are required for confirmation. A larger number of reactant ions would increase the dynamic range and linearity of the system. A stronger Ni-63 source would help, but the number of reactant ions does not proportionally increase with the activity. Thus, a significantly stronger Ni-63 source would be required. As another example, Figure 6 shows the ion mobility spectra of nitrogen mustard at different concentrations. Both, positive and negative nitrogen mustard product ions are formed. Again, because of the limited number of reactant ions in our system and the fact that reactant ions deplete when forming product ions, both reactant ion peaks significantly decrease with an increasing nitrogen mustard concentration. At higher nitrogen mustard concentrations, a single positive product ion peak is visible at Ud ) 11.4 V. This peak is most likely the proton-bound monomer of nitrogen mustard. Again, further investigations by mass spectrometry are required for confirmation. No discrete peaks are visible in Figures 5 and 6 since the resolution of our system is still low compared to time-of-flight IMS. This applies to all substances. Thus, substances or product ion species are hard to identify by their peak positions. However, pattern recognition can be used for single substance identification. Therefore, each drift voltage step Vn represents a component, each I(Vn)S,c the corresponding value, and each I(V)S,c curve the characteristic pattern for a certain substance S at a concentration c. Representative patterns result from averaging a number of measured I(V)S,c curves and I(Vn)S,c values, respectively, as in eq 6. Here we used v ) 10 measurements per substance S and concentration c. Furthermore, just a limited number m of concentrations cm could be trained, so that linear interpolation was used for pattern generation at concentrations not trained, see eq 7. Figures 7 and 8 show the ion mobility spectra of sarin and nitrogen mustard at different concentrations. The ion mobility

Figure 7. Differentiated and interpolated jI(V) curves (ion mobility spectra) of sarin at various concentrations. The averaged and interpolated spectra were obtained as in eqs 6 and 7.

Figure 8. Differentiated and interpolated jI(V) curves (ion mobility spectra) of nitrogen mustard at various concentrations. The averaged and interpolated spectra were obtained as in eqs 6 and 7.

spectra were derived from differentiating the recorded and interpolated ¯I(V)S,c curves. The models are based on m ) 5-6 measured concentrations per substance.

Figure 9. Sammon map of all trained substances.

S, see eq 8. In the following, zero substance concentration is indexed with RIP since only reactant ion peaks (RIP) are present in the spectrum. To account for the noise of our system, the mean value ¯I(V)RIP of k measured patterns ¯I(V)c-0,k at zero concentration is calculated as in eq 9. Furthermore, the Euclidean distance dRIP,k of this mean value to all measured patterns at zero concentration is calculated as in eq 10. Finally, a threshold value did is defined as in eq 11, so that 99% of all measured patterns at zero concentration have a Euclidean distance to their mean value smaller than the threshold value. Considering the noise independent of the substance concentration, reliable and reproducible substance identification with 0.99 confidence is possible as long as eq 12 is valid for a single substance S, so that X ) S and cx ) cp. If eq 12 is valid for multiple substances, reliable identification is not possible but a substance specific uncertainty could be calculated. However, this is not part of the following discussion. 2 dS,min ) minc

[∑ (

)

]

)2

I(Vn)X,cx - I(Vn)S,c

n

∑ (I(V )

n X,cx - I(Vn)S,cp

)2

(8)

n

I(Vn)S,c )

I(Vn)S,c )

1 ν

I(Vn)S,cm - I(Vn)S,cm-1 cm - cm-1

∑ I(V )

n S,c,ν

(6)

ν

(c - cm-1) + I(Vn)S,cm-1, for cm-1 < c e cm (7)

Figure 9shows a two-dimensional Sammon map of all trained and interpolated patterns. Sammon mapping is a nonlinear method that maps a high-dimensional vector space onto two dimensions while preserving the distance of data points as good as possible.21 It gives a first impression of the pattern structure and the separation power of our miniaturized ion-focusing aspiration condenser IMS. As seen from Figure 9, the patterns of tabun, sarin, and soman are very similar whereas the patterns of US-VX, sulfur mustard, nitrogen mustard, and lewisite are quite different. For better evaluation of the separation power, a simple approach is suggested, where substance identification and quantification are based on the minimum Euclidean distance dS,min of a measured pattern ¯I(V)S,cx to the patterns ¯I(V)S,c of substance (21) Sammon, J. W. IEEE Trans. Comput. 1969, 18 (5), 401–409.

I(Vn)RIP )

2 ) dRIP,k

1 k

∑ I(V )

(9)

n c)0,k

k

∑ (I(V )

)2

n c)0,k - I(Vn)RIP

(10)

n

did : P(dRIP < did) ) 0.99

(11)

dS,min e did

(12)

Table 1shows the identification limits of our miniaturized aspiration condenser IMS for all trained substances. The identification limit defines the minimum concentration, where a substance can be clearly identified with 99% probability according to the above approach. In other words, any trained substance at a concentration higher than the identification limit cannot be confused with any other trained substance. In this case eq 12 does apply to only one of the trained substances. The apparent contradiction between Figure 9 and Table 1 results from individually parametrized curves in the Somman map. Each curve spans a substance specific concentration range that depends on the substance specific sensitivity. Because of significantly Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

6675

Table 1. Identification Limits cid and Lowest Concentrations Trained cmin substance

cid (µg/m3)

cmin (µg/m3)

tabun sarin soman US-VX sulfur mustard nitrogen mustard lewisite

2 6 7 3 34 24 93

3 15 30 5 70 150 63

higher sensitivities for tabun, sarin, and soman compared to sulfur mustard, nitrogen mustard, and lewisite, significantly lower concentrations are required to move along the curves and to generate distinct patterns of tabun, sarin, and soman. Furthermore, the high signal-to-noise ratio allows substance identification close to the origin of the Somman map. It is important to note that the presented identification limits are theoretical values based on the signal-to-noise ratio and the linear approach for pattern interpolation at concentrations not trained. However, this model gives a very good estimation as long as the minimum concentration trained is close to the calculated identification limit, see Table 1. It is also important to note that the detection limits are significantly lower than the identification limits. The identification limit is defined as the lowest concentration where reliable and reproducible substance identification is possible. Figure 10 shows the response of our IMS when challenged with lewisite at concentrations that were not trained beforehand to demonstrate reliable and reproducible substance identification based on interpolated patterns. All of the recorded I(V) curves were clearly identified as lewisite when lewisite was present. As mentioned above, a pattern can be recorded in 30 s by stepping the deflection voltage in 300 steps and integrating the ion current for 100 ms per step. It can be shown that 30 deflection voltages Vn are sufficient for clear pattern recognition so that substance identification is possible in 3 s. A major drawback of the above pattern recognition approach is the limited performance when target substances are present in a mixture. Basically, any substance mixture can be identified if trained. However, this is impossible in practice due to the immense training effort of testing all possible mixtures and concentration variations. Furthermore, interferences are not known beforehand in real environment

applications. If an individually trained target substance can still be identified in a mixture strongly depends on the composition of the mixture. Substance identification is always difficult when interferences are present that form a significant number of product ions. This would lead to untrained patterns and uncertain substance identification. Because of the high proton affinity of tabun, sarin, and soman, ion formation is suppressed for substances with low proton affinity. Very high concentrations would be required to form a critical number of interfering ions. Furthermore, dopant gases could be used to effectively prevent ionization of interferences by modifying the gas-phase ion chemistry in the ionization region. This significantly reduces the number of interfering compounds. Dopant gases in IMS are discussed elsewhere.1,22,23 Thus, substance identification in mixtures is possible as long as present interferences are effectively suppressed by a dopant gas or the target substance itself. Substance identification by peak position generally performs better in real environments since ionized interferences are just an issue if their ion mobility is similar to a target substance. However, substance identification by peak position requires higher resolution of R > 10 and mixtures are still hard to separate since the resolution of IMS is generally low compared to MS with R > 1000, as required for analyzing complex mixtures. CONCLUSIONS The simple concept of our miniaturized ion-focusing aspiration condenser IMS allows easy manufacturing at low cost. All major components of the IMS unit are made from PEEK by standard manufacturing techniques. The integrated electrodes are MEMS fabricated. Compared to other aspiration condenser type IMS, the separation power of our system is improved due to effective ion focusing before ion separation. The ions are focused by means of geometric constraints and fluid dynamics. However, the resolution is still low and substance identification by peak position not possible yet. The resolution is R ) 2.4 for the positive reactant ion peak. Pattern recognition is used for the identification of trained substances. The device was tested at the Armed Forces Scientific Institute for Protection Technologies-NBC-Protection, Germany, to evaluate the analytical performance. The spectra of different chemical warfare agents, such as sarin, tabun, soman, US-VX, sulfur mustard, nitrogen mustard, and lewisite were recorded at various concentrations. From earlier tests it is known that lower concentrations of sulfur mustard, nitrogen mustard, and lewisite are not detectable under humid conditions. Thus, purified dry air with an absolute humidity of 20 ppm was used for all experiments to evaluate the separation power of the IMS concept independent of any inlet system. On the basis of these measurements, it is demonstrated that the separation power of our miniaturized ion-focusing aspiration condenser IMS is sufficient to clearly identify all trained substances down to a few micrograms per cubic meter. However, the used identification algorithm does not allow substance identification in the presence of ionized interferents yet. As mentioned above, a dopant gas could be used to chemically mask interferents. Received for review March 18, 2008. Accepted June 25, 2008.

Figure 10. Response of the miniaturized IMS to lewisite at concentrations not trained. The given concentrations were 220 and 150 µg/m3. The recorded I(V) curves were clearly identified as lewisite when lewisite was present. 6676

Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

AC800559H (22) Proctor, C. J.; Todd, J. F. J Anal. Chem. 1984, 56, 1794–1797. (23) Eiceman, G. A.; Wang, Y. F.; Garcia-Gonzales, L.; Harden, C. S.; Shoff, D. B Anal. Chim. Acta 1995, 306, 21–33.