Plasma Chromatography - Analytical Chemistry (ACS Publications)

---

Figure 1. PC instrumentation system composed of commonly used elec­ tronic components applied to ion-drift tube technology

Plasma Chromatography Francis W. Karasek Department of Chemistry University of Waterloo Waterloo, Ont., Canada

Plasma chromatography (PC) shows much promise as a method for ultratrace analysis of organic com­ pounds. The technique is based on creating an ion-molecule reaction and observing mobility spectra which re­ veal both the kind and relative abun­ dances of the charged particles formed, with all steps being carried out at atmospheric pressure. The name was suggested by the use of ions (plasma) and a relation of ion mobility separation to chromatogra­ phy. Ions for the reaction are gener­ ated in a purified nitrogen or air car­ rier gas containing a trace of water vapor by the action of 60-keV elec­ trons emitted from a 6 3 Ni foil. In nitrogen the primary ions formed undergo a series of reaction steps to evolve the stable species of (H 2 0)„H+ and (H 2 0)„NO+ ions, whose relative abundance and value of η depend upon water concentration and temperature. The negative parti­ cles are low-energy (~0.5 eV) elec­ trons. When air is used as a carrier gas, (Η2θ) η θ2~ ions are added to these groups. A trace organic mole­ cule injected into the carrier gas will undergo reactions with these ions and electrons, forming reaction product ions which show both quasi-molecular and simple dissociated ions in their mobility spectra. The method is ex­ tremely sensitive and reasonably

qualitative. It is capable of detecting 10~ 12 gram or less of a compound and can provide identification through the characteristic positive and negative mobility spectra. The PC method is a combination of well-established techniques. The in­ strumentation is also a combination of well-known instrumental compo­ nents. The PC tube consists of two coupled ion-drift tubes: one contain­ ing the 63 Ni foil functions as a reactor to generate the ion-molecule com­ plexes; the other equipped with an ion-injection grid and electrometer detector functions as an ion-drift spectrometer to produce the mobility spectra. Illustrated in Figure 1 are the simple components and electron­ ics needed to produce a working sys­ tem. The 63 Ni is the same foil ele­ ment used in an electron-capture gas chromatographic detector (1). The ion-drift tubes and ion-injection grid are similar to designs long employed by physicists (2). The electronic com­ ponents (electrometer, dc voltage supply, boxcar integrator, digital sig­ nal averager, and recorder) are all commonly available components whose functions can be duplicated by a number of different available de­ vices. A mobility spectrum is produced by using a voltage gradient to move the charged particles generated in the

710 A • ANALYTICAL CHEMISTRY, VOL. 46, NO. 8, JULY 1974

reactor tube toward the drift spec­ trometer tube and by injecting a pulse of these ions through an injec­ tion grid into the spectrometer. As this group of ions moves through the spectrometer in an inert N 2 atmo­ sphere toward the electrometer detec­ tor, separation of individual ionic types occurs because of their differing mobilities in N 2 . This action can be thought of as similar to that of a time-of-flight mass spectrometer, ex­ cept that the ionic velocity at the at­ mospheric pressure PC conditions is determined by the collision interac­ tion between an ion and neutral ni­ trogen molecules, rather than by the kinetic energy velocity of the ion as in mass spectroscopic conditions. The results are similar in either case; the heavier ions move more slowly. Posi­ tive or negative ions can be indepen­ dently observed merely by choice of the electrical field polarity. A single mobility spectrum (Figure 2) for ions with masses up to 400 oc­ curs in 20 msec. A convenient way to record these data is with a digital sig­ nal averager whose output will be the integrated sum of several hundred 20-msec spectra. An alternate, much simpler way to record spectra in the several minute time span needed to use the usual laboratory Χ- Υ recorder can be accomplished with a boxcar integrator (Model CW-1 Princeton

Report

Figure 2. Mobility spectrum

Reactant charged particles created by 63 Ni ionizer containing (H 2 0)„H + and ( H 2 0 ) „ N O + ions and low-energy electrons. Spectra, positive and negative, produced by N 2 with ~ 1 0 ppm H 2 0 at 200°C and 760 torr. Relative abundance of positive ions changes with temperature and water concentration; more water and lower temperature favor formation of larger water cluster ions. Reduced mobility values, K0, calculated by Equation 6

Applied Research, P.O. Box 565, Princeton, N . J . 08540). Boxcar integrators are instruments that average a repetitive waveform by examining it one piece at a time. Triggered by the sample injection pulse, a delay applied to a gating pulse is slowly scanned across the 0-20 msec mobility spectrum time domain. The gate pulse only permits ions from the portion of the spectrum corresponding to the delay time to be recorded. Thus, if the gating pulse is scanned across the 0-20 msec mobility spectrum in 1 min, the recorded mobility spectrum will be an integrated average of 3000 20-msec spectra. The scanning gate pulse can either be applied after the electrometer (Figure 1) or to another grid immediately in front of the detector element in the PC tube. In the latter case, the response speed requirement of the electrometer is less stringent. A useful advantage of the boxcar integrator electronics is also its ability to tune the system to a given ionic peak and observe peak behavior as a function of sample concentration and time. For the past several years there has been limited commercial availability of a PC instrumentation system. Only slightly over 10 custom-built units, each somewhat different, have been produced to date. Several of the early models incorporated a quadrupole mass spectrometer directly interfaced

to the drift spectrometer to give the mass of ionic peaks in the mobility spectrum (3). This system proved to be too complicated and expensive to warrant the slight additional data provided and was soon abandoned. The present model (BETA V l l - S , Franklin GNO Corp., West Palm Beach, Fla. 33402) uses the PC tube, shown in Figure 3 designed into a standard Blue M laboratory oven, and incorporates commercially available components as illustrated in Figure 1 to provide a total system costing between $30,000 to $40,000. Except for slight changes in construction of the PC tube, the basic instrumentation has received little development since its inception in 1969. Most of the activity in PC has centered around exploratory studies to demonstrate and understand the scope of the technique. Scope of Plasma Chromatography The literature contains about 30 publications on PC. Several of these have been state-of-the-art reviews (4-6), but most have reported exploratory studies. Of these, 25 have come from our Waterloo laboratories where exploratory analytical work has been conducted on understanding the phenomena involved and the capabilities of the technique. From our work two general observations have evolved:

the data contained in positive mobility spectra bear a close similarity to those obtained in chemical ionization mass spectrometry, and the data contained in negative mobility spectra correspond to those obtained with the GC electron capture detector (ECD), if one were able to observe the negative ions formed rather than the decrease in standing current. These two broad generalizations provide considerable guidance in understanding both the technique and its applicability to analytical problems. Chemical ionization mass spectrometry (CIMS) is a well-established technique (7). With its ability to produce mass spectra more abundant in molecular-type and simpler fragmentation ions than those from electron impact ionization, the technique is widely used in G C / M S systems as a valuable interpretative aid for mass spectra. With a CH4 reactant gas, CIMS ionization of sample molecules occurs via an ion-molecule reaction at 1 torr between CH 5 +, C 2 H 5 +, C 3 H 5 + ions and the organic sample molecules to give additive product ions of MH+, (M+C 2 H 5 )+, (M+C 3 H 5 )+ types for classes of compounds such as aromatics, and to give fragment ions of ( M - H ) + and (M-CxH y )+ types for alkanes. In like manner, PC ionization of sample molecules occurs via an ion-molecule reaction at 760 torr between (H 2 0)„H+, (H 2 0)„NO+

A N A L Y T I C A L CHEMISTRY, VOL. 46, NO. 8, JULY 1974 • 711 A

Figure 3. C u t - a w a y v i e w of PC tube Two glass inlet tubes lead into 6 3 Ni ionizer, seen as bright foil on first cir­ cular element. Each circular element at different electrical potential to maintain uniform field gradient across tube. Ion-drift spectrometer uses larger circular elements than ion-molecule reactor section; ion-injection grid separates two ion-drift tubes. Electrometer detector only partially observable behind drift gas entrance hole in end plate. Total length of drift tubes, 14 cm with 200 V / c m gradient

ions and the organic sample mole­ cules to give similar results. Interpretation of the data can be treated in comparable ways. For ex­ ample, in CIMS with CH4 reactant gas, one locates the quasi-molecular ion of many compounds by looking fortheMH+, (M+29)+, (M+4D+ series; in PC one looks for the MH+, (M)NO+ pair (Figure 4). PC studies illustrating a relationship of PC to CIMS spectra have been reported for alkanes (8), alkyl halides (-9), halogenated benzenes (10), alkyl alcohols (11), esters (12), halogenated nitroaromatics (13), and nitroaromatics (14). Although there are similarities between CIMS and PC, there are im­ portant differences, primarily attrib­ utable to the differences in pressure conditions and reactant, ion species. One aspect of this difference is the ease of formation of M2H+ and M 3 H+ cluster ions in PC, particularly with polar compounds such as alco­ hols and esters. In gas chromatographic analysis the ECD is a useful and well-established detector because of its great sensitivi­ ty and selectivity to many classes of analytically important compounds. Studies of ECD behavior have led to a number of theories of its mecha­ nism and to designs that alleviate some anomalous behaviors observed. Pellizzari has given an extensive.re­ view of the characteristics and perfor­ mance of the ECD (15). Since in the negative mode the PC instrument can

Figure 4. PC data for n-alkanes Comparison one can make to similar data obtained for compounds in CI MS by appearance of quasi-molecular ions of (M —H) + , additive ions of ( M + N O J + , and series of fragment ions whose composition can be in­ ferred by coincident mobilities among compounds in series

be thought of as an ECD in which the negative mobility spectra reveal the ionic products formed as a result of electron capture by a compound, use­ ful information can be developed to test previous mechanisms and ad­ vance our understanding of electron attachment phenomena. Negative PC mobility data have demonstrated the occurrence of both dissociative and associative electron attachment (10, 16). Both the monohalogenated benzenes and alkyl halides exhibit simple dissociative elec­ tron capture to give only the halide ion. Nitrobenzene exhibits only asso­ ciative electron capture to give the negative molecular ion (9), whereas the halogenated nitrobenzenes and nitrated toluenes show both associa­ tive and dissociative electron capture (13, 17). Traces of air in the carrier gas to the ECD have a deleterious ef­ fect on sensitivity and linearity. PC data show the effect is due to a tem­ perature-dependent depletion of reac­ tive electrons by formation of (H 2 0) r e 0 2 "' complexes (18). Not only

712 A • ANALYTICAL CHEMISTRY, VOL. 46, NO. 8, JULY 1974

are there fewer electrons now avail­ able to provide a response for an elec­ tron-capturing compound entering the detector, but this compound will undergo a competing ion-molecule reaction with the (Η2θ),ιθ2~ ions present. This is also true for electron capturing column bleed compounds. Although negative PC mobility spectra are useful in understanding the behavior of an ECD, these spectra are far more useful in a fundamental analytical sense. Only those com­ pounds that undergo electron attach­ ment produce negative mobility spec­ tra. The spectra of those that do pro­ vide detection with equal or greater sensitivity than given by the ECD and also provide a qualitative spec­ trum for identification. The mobility data shown in Figure 5 for a series of halogenated nitrobenzenes show the differing spectral patterns given for isomers as well as the presence of an M~ ion. In many compounds the co­ incidence of the M~ and MH+ ions in the negative and positive mobility spectra provides a means for interpre-

taking into account other interaction forces (22). After subsequent modifi­ cation by Hassé (23), this theory yielded two equations for mobility, depending upon whether ion-neutral molecule repulsive forces or pure polarization forces predominated upon an interactive collision. These equations are: K

0.75 e

° =

[\

M V*

L1 + m \

2 /

,

v

(2)

(elastic sphere limit) where e is the ionic charge, D12 is the sum of the ionic and molecular radii, ρ is the gas density, ρ is the gas pres­ sure, M is the molecular mass, and m is the ionic mass, and w

0.5105

|\

M l 1/2

(polarization limit)

Figure 5. Normalized plots of negative ion intensities vs. Ko values in mobility spectra for series of halo- and nitrosubstituted benzenes

tation of the spectra and determination of molecular weight. Taken together, the positive and negative mobility spectra of a compound show characteristic patterns that can identify that compound. Applicable Theory Needs Refinement The theory applicable to PC involves that of gas-phase ion-molecule reactions and ionic mobilities. Extensive studies have been made in both fields, but attempts to apply these theoretical results to PC reveal the need for extension and modification of the theory to cover the experimental conditions of PC: namely, ionmolecule reactions of polyatomic molecules and ionic mobilities of large ions at 760 torr. This situation can be illustrated by a brief review of current theory and its relationship to PC. A more detailed treatment can be found elsewhere (19, 20). To begin a discussion of ionic mobility, we can consider a localized collection of ions of a single type contained in a gas of uniform tempera-

ture and total pressure for which the number density of the ionic species is low enough that coulombic repulsion forces may be ignored. If a weak uniform electric field is applied throughout the gas, a steady flow of ions along the field lines will develop, superimposed on the much faster random diffusive motion. The velocity of the center of mass of the ion cloud, or equivalently the average velocity of the ions, is known as the drift velocity, Vd, which is directly proportional to the electric field intensity, E, provided the field is kept weak. Thus, Vd = ÈE

(1)

where the constant of proportionality, K, is called the (scalar) mobility of the ions; Κ is a joint property of the ions and the gas through which the motion occurs. The first theory of ionic mobility was presented in 1903 by Langevin and was based on the kinetic theory of gases (21). However, calculated values of Κ compared to experimen­ tal data were always too high by about a factor of 4. In 1905 Langevin published a more rigorous theory,

714 A • ANALYTICAL CHEMISTRY, VOL. 46, NO. 8, JULY 1974

where κ is the dielectric constant of the gas. Since the publication of these theories, several other investigators have altered the equations to account for various shortcomings in the pre­ diction of mobilities (24-30), the most rigorous theory being that of Mason and Schamp (30). The disadvantage of classical mo­ bility theory when applied to the phenomena observed in PC is that all experimental mobilities used to im­ prove the theory were measured on single ions in their own vapor at low pressures (about 0.1-10 torr), e.g., He+ in He, Kr+ in Kr, Xe+ in Xe, or single ions in an inert gas, e.g., Na + in He, Cs+ in He, K+ in Ar. In PC, polyatomic ions and ion molecules drifting in nitrogen gas are the species of interest, and little theory has been developed to explain the mobilities of these types of ions. Using the data of Patterson for ions observed in sulfur hexafluoride (3/), Mason and coworkers modified the earlier Mason-Schamp theory to ac­ count for the observed mobilities (32). They modified the calculations to allow for the bulkiness of the ion and the fact that its charge is not lo­ calized at its geometric centre. With their model they attempted to ex­ plain why large polyatomic ions show mobilities falling below the polariza­ tion limit mobility derived by Langevin (Equation 3). The equation for mobility of Mason and coworkers is:

~~ 16 Nim Γ2πΊ

+

M]

1/2

* 1 + Δ

L^J ~^r^

...

(4)

where Ν is the molecular number density, k is the Boltzmann constant,

Τ is the temperature in °K, Δ is a small correction term for higher ap­ proximations, rm is the position of minimum potential for interaction between the ion and molecule, and fit i',i) * i s the first order collision inte­ gral.fi(1 · 1 ) * values were tabulated as a function of the reduced temperature T* = kT/t, where e is the depth of the potential well at rm, and core di­ ameter a* = a/rm, where a is the rigid core diameter. The results showed that the effect of the added rigid core to the mobility was marked. Their model gave a good fit between theoretical and experimental mobilities for both clustered and unclustered ions. Although the theory of Mason and coworkers represented by Equation 4 is the most generally applicable and accurate theory of ionic mobility cur­ rently developed, it still has several deficiencies. Calculated mobilities are too high for small ions and too low for very large ions. This is because rm and α values, used to calculate values fore andS^ 1,1 **, are difficult to esti­ mate with reasonable accuracy. These values can only be estimated by con­ sidering ionic and molecular size ob­ tained from available viscosity data or consideration of ion size by sum­ mation of bond lengths, ring sizes, and other molecular parameters. Also the theory predicts that mobility varies as Λ / 1 / Τ , but no temperature dependence is observed in PC for the mobilities of large ions; and where temperature dependence is observed, i.e., for small ions as Cl~, HsO + , and (H 2 0)NO+, the mobility increases with increasing temperature instead of decreasing as theory predicts. Thus, there is need for the theory to be further extended and refined to

Figure 7. Two-min scan of positive mobility spectrum of Δ 9 -ΤΗΟ showing single MH+ ion produced at 205°C. First scan is of reactant ions only. Sample in­ jected past midpoint of second scan

explain behavior at PC conditions. The great sensitivity of PC to trace compounds arises because the ionmolecule reactions occur at atmo­ spheric pressure, where many mil­ lions of collisions of reactant ions with trace gas molecules are possible. This means that the probability of detecting all of the product ions formed from the trace amount of sample is high. As the collision fre­ quency is increased by several orders of magnitude over studies conducted at lower pressures, a terminal ion dis­ tribution is established, which is rep­ resentative of the gas composition under investigation. Furthermore, the mobility spectra of ions are far supe­ rior to spectra measured at lower pressure owing to the suppressed

transversal diffusion of ions which causes an increasing spectral band width with decreasing pressure. The high operating pressure of PC results in narrow, gaussian-shaped ionic peaks. The PC instrument measures the time required for a given charged species to migrate a fixed distance. Vd in Equation 1 can be replaced by cell length, d, over drift time, τ, - = KE (5) τ Solving Equation 5 for Κ and cor­ recting to 760 torr and 273°K results in Equation 6 which is applied to ex­ perimental data obtained with the PC instrument. Λ

Figure 6. Correlation curve of K u vs. ionic mass for aliphatic and aromatic compounds containing oxygenated groups Experimental Ko value for ionic peak in mobility spectrum of A M e t r a h y d r o cannabinol gives approximate molecular weight of 300 for ion, indicating ion to be M H +. Exceptional mobility behavior of halogen ions also shown

716 A • ANALYTICAL CHEMISTRY, VOL. 46, NO. 8, JULY 1974

°

_

τΕ

760

Τ

{b

>

K0, the reduced mobility, converts mobility measurements to a common base of temperature, pressure, and electric field strength and provides the best parameter to use for plots of mobility spectra. Ionic mobility depends upon many factors. These include ionic charge, mass, and size of the ions and molec­ ular mass, size, and polarizability of the drift gas. In the current theoreti­ cal treatment, ionic size is an impor­ tant parameter for determining theo­ retical mobility values, but this pa­ rameter is often difficult to establish accurately, particularly if clustering occurs. Correlation plots of mobility values vs. ionic mass alone have been shown to be accurate to only ±20% (33). Kane has measured the ionic mobilities for different series of many organic compounds by PC and has plotted mobility vs. mass curves for

each of the different classes (19). A typical curve of this type is shown in Figure 6 for compounds containing oxygenated groups. Such curves for different classes form a reasonably close and parallel family of curves, the halogenated compounds being an exception. Although this type of curve is unsuitable for predicting ac­ curate ionic masses, it can be quite useful for spectral interpretation ; for example, to determine whether a par­ ticular compound forms a molecular ion, an ion cluster, or fragment ions. Figure 6 illustrates this use for the compound Δ ^tetrahydrocannabinol. This molecule with a molecular weight of 314 gives a single positive ion peak in its spectrum with a K0 mobility of 1.06. This corresponds to an ionic mass of 300 from the plot in Figure 6. This value is within 5% of the molecular weight, indicating the ion to be a molecular type, most probably MH+. One also finds that the voluminous experimental and theoretical work on ion-molecule reactions has been done at low pressures. In spite of a large pressure difference and the fact that many ion-molecule reactions are com­ plex and are functions of concentra­ tion, temperature, and pressure, es­ tablished ion-molecule theory and data can be used for guidance in in­ terpreting PC results. Primarily, one must take into account the differ­ ences owing to pressure conditions; for example, the higher PC pressure can lead to simpler results by the for­ mation of collisionally stabilized ions that may not be observed at lower pressure. The ion-molecule work most directly related to PC is found in CIMS, and it appears that this can be relied upon for use in interpreta­ tion of PC data. The abundance of the positive reactant ions shown in Figure 2 is a function of temperature and water concentration and is the same species whether air or nitrogen is used as the carrier gas. It appears that even though formation of the reactant ions occurs through a complicated series of competing reactions, it is possible to postulate a series of well-estab­ lished reaction steps beginning with 60-keV electrons, nitrogen, and water to explain formation of the stable reactant ions observed. This has been done with some experimental verifi­ cation by Griffin et al. (33), Karasek and Denney (34), and Huertas et al. (35). Analytical Aspects and Applications The analytical strength of PC lies in its ability to provide both a posi­ tive and negative mobility spectrum containing qualitative information for trace amounts of an organic com-

Figure8. Comparative response of FID and plasma chromatograph used as GC detector for sample of 1 0 ~ 1 0 gram of C22H26O3 (resmethrin) from peanut oil extract. Equal amounts of GC efflu­ ent enter each detec­ tor. PC instrument tuned to MH + ion in positive spectrum of compound

pound. All compounds studied have exhibited intense positive mobility spectra. Although only those under­ going electron attachment reactions produce a negative mobility spec­ trum, the absence of a negative spec­ trum is still significant analytically. Generally, the intensity of the posi­ tive spectrum will exceed that of the negative, although the relative inten­ sity of the positive-to-negative spec­ trum is a specific characteristic of the compound involved. For some drugs (Figure 7) the positive spectrum is far more intense than the negative; for trinitrotoluene the reverse is true. The PC technique is strictly limit­ ed to the introduction of trace quan­ tities of a compound. The quantities detectable and identifiable lie be­ tween 10~ 8 to 10 _ 14 gram. A sample of 10 6 gram or greater saturates the instrument, completely consumes the reactant ions, and gives rise to the formation of cluster ions and complex reactions between the product ions themselves so that the changing spec­ tral pattern with sample concentra­ tion is uninterpretable. This problem is well illustrated by Metro and Kel­ ler in their attempt to interpret PC data for alkyl ethers after introduc­ tion of 10" 4 gram samples (36). A sample quantity of 10~ 8 gram or less and the presence of some unreacted reactant ions are essential to obtain valid, reproducible PC data (41). The PC instrumentation is in an early stage of development, making it primarily suitable for research stud-

718 A • ANALYTICAL CHEMISTRY, VOL. 46, NO. 8, JULY 1974

ies. Almost no quantitative applica­ tions have been reported. As with any ultra trace method, much refinement of the sampling technique, analytical procedure, sample inlet design, and instrumentation design is needed to consistently give quantitative results of acceptable precision and accuracy. Most of the PC work reported has been concerned with the broad quali­ tative aspects of the method, in which the types of mobility spectra produced by different classes of com­ pounds, limits of detectability, and possible analytical procedures are ex­ plored. The present instrumentation produces good, reproducible qualita­ tive data. The qualitative studies made have included detection and identification of such compounds as n-alkyl alcohols (11, 37), oxygenated organic compounds (12, 38), substi­ tuted aromatics (10, 13, 16), rc-alkanes (8), alkyl halides (9), nitrotoluenes (17), musk ambrette (39), polychlorinated biphenyls (40), aliphat­ ic N-nitrosamines (41), and pesticides (42). In these studies a sample of 10 8 gram or less of a pure compound is introduced via a platinum wire coat­ ed with a l-μΐ methanol solution of the compound after the solvent had evaporated. Interpretation of the mo­ bility spectra is accomplished by ob­ servation of coincident mobilities of positive and negative ions, occurrence of coincident ions in homologous se­ ries, and relating PC results to the stable ionic species observed in CIMS

Figure 9. Normalized plots of positive product ion intensities vs. Ko for series of substituted hexane c o m pounds illustrate type of reference spectra being developed. Only 1-bromohexane produces negative mobility spectrum of compounds shown

and positive and negative ion mass spectrometry. Even with ambiguities, these spectral interpretations increase our basic knowledge of PC, although it is the characteristic pattern in the spectra of each compound that is more important and useful analytically. The ability of the plasma chromatograph to function as a GC detector with identification capabilities is obvious and has been explored (4, 6, 9, 39, 43, 44). The instrument is more sensitive than the flame ionization detector (Figure 8) (45), and is also able to detect a specific compound in the presence of a large concentration of another, such as a solvent (6). This selectivity and the use of the wire sample injection technique have led to a demonstration of its ability to function as a detector for liquid chromatography (42). Although the feasibility of the GC/PC detector system is clear, a number of developments are necessary before any widespread use can come about. Reference mobility spec-

Figure 10. Normalized mobility spectra of positive reactant ions and TNT product ions observed with nitrogen carrier gas at 193° and 152°C. Reactant ions change mobility with temperature; product TNT ions do not

tra of many compounds must be produced (Figure 9); only a few are now available (9, 46). The ion-drift spectrometer produces mobility spectra of low resolution. One approach to resolution improvement is to apply computerized deconvolution of the essentially gaussian ion peaks observed for a single ion to unresolved peak pairs. This approach has been used to separate the ions of the halogen isotopes in PC spectra (47). The most important development needed to make the PC method practical for GC detection is to develop an instrumental package of reasonable cost and simplicity, specifically designed as a GC detector with higher resolution obtained instrumentally. Although interfacing to a GC via a splitter is straightforward, a major problem to be solved is reduction of comparatively large inlet surfaces exposed to the sample prior to ionization on which compound absorption and peak tailing occur. This can be minimized in the present design by operation at 200°C, but this limits

flexibility and applicability to thermally unstable compounds. Detection of trace amounts of trinitrotoluene in air is a currently relevant and difficult analytical problem that has been explored by PC with favorable results (17, 48). The consensus is that the most viable approach to detection of hidden explosives in either civilian or military situations is through a trace analysis for the vapors evolved by the chemicals of the explosive. Such trace analyses stretch the detection limits and selectivity capabilities of most established methods, which generally involve lengthy laboratory procedures with a preconcentration step from a large air sample that is followed by such instrumental methods as GC with ECD, mass spectrometry, or GC/MS. In the case of T N T detection, the vapors evolving from the compound in equilibrium at room temperature place only 10~ 9 gram in each 100 ml of air directly over the sample. Because this quantity represents that present before al-

A N A L Y T I C A L CHEMISTRY, VOL. 46, NO. 8, JULY 1974 • 719 A

lowing for losses by air convection and from container barriers and ab­ sorptions at other interfaces, the de­ tection capability of a useful instru­ ment must be in the 10~ l a to 10~ 14 gram range. As a basis for development of a monitoring instrument for TNT va­ pors in air, PC appears to match the requirements well. T N T gives strong, characteristic positive and negative mobility spectra (Figure 10). The most abundant ion of both spectra is the M~ molecular ion. Direct detec­ tion of T N T in air is facilitated by the use of an air carrier gas, which produces a mixture of (Η2θ) η θ2~ ions and electrons as the negative reactant species. These give essential­ ly the same negative mobility seen in Figure 9 for a nitrogen carrier. By tuning the plasma chromatograph to the M~ ion peak, the sensitivity at­ tainable for T N T in air for 10~ 9 gram injection of T N T displays a signal/ noise ratio greater than 1000, giving detectability in the picogram range. It is quite possible to design a PC in­ strument as a compact, simple moni­ tor of air with a sensitivity greater than that demonstrated by these data. Only the surface of the potential of the PC method has been explored. Largely unstudied are the many car­ rier gases other than nitrogen and air (gases like NO, HC1, NH 3 , C 0 2 , and SO2) that either alone or as additives to nitrogen could greatly enhance the selectivity and sensitivity of the method. The promise of PC for ultratrace analysis and identification of compounds akin to that already es­ tablished for the competing technique of GC/MS remains to be developed. The advantage of simple, atmospher­ ic pressure instrumentation is consid­ erable, but that instrumentation re­ quires much development in areas of improved resolution, electronics, and system development, both for units of more general capabilities and for units of simple, specific uses. Perhaps the most fruitful application of PC will be its interfacing with the highly developed GC/MS/computer system where the large capacity computing capability can provide both positive and negative mobility spectra in par­ allel with the mass spectra to give data that can greatly aid the inter­ pretation for compound identifica­ tion.

References (1) P. G. Simmonds, D. C. Fenimore, B. C. Pettitt, J. E. Lovelock, and A. Zlatkis, Anal. Chem., 39,1428 (1967). (2) E. W. McDaniel, "Collision Phenome­ na in Ionized Gases," pp 457-61, Wiley, New York, N.Y., 1964. (3) F. W. Karasek, Res./Develop., 21 (3), 34(1970).

(4) F. W. Karasek, Int. J. Environ. Anal. Chem., 2, 157 (1972). (5) F. W. Karasek, "Plasma Chromatog­ raphy," 1973 McGraw-Hill Yearbook of Science and Technology, McGraw-Hill, New York, N.Y., 1973. (6) F. W. Karasek, Can. Res. Develop., 6 (4), 22 (1973). (7) M. S. B. Munson, Anal. Chem., 43 (13), 28A(1971). (8) F. W. Karasek, D. W. Denney, and E. H. DeDecker, ibid., 46, in press (1974). (9) F. W. Karasek, 0 . S. Tatone, and D. W. Denney, J. Chromatogr., 87, 137 (1973). (10) F.W. Karasek and O . S . Tatone, Anal. Chem., 44, 1758 (1972). (11) F.W. Karasek and D . M . K a n e , J. Chromatogr. Sci., 10, 673 (1972). (12) O. S. Tatone, MS thesis, University of Waterloo, Waterloo, Ont., Canada, June 1973. (13) F. W. Karasek and D. M. Kane, Anal. Chem., 46, 780 (1974). (14) D. W. Denney, MS thesis, University of Waterloo, Waterloo, Ont., Canada, April 1974. (15) E. D. Pellizzari, "Electron Capture/ Gas-Liquid Chromatography—A Tech­ nical Review of the Literature," RTI No. 43V-817, Research Triangle Institute, Research Triangle Park, N.C. 27709. (16) F. W. Karasek, O. S. Tatone, and D. M. Kane, Anal. Chem., 45, 1210 (1973). (17) F. W. Karasek and D. W. Denney, J. Chromatogr., 93,141 (1974). (18) F. W. Karasek and D. M. Kane, Anal. Chem., 45,576(1973). (19) D. M. Kane, PhD thesis, University of Waterloo, Waterloo, Ont., Canada, July 1974. (20) E. W. McDaniel and E. A. Mason, "The Mobility and Diffusion of Ions in Gases," Wiley, New York, N.Y., 1973. (21) P. Langevin, Ann. Chim. Phys., 28, 289(1903). (22) P. Langevin, ibid., 5, 245 (1905). (23) H. R. Hassé, Phil. Mag., 1, 139 (1926): (24) D. Enskog, PhD dissertation, University of Uppsala, Uppsala, Sweden, 1917. (25) S. Chapman, Phil. Trans. Roy. Soc, A216, 279(1916). (26) S. Chapman, ibid., A217, 279 (1917). (27) T. Kihara, Rev. Mod. Phys., 24, 45 (1952). (28) T. Kihara, ibid., 25, 844 (1953). (29) S. Chapman and T. G. Cowling, "The Mathematical Theory of Non-Uniform Gases," 2nd éd., Cambridge University Press, London, England, 1952. (30) E. A. Mason and H. W. Schamp, Ann. Phys. (New York), 4, 233 (1958). (31) P. L. Patterson, J. Chem. Phys., 53, 696(1970). (32) E. A. Mason, H. O'Hara, and F. J. Smith, J. Phys. B., 5, 169 (1972). (33) G. W. Griffin, I. Dzidic, D. I. Carroll, R. N. Stillwell, and E. C. Horning, Anal. Chem., 45, 1204 (1973). (34) F. W. Karasek and D. W. Denney, ibid., 46,633(1974). (35) M. L. Huertas, A. M. Marty, J. Fontan, I. Alet, and E. Duffa, Aerosol Sci., 2,145(1971). (36) M. M. Metro and R. A. Keller, J. Chromatogr. Sci., 11, 520 (1973). (37) F. W. Karasek, M. J. Cohen, and D. I. Carroll, ibid., 9, 390 (1971). (38) F. W. Karasek, W. D. Kilpatrick, and M. J. Cohen, Anal. Chem., 43, 1441 (1971). (39) F.W. Karasek and R.A.Keller, J. Chromatogr. Sci., 10,626(1972). (40) F. W. Karasek, Anal. Chem., 43, 1982 (1971). (41) F.W. Karasek and D. W. Denney, ibid., 46, in press (1974). (42) F. W. Karasek and D. W. Denney, Anal. Lett., 6 (11), 993 (1973). (43) M. J. Cohen and F. W. Karasek, J. Chromatogr. Sci., 8, 330 (1970).

720 A • ANALYTICAL CHEMISTRY, VOL. 46, NO. 8, JULY 1974

(44) S. P. Cram and S. Ν. Chester, ibid., 11,391 (1973). (45) Application Report G-427, Franklin GNO Corp., West Palm Beach, Fla. 33402, November 1973. (46) F. W. Karasek and D . M . K a n e , J. Chromatogr., 93, 129(1974). (47) F. W. Karasek, R. J. Laub, and E. DeDecker, ibid., ρ 123. (48) F. W. Karasek, Res./Develop., 25 (5), 32(1974).

Francis W. Karasek is a professor of chemistry at the University of Water­ loo in Waterloo, Ontario. He received his BS degree in 1942 from Elmhurst College and his PhD degree in 1952 from Oregon State University. For the next 16 years he led an automatic analytical instrumentation develop­ ment group at the Phillips Petroleum Research Center in Oklahoma. That work has resulted in many analytical instruments in direct control of chemical processes. In 1968 Dr. Kar­ asek joined the faculty at Waterloo where he does research and teaching in analytical instrumentation, GC/MS systems, surface analysis in­ strumentation, and plasma chroma­ tography. As an editorial associate of the Research/Development magazine, he writes a monthly article on new analytical instrumentation. He origi­ nated the present Graduate Fellow­ ship program of the ACS Analytical Division and is Councilor for the Di­ vision. Dr. Karasek is a member of the Editorial Board οι Analytical Let­ ters and is also chairman of the American Society of Mass Spectrom­ etry (ASMS) Committee V on New Instrumentation and of the ASTM Subcommittee E-19. He is also the father of seven sons, four of whom are active in the fields of science and medicine.